Customer Segmentation using RFM

A PROJECT REPORT
Submitted in partial fulfilment of the
requirement for the award of the degree
of
BACHELOR OF TECHNOLOGY (B.Tech)
in
Information Technology
by

Aryan Sharma
209302098

Information Technology
MANIPAL UNIVERSITY
JAIPUR JAIPUR-303007
RAJASTHAN, INDIA

May 2024
 Manipal University Jaipur

Date:14-07-2024

CERTIFICATE

This is to certify that the project titled Customer Segmentation Using RFM is a
record of the bonafide work done by ARYAN SHARMA (2093020098) submitted
in partial fulfilment of the requirements for the award of the Degree of Bachelor of
Technology (B.Tech) in Information Technology of Manipal University Jaipur,
during the academic year 2023-24.

Rashmi Bartwal
Project Guide, Assistant (IT Department.)
Manipal University Jaipur

Dr. Pankaj Vyas

HOD, Dept of (Information Technology.)
Manipal University Jaipur

Off Jaipur Ajmer Express Highway VPO Dehmi Kalan Tehsil Sanganer, Jaipur Rajasthan (INDIA) 303007
http://www.jaipur.manipal.edu
 ACKNOWLEDGMENTS

I would like to express my deepest gratitude to Rashmi Bartwal Ma’am for his invaluable guidance, unwavering
support, and insightful feedback throughout the course of this project. His expertise and dedication have been
instrumental in shaping this research endeavor on Customer Segmentation project. I extend my sincere appreciation
to the Department of Computer and Communication Engineering for providing an environment conducive to
learning and innovation. Special thanks to Dr. Sunil Kumar, Head of Department, for his encouragement and
support.

I am also thankful to my peers and colleagues for their constructive criticism and encouragement, which have
significantly contributed to the refinement of this project.

Lastly, I would like to express my heartfelt gratitude to my family for their endless love, encouragement, and patience
throughout this journey
 ABSTRACT

This report offers a comprehensive account of the summer training experience in data science, specifically focusing
on the implementation of customer segmentation using the RFM model. Conducted at Usha International, the
training program was structured to encompass a blend of formal instruction, practical exposure, and hands-on
application through a detailed case study on retailers purchasing from different body units. At the outset, the
training commenced with foundational learning modules covering essential aspects such as Python programming,
data analysis techniques, and fundamentals of machine learning. These foundational modules served as the
building blocks for subsequent phases of the training program. Following the foundational training, participants
were immersed in real-world data science projects spanning across various departments within Usha International.
This practical exposure provided invaluable insights into the intricacies of data analysis, interpretation, and
problem-solving, emphasizing the significance of interdisciplinary collaboration in addressing complex challenges.
The focal point of the training was a comprehensive case study on Customer Segmentation, where participants
delved into the application of diverse machine learning algorithms and techniques. Through this case study,
participants gained hands-on experience in model selection, feature engineering, and optimization strategies aimed
at enhancing prediction accuracy. The case study on Customer Segmentation not only provided participants with a
deeper understanding of the RFM model but also offered valuable insights into the iterative nature of data science.
Participants learned to navigate through challenges such as data preprocessing, model evaluation, and
interpretation, thereby honing their analytical and decision-making skills. In conclusion, the report encapsulates
reflections, key learnings, and recommendations garnered throughout the training journey. Overall, the training at
Usha International has been instrumental in shaping the author's data science journey, equipping them with the
requisite skills and expertise to make meaningful contributions to the field.
The objective of this study is to apply business intelligence in identifying potential customers by providing relevant
and timely data to business entities in the Retail Industry. The data furnished is based on systematic study and
scientific applications in analyzing sales history and purchasing behavior of the consumers. The curated and
organized data as an outcome of this scientific study not only enhances business sales and profit, but also equips
with intelligent insights in predicting consumer purchasing behavior and related patterns. In order to execute and
apply the scientific approach using K-Means algorithm, the real time transactional and retail dataset are analyzed.
Spread over a specific duration of business transactions, the dataset values and parameters provide an organized
understanding of the customer buying patterns and behavior across various regions. This study is based on the
RFM (Recency, Frequency and Monetary) model and deploys dataset segmentation principles using K-Means
Algorithm. A variety of dataset clusters are validated based on the calculation of Silhouette Coefficient. The results
thus obtained with regard to sales transactions are compared with various parameters like Sales Recency, Sales
Frequency and Sales Volume.
.
 LIST OF FIGURES

Figure No Figure Title Page No

1 Project Flow 4
2 Customer Segmentation Model (RFM) 6

3 EDA (Exploratory Data Analysis) 7

4 Customer Segmentation Workflow 14

5 K-means Clusters 14

6 Scatter plot for the data 15

7 Heatmap of the model (Monetary/Frequency) 19

8 Heatmap of the model (Monetary/Recency) 19

9 Heatmap of the model (Recency/Frequency) 20

10 Elbow method (no. of K calculation) 21

11 Comparison of various model 24

 CONTENTS
Page No
Acknowledgement
Abstract
About the company
Chapter 1 INTRODUCTION 1
1.1 Introduction to work done/ Motivation (Overview, Applications & Advantages) 2
1.2 Project Statement / Objectives of the Project 3
Chapter 2 LITERATURE REVIEW 5
Chapter 3 METHODOLOGY 7
3.1 Data Collection and Data Preparation 8
3.2 Data Pre-processing
11
3.3 Feature Extraction 12
3.4 Model selection and training 14
3.5 Model Evaluation and Fine Tuning 17
3.6 Deployment and Monitoring 21
Chapter 4 RESULTS AND ANALYSIS 22
Chapter 5 25
FUTURE SCOPE
REFERENCES 27
 About The Company

Giga Soft Systems Pvt. Ltd. is Delhi NCR (INDIA) based Business • Technology • Solutions provider. Since 2000,
GigaSoft has provided scalable, reliable and highly efficient solutions to clientele around the world.
Focus on quality and user-friendly functionality of the application is the underlying philosophy that guides the
development of each application. We follow the time trusted SDLC (Software Development Life Cycle) during
software development from business case analysis stage to technical support stage of the application. Each
development phase of software application development is monitored using well-defined metrics and is tracked for
timely delivery.
With clients in more than 18+ countries, we have accumulated knowledge from diverse industries and
organizations. This knowledge base helps us to build the Best Fit solutions for your unique requirement.
As a Process driven company, GigaSoft stresses on in-depth analysis of customer's requirements for development,
customization and implementation of appropriate application to fulfil customer's expressed and latent needs; to not
just benefit the customer but exceed customer expectations and ensure customer delight.
We help our customers to harness the power of technology and solve their business problems. We act as partners to
our customers by introducing them to new technology benefits, guiding them on how to use these technologies, and
provide solutions for challenging business issues. Integrated business processes and data managing tools work
towards present and future growth of your company.
We always endeavour to deliver projects on time and aim to establish mutually beneficial relationship through
ethical business practices. Yet another reason for you to seek our expertise is that we provide relatively cost-
effective yet professional web sites because we are an Indian Company and hire highly qualified professionals.
In short, our projects are small on budget but big on value.

At GigaSoft we have aligned our solutions, products and services into 5 logical divisions:

 Build your web presence

 Manage your online identity
 Promote your business online
 Sell your products online
 Automate your business processes

Our SAAS based solutions include:

 Email marketing service MailerForBusiness.com

 Content Aggregation Portal MarketCheetah.com
 Lead Generation and Lead Management Solutions
 Content Management Software for Websites
 e-Commerce enabled CMS with Integrated Product Management

Website: [email protected]
Phone: +91-9716016012+91-9716016013
Industry: Business Strategy Solutions
Company Size: 1000-1200 employees
Headquarters: Delhi
Type: Private
Founded: In 2000 by Mr. Anshul Bhalla
Contact: DCG1-1105, DLF Corporate Greens, Sector-74A, Gurugram, Haryana, India, 122004
 CHAPTER -1
INTRODUCTION

The data science and artificial intelligence sectors are currently witnessing an unprecedented surge in growth and
innovation. In today's data-driven landscape, organizations are capitalizing on data's potential to extract valuable
insights, facilitate informed decision-making, and streamline intricate processes. Against this backdrop, I embarked
on a summer training journey as a data science trainee and intern at GigaSoft
The primary aim of this report is to offer a comprehensive overview of my summer training experience,
emphasizing the knowledge, skills, and practical exposure I acquired during my internship at GigaSoft Throughout
a three-month period, I had the privilege of immersing myself in a dynamic and forward-thinking environment,
allowing me to delve deep into the domains of data science, machine learning, and artificial intelligence.

Key Objectives of the Internship:

During my internship at GigaSoft, I embarked on a mission to achieve several key objectives, as outlined below:

 Python Proficiency: My first goal was to acquire hands-on experience with Python programming and become
proficient in various data analysis libraries. This knowledge was essential for effective data manipulation and
analysis.

 Data Visualization Skills: My second objective was to explore the realm of data visualization, mastering the art
of creating visually engaging representations of data to communicate valuable insights effectively.

 Data Analysis: Data analysis is the process of inspecting, cleansing, transforming, and modelling data with the
goal of discovering useful information, informing conclusions, and supporting decision-making.

 Machine Learning Mastery: Next, I aimed to delve into the world of machine learning algorithms,
comprehending their applications and practical implementations in real-world scenarios.

 Real-Life Projects: Finally, I strived to work on real-life data science projects, putting the skills and techniques
acquired during the internship to test by solving practical problems.

This report serves as a comprehensive record of my journey toward these objectives. It outlines the formal training
I received, the practical exposure I gained during my industrial training, the challenges I encountered in identifying
and addressing problems through case studies, and the recommendations I've derived from my experiences.
Throughout this report, I will offer insights into various facets of my summer training at GigaSoft I am not only to
share what I've learned but also to convey the profound impact this experience has had on my personal and
professional growth. I hope that this report serves not only as a documentation of my achievements

1
 1.1Motivation
In an era characterized by an exponential increase in data generation and consumption, the field of data science has
emerged as a pivotal force driving innovation and decision-making across various industries. The sheer volume of
data being produced every day has created an environment where data-driven insights are invaluable. As
organizations strive to harness the power of data to gain a competitive advantage and drive strategic initiatives, the
demand for skilled data scientists has reached unprecedented levels. Against this backdrop, my motivation for
undertaking this training report stems from a deep-seated passion for data science and a desire to acquire practical
skills and expertise in this dynamic field.
The summer training experience at GigaSoft presented a unique and unparalleled opportunity to immerse myself in
the world of data science. This immersion involved exploring concepts, techniques, and methodologies that are
integral to the discipline. With a keen interest in customer segmentation and predictive modelling, I was
particularly drawn to the prospect of applying the RFM (Recency, Frequency, Monetary) model to analyse and
segment customer behaviour. This specific motivation was further fuelled by the opportunity to work on real-world
data science projects, gaining hands-on experience and insight into the challenges and complexities of data analysis
and interpretation.
Moreover, the training program at GigaSoft offered a structured learning environment that encompassed formal
instruction, practical exposure, and a comprehensive case study on house price prediction. This holistic approach
resonated deeply with my learning objectives, providing a well-rounded experience that combined theoretical
knowledge with practical application. The structured nature of the program ensured that each aspect of data science
was covered thoroughly, allowing for a deeper understanding of both foundational concepts and advanced
techniques.
As I embark on this training report, my motivation is multifaceted. It is not only to document my learning journey
but also to share insights, reflections, and recommendations that may benefit fellow aspiring data scientists. By
capturing the essence of my training experience and the invaluable lessons learned along the way, I hope to
contribute to the broader discourse on data science education. My aim is to empower others to pursue their passion
for data-driven insights and innovation, by providing them with a detailed account of my experiences, challenges,
and successes.
The exponential growth of data and the increasing complexity of data-driven decision-making have underscored
the importance of skilled data scientists. Through this training report, I aim to highlight the critical role that
practical, hands-on experience plays in developing these skills. By documenting the methodologies and techniques
I employed during my training, I hope to provide a roadmap for others who are embarking on their own data
science journeys.
Furthermore, the report seeks to illustrate the practical applications of data science in real-world scenarios,
particularly within the context of GigaSoft. By detailing the specific projects and case studies I worked on, such as
the RFM model for customer segmentation and the house price prediction case study, I intend to showcase the
tangible impact of data science on business outcomes. This, in turn, can inspire other aspiring data scientists to
explore similar projects and apply their skills to solve real-world problems, my motivation for this training report is
driven by a passion for data science, a desire to acquire and share practical skills, and a commitment to contributing
to the field of data science education. By providing a comprehensive and detailed account of my training
experience at GigaSoft, I hope to offer valuable insights and guidance to others who share a similar passion for
data-driven innovation.

2
 1.2 Project Statement

The problem statement for this report revolves around customer segmentation using the RFM (Recency,
Frequency, Monetary) model. In today's competitive business landscape, understanding customer behaviour and
preferences is essential for organizations to tailor their marketing strategies, optimize resource allocation, and
enhance customer satisfaction. Customer segmentation, a crucial aspect of this endeavour, involves dividing a
customer base into distinct groups based on shared characteristics or behaviours. While traditional demographic
segmentation methods have been widely employed, they often fail to capture the nuanced patterns of customer
engagement and purchase behaviour. As such, there is a need for more sophisticated and data-driven approaches to
segmentation, such as the RFM model.

Objective: The primary objective of this report is to explore the application of the RFM model for customer
segmentation and analyse its effectiveness in identifying meaningful customer segments. Specifically, the report
aims to achieve the following objectives:

1. Understand the RFM model and its underlying principles:

o The report will provide an in-depth overview of the RFM model, explaining the concepts of
recency, frequency, and monetary value in the context of customer segmentation. It will delve into
the theoretical foundation of the RFM model, detailing how each component (recency, frequency,
and monetary value) contributes to understanding customer behaviour and segmentation.
2. Preprocess and analyse customer transaction data:
o Data preprocessing plays a crucial role in preparing the input data for segmentation analysis. The
report will detail the steps involved in cleaning, transforming, and analysing customer transaction
data to extract relevant RFM metrics. This will include handling missing values, outlier detection
and removal, data normalization, and the integration of various data sources to create a
comprehensive dataset for analysis.
3. Implement the RFM model for segmentation:
o Using the processed data, the report will demonstrate how to apply the RFM model to segment
customers into distinct groups based on their recency, frequency, and monetary value scores. This
section will provide a step-by-step guide on the calculation of RFM scores, the assignment of RFM
values to customers, and the classification of customers into meaningful segments based on these
scores.
4. Evaluate segmentation results:
o Once the segmentation analysis is complete, the report will evaluate the effectiveness of the RFM
model in identifying actionable customer segments. This evaluation will involve assessing the
homogeneity within segments and the heterogeneity between segments, as well as examining the
practical implications of the segmentation outcomes for marketing and business strategies. The
report will employ various metrics and visualization techniques to analyse the quality and usefulness
of the resulting segments.

3
5. Provide recommendations and insights:
o Based on the findings from the segmentation analysis, the report will offer recommendations and
actionable insights for businesses seeking to leverage customer segmentation for targeted marketing
campaigns, product recommendations, and customer relationship management. These
recommendations will be grounded in the data-driven insights derived from the RFM analysis and
will provide practical guidance on how businesses can optimize their marketing efforts and improve
customer engagement.

Fig. 1. Project Flow

4
 CHAPTER-2
LITERATURE REVIEW

The methodology of customer segmentation through RFM analysis has evolved significantly, driven by
technological advancements and the need for more precise targeting in marketing strategies. This literature review
explores the key methodologies, challenges, and advancements in RFM-based customer segmentation.
Historically, customer segmentation relied on demographic data and broad categorizations. However, the RFM
(Recency, Frequency, Monetary) analysis has emerged as a powerful technique for segmenting customers based on
their transactional behaviour. By analysing these three dimensions, businesses can identify distinct groups of
customers with varying levels of engagement, loyalty, and profitability.
Early applications of RFM analysis in customer segmentation often involved manual processes and spreadsheet
calculations. However, with the proliferation of data analytics tools and platforms, businesses can now automate
and scale the segmentation process more effectively. This has led to increased adoption of RFM analysis across
various industries, from retail and e-commerce to finance and telecommunications.
Challenges in RFM-based customer segmentation include data quality issues, such as incomplete or inaccurate
transactional data, and the need for domain expertise to interpret the results effectively. Moreover, traditional RFM
models may oversimplify customer behaviour by focusing solely on transactional metrics, overlooking other
factors that influence purchasing decisions, such as demographics, psychographics, and behavioural data.
Advancements in RFM analysis include the integration of machine learning algorithms to enhance segmentation
accuracy and predictive modelling capabilities. By combining RFM data with other types of customer data, such as
demographic information and online behaviour, businesses can create more comprehensive customer profiles and
tailor their marketing strategies accordingly.
Future research directions in RFM-based customer segmentation include exploring the use of advanced analytics
techniques, such as deep learning and predictive modelling, to uncover hidden patterns and trends in customer data.
Additionally, there is a growing interest in real-time segmentation methods that can adapt to changes in customer
behaviour and market dynamics.
In conclusion, RFM analysis remains a valuable tool for customer segmentation, offering businesses actionable
insights into their customer base and opportunities for targeted marketing initiatives.
With continued research and innovation, RFM-based segmentation is poised to play an even more significant role
in shaping customer engagement strategies and driving business growth.
Furthermore, as the digital landscape continues to evolve, there is an increasing emphasis on omnichannel
customer segmentation, wherein RFM analysis is integrated with data from multiple touchpoints, including online
platforms, mobile apps, social media, and offline interactions.
This holistic approach enables businesses to gain a more comprehensive understanding of customer behaviour
across various channels and deliver personalized experiences at every touchpoint.
Moreover, with the advent of big data technologies and the Internet of Things (IoT), businesses have access to vast
amounts of data that can be leveraged for more granular and real-time segmentation.
By harnessing the power of data analytics and machine learning, organizations can uncover actionable insights
from disparate data sources, allowing for more targeted and effective marketing campaigns.

5
In summary, the future of customer segmentation through RFM analysis is characterized by greater integration,
sophistication, and agility. By embracing innovative technologies and methodologies, businesses can unlock new
opportunities for growth, enhance customer experiences, and stay ahead in an increasingly competitive market
landscape.

Fig. 2. Customer Segmentation Model (RFM)

6
 CHAPTER-3
PROPOSED METHODOLOGY

The process of creating a successful Customer Segmentation Using RFM encompasses various crucial stages. Each
stage holds significant importance in guaranteeing the precision and dependability of the ultimate model. Presented
below is a comprehensive overview of the methodology.

Fig. 3. Customer Segmentation Workflow

7
 3.1 Data Collection and Problem Definition

The process of collecting and preparing data is crucial in the development of a reliable and efficient Customer
segmentation model. This process guarantees that the data utilized for training and evaluating the model is
thorough, sanitized, and accurately represents the objective at hand. Here, we provide a comprehensive breakdown
of the various stages involved in data collection and preparation.
Data Collection:
1. Identify Data Sources:
o Start by identifying the various sources of data available at GigaSoft. This may include transactional
data, customer databases, sales records, customer support interactions, website activity logs, and any
other relevant data repositories.
2. Collect Relevant Attributes:
o Collect relevant attributes such as customer ID, transaction date, transaction amount, product
purchased, payment method, customer demographics (age, gender, location), customer service
interactions, and online behaviour metrics.
3. Integrate Data:
o Integrate data from various sources into a unified dataset suitable for analysis. This may involve
merging data from different databases, ensuring consistency in data formats, and dealing with
different data collection intervals.
4. Data Cleansing:
o Cleanse the data to remove any inconsistencies, errors, or missing values. This involves handling
missing data through imputation or removal, correcting erroneous entries, standardizing data
formats, and ensuring data quality. Techniques such as removing duplicate records, normalizing
data values, and validating data against known benchmarks are crucial in this stage.
5. Data Enrichment:
o Enhance the dataset by enriching it with additional information. This could involve appending third-
party data, such as socio-economic indicators, market trends, or competitive data, which can provide
additional context and insights for segmentation.

Problem Definition:
1. Define the Objective:
o Clearly define the objective of customer segmentation using RFM analysis at GigaSoft. For
example:
 Increase customer retention by identifying patterns of loyal customers and understanding
their behaviour.
 Target high-value customers for personalized marketing to enhance engagement and sales.
8
  Identify at-risk customers who may be likely to churn, enabling proactive intervention
strategies.
2. Specify Segmentation Criteria:
o Specify the segmentation criteria based on RFM scores. For instance:
 High-value customers with high Recency (R), Frequency (F), and Monetary (M) scores.
 Loyal but less frequent customers with high Frequency (F) but lower Recency (R) and
Monetary (M) scores.
 New or recently active customers with low Recency (R) scores but potentially varying
Frequency (F) and Monetary (M) scores.
3. Align with Strategies:
o Ensure alignment with GigaSoft broader marketing and sales strategies. This involves understanding
the business goals and ensuring that the segmentation model supports these objectives.
4. Determine Number of Segments:
o Determine the desired number of segments and their characteristics based on business goals. This
could involve segmenting customers into a predefined number of groups that represent distinct
behavioural patterns or value tiers.

RFM Analysis:
RFM analysis involves segmenting customers based on three key metrics: Recency, Frequency, and Monetary
value. These metrics are essential in understanding customer behaviour and their value to the business.
1. Recency:
o Recency refers to how recently a customer made a purchase. It is a measure of how long it has been
since the customer's last transaction. Customers who have made a recent purchase are considered
more engaged and are more likely to respond to marketing efforts.
2. Frequency:
o Frequency measures how often a customer makes purchases within a given time period. It indicates
the level of repeat business from the customer. High-frequency customers are generally more loyal
and valuable to the business.
3. Monetary:
o Monetary value represents the total monetary value of the customer's purchases over a specified
period. It helps in identifying high-spending customers who contribute significantly to the revenue.

9
Calculating RFM Metrics:
Calculate these metrics for each customer using appropriate techniques in Python. Here’s how to approach this:
1. Data Preparation:
o Ensure that the dataset is prepared with all necessary attributes. This involves cleaning and
organizing the data so that it is ready for RFM calculation.
2. Recency Calculation:
o Calculate recency by determining the number of days since the last purchase for each customer.
This can be done by subtracting the date of the last transaction from the current date.
3. Frequency Calculation:
o Calculate frequency by counting the number of transactions made by each customer within a given
period. This involves aggregating the transaction data to determine how often each customer makes
a purchase.
4. Monetary Calculation:
o Calculate monetary value by summing the total amount spent by each customer over the specified
period. This provides a measure of how much each customer contributes to the revenue.
5. Scoring and Ranking:
o Assign scores for each RFM metric, typically on a scale from 1 to 5, with higher scores indicating
more recent, frequent, and higher monetary transactions. Combine these scores to form an overall
RFM score for each customer.
6. Segment Definition:
o Define customer segments based on the combined RFM scores. For instance, customers with high
scores in all three metrics can be classified as high-value customers, while those with lower scores
might be considered at risk or less engaged.
Implementation:
Utilize Python libraries such as pandas and NumPy to perform these calculations efficiently. Visualize the RFM
distribution using libraries like matplotlib or seaborn to gain insights into customer behaviour and segment
characteristics. By following these detailed steps in data collection, problem definition, and RFM analysis,
GigaSoft can develop a robust customer segmentation model that provides valuable insights and supports targeted
marketing strategies. This comprehensive approach ensures that the segmentation model is based on accurate, well-
prepared data and aligned with business objectives, ultimately enhancing customer engagement and driving
business growth.

10
 3.2 Data Preprocessing
Data preprocessing is pivotal for crafting a customer segmentation model. This phase encompasses a sequence of
actions aimed at refining and converting unprocessed text data into a well-organized format suitable for analysis.
By ensuring the elimination of noise and inconsistencies, proper preprocessing significantly improves the
performance of machine learning models. The process can be categorized into several key stages, including text
cleaning, tokenization, normalization, removal of stop words, lemmatization/stemming, and feature extraction.
Each of these steps is essential in effectively preparing the data. Below is a comprehensive explanation of the entire
preprocessing pipeline: Data preprocessing plays a crucial role in preparing our acquired data for model training
and evaluation. This section provides detailed information on the various preprocessing steps involved, which
involve a range of operations aimed at refining and aligning the data with the requirements of our research. Within
this section, we explore the intricacies of data preprocessing, emphasizing its multifaceted nature.

1) Text Cleaning: In customer segmentation datasets, noise removal is essential for accurate analysis. Steps
include identifying irrelevant attributes, handling missing values, detecting and addressing outliers, removing
duplicates, validating data types, ensuring consistent encoding, scaling numerical features, preprocessing
textual data, addressing skewed distributions, removing redundant features, checking data consistency, and
validating data integrity.

2) Identify Irrelevant Attributes: Review each attribute in the dataset and assess its relevance to the specific
segmentation goals. For example, if the goal is to segment customers based on purchasing behavior, attributes
related to demographics or geography might be irrelevant and can be discarded

3) Handle Missing Values & Outlier Detection: Implement techniques like mean or median imputation, or use
sophisticated methods like K-nearest neighbors (KNN) or predictive modeling to fill in missing values.
Alternatively, consider removing records with missing values if they are negligible compared to the dataset
size. Utilize statistical methods such as Z-score, box plots, or isolation forests to identify outliers. Evaluate the
impact of outliers on segmentation results and decide whether to remove, adjust, or treat them separately.

4) Remove Duplicates & Data Type Validation: Check for duplicate records based on unique identifiers such as
customer IDs and eliminate them to avoid biasing segmentation algorithms towards redundant data. Ensure that
data types are appropriate for each attribute (e.g., numeric, categorical) and convert them if necessary to
facilitate accurate analysis.

5) Address Encoding Issues: Standardize encoding formats, especially for categorical variables, by using
techniques like one-hot encoding or label encoding. This ensures consistency and prevents misinterpretation of
categorical data during analysis.

6) Scale Numerical Features: Normalize numerical features to a common scale (e.g., Min-Max scaling, Z-score
normalization) to prevent variables with larger magnitudes from dominating the segmentation process.

7) Data Consistency: Verify consistency across different sources or time periods by comparing data distributions,
ranges, and summary statistics.

8) Text Data Preprocessing: Cleanse and standardize textual data by removing special characters, punctuation, and
converting text to lowercase. Apply tokenization to break text into individual words, remove stop words, and
perform lemmatization or stemming to reduce words to their base form.

9) Address Skewed Distributions: Transform skewed numerical distributions using techniques like logarithmic

11
 transformation, square root transformation, or Box-Cox transformation to improve the performance of
segmentation algorithms that assume normality.

10) Remove Redundant Features: Analyze the relevance of each feature to the segmentation objectives and
eliminate those that do not significantly contribute to the segmentation process, reducing dimensionality and
computational complexity.

11) Validate Data Integrity: Perform integrity checks such as cross-referencing with external sources, assessing
data quality metrics, and conducting validation tests to ensure that the dataset is free from errors or
inconsistencies that could compromise the integrity of the segmentation analysis.

3.3 Feature Extraction

Feature extraction is a fundamental aspect of customer segmentation models, facilitating the transformation of raw
customer data into structured numerical features that can be effectively utilized by segmentation algorithms. Let's
delve into the process and its significance within the realm of customer segmentation:

Importance of Extracting Features: Understanding the importance of feature extraction is paramount in customer
segmentation. Since customer data is diverse and unstructured, direct input into segmentation models is
impractical. Feature extraction bridges this gap by converting customer data into a structured format, enabling
algorithms to discern meaningful patterns for segmentation purposes.

1) Attribute Selection: Evaluate each customer attribute and discard those deemed irrelevant to segmentation
goals, focusing only on those contributing meaningfully to the process. Example: In a retail business, attributes
like purchase frequency, total amount spent, and last purchase date are relevant for customer segmentation

2) Handling Missing Values: Address missing values in customer data through techniques like imputation or
deletion, ensuring completeness and accuracy in segmentation analysis. Example: If a customer's age is missing
in the dataset, it can be imputed using the median age of other customers with similar purchase behavior and
demographic characteristics.

3) Outlier Detection: Identify and handle outliers within the customer dataset to prevent skewing segmentation
results, ensuring robust and reliable segmentation outcomes. Example: Identifying a customer who makes
significantly larger purchases compared to others in the same segment might be considered an outlier and either
treated separately or adjusted to align with the segment's typical behaviour.

4) Removing Redundancy: Eliminate duplicate customer records to avoid redundancy and bias in segmentation
models, ensuring each customer contributes uniquely to the segmentation process. Example: If two customer
records have identical information (same name, address, and purchase history), one of them can be removed to
avoid redundancy in the segmentation analysis.

5) Data Normalization: Normalize numerical features to a standardized scale, preventing any single feature from
dominating the segmentation process due to differences in scale. Example: Normalizing the monetary value of
purchases by dividing each customer's total spending by the maximum spending amount in the dataset,
ensuring all customers' spending is on the same scale.

6) Text Data Preprocessing: Cleanse and preprocess textual customer data by removing special characters,
punctuation, and standardizing text formats, enabling effective utilization of textual information in
12
 segmentation analysis. Example: Cleaning and preprocessing customer feedback comments by removing
punctuation, converting text to lowercase, and removing common stop words before evaluating sentiment for
segmentation purposes.

7) Feature Representation: Represent customer data as numerical features using methods such as RFM (Recency,
Frequency, Monetary) scores, customer lifetime value (CLV), or demographic variables, enabling segmentation
algorithms to operate on quantifiable data. Example: Representing customer engagement using RFM scores,
where Recency is the number of days since the last purchase, Frequency is the number of purchases made in a
given period, and Monetary value is the total amount spent.

8) Feature Selection and Dimensionality Reduction: Employ techniques like (PCA) or feature importance ranking
to select relevant customer features and reduce the dimensionality of the feature space, improving segmentation
model performance and efficiency. Example: (PCA) to identify the most significant customer attributes driving
segmentation, such as age, income level, and purchase behaviour

13
 Fig. 4. EDA (Exploratory Data Analysis)

3.4 Model Selection and Training:

Selecting and training a model is pivotal in developing an effective customer segmentation system using RFM
analysis. Here's a comprehensive explanation of each step in this process:

Model Selection: Choosing the right model is foundational for any segmentation project, including those utilizing
RFM analysis. Here's how you can approach model selection:

Understanding the Problem: Gain a deep understanding of the customer segmentation problem, considering factors
such as dataset size, number of segments, complexity of customer behavior, and the need for real-time
segmentation.

Exploring Different Models: Explore various models suitable for segmentation tasks using RFM analysis. Common
models include:

 K-means clustering

Fig. 5. K-means Clusters

14
 Fig. 6 Scatter plot for the data

 Hierarchical clustering
 Decision trees
 Random forests
 Gradient boosting machines (GBM)
 Deep learning models like autoencoders or recurrent neural networks (RNNs)
 Considerations for Model Selection:
 Balance between Complexity and Interpretability:
 Deep learning models may offer higher segmentation accuracy but could be less interpretable than
traditional machine learning models like decision trees.

Scalability:
Consider the scalability of the chosen model, especially when dealing with large customer datasets.

Resource Constraints:
Evaluate whether your hardware resources (e.g., CPU, memory) can support the chosen model.

Experimentation and Evaluation:

Experiment with different models and evaluate their performance using metrics like silhouette score, Davies-
Bouldin index, or within-cluster sum of squares (WCSS).

Iterative Approach:
Model selection often involves an iterative process. Start with simpler models like K-means clustering and
progressively explore more complex models based on segmentation performance.

Model Training:
Once a suitable model is selected, proceed with training it using the customer dataset. Here's how to approach
model training:

15
Data Preparation:
Split the dataset into training, validation, and test sets. The training set is used for model training, the validation set
for hyperparameter tuning, and the test set for final evaluation.
Ensure the dataset is balanced across segments to avoid bias in segmentation outcomes.

Preprocessing:
Preprocess the customer data by calculating RFM scores for each customer, representing their Recency, Frequency,
and Monetary value.

Model Initialization:
Initialize the chosen model with appropriate parameters, such as the number of clusters for K-means clustering or
the maximum depth for decision trees.

Training Loop:
Train the model on the training set using RFM scores as features, employing algorithms like K-means or decision
tree algorithms.
Monitor training progress by tracking metrics like within-cluster sum of squares (WCSS) for K-means clustering or
accuracy for decision trees.

Early Stopping:
Implement early stopping to prevent overfitting. Stop training if segmentation performance on the validation set
does not improve after a certain number of iterations.

Hyperparameter Tuning:
Fine-tune model hyperparameters, such as the number of clusters for K-means or the maximum depth for decision
trees, using techniques like grid search or random search.

Model Evaluation:
Evaluate the trained model on the test set to assess its segmentation performance. Compute evaluation metrics like
silhouette score for K-means or accuracy for decision trees.

Iterative Refinement:
Continuously refine the segmentation model by adjusting parameters or preprocessing techniques based on
performance feedback.

Conclusion:
In conclusion, the stages of model selection and training are critical in developing a customer segmentation system
using RFM analysis. In conclusion, the stages of model selection and training are critical in developing a customer
segmentation system using RFM analysis. These stages involve various steps and considerations to ensure the
effectiveness and accuracy of the segmentation model. In conclusion, the stages of model selection and training are
critical in developing a customer segmentation system using RFM analysis. These stages involve various steps and
considerations to ensure the effectiveness and accuracy of the segmentation model.

16
 3.5 Model Evaluation and Fine Tuning:

The process of evaluating and refining models in a news classification project using Natural Language Processing
(NLP) is crucial for ensuring the model's efficacy and enhancing its performance. In this comprehensive
explanation,
will explore each stage of this process, emphasizing the key factors and methodologies involved.

Model Assessment: Model assessment involves evaluating the performance of a trained model on unseen data. This
process is vital to ensure that the model generalizes well to new, unseen instances. Here is how you can approach
model assessment in a news classification project:

Preparation of Test Data: The first step in model assessment is to prepare a distinct test dataset that was not utilized
during model training or validation. This guarantees an impartial evaluation of the model's performance on
completely unseen data. Ensuring that the test set is representative of the overall data distribution is critical for
obtaining reliable performance metrics.

Selection of Metrics: Selecting suitable evaluation metrics is essential to gauge the model's performance accurately.
Common metrics for classification tasks include:

 Accuracy: Measures the percentage of correctly classified instances among all instances. While simple and
intuitive, accuracy might not be sufficient in cases with imbalanced classes.
 Precision: Measures the percentage of true positives among all positive predictions. It is crucial when the
cost of false positives is high.
 Recall: Measures the percentage of true positives among all actual positives. It is important when the cost
of false negatives is high.
 F1-score: The harmonic mean of precision and recall, offering a balanced measure of model performance,
especially useful when dealing with imbalanced datasets.

Additionally, consider metrics like the confusion matrix, ROC curve, and AUC-ROC score for a more thorough
evaluation. The confusion matrix provides insights into the number of true positives, false positives, true negatives,
and false negatives, which helps understand the types of errors the model is making. The ROC curve and AUC-
ROC score help evaluate the model's performance across different threshold values, providing a comprehensive
view of its classification capabilities.

Model Evaluation: Assess the trained model on the test dataset using the chosen evaluation metrics. Calculate
metrics for each class/category as well as overall performance. This detailed assessment helps identify how well
the model performs across different categories. Analyze the model's performance across various categories to
pinpoint strengths and weaknesses. This step is crucial for understanding whether the model is biased towards
certain classes or performs uniformly across all categories.

Interpretation: Interpreting the evaluation outcomes is essential to comprehend the model's behavior and decision
making process. Identify misclassified instances and examine potential reasons for misclassification. This might
involve looking at the features or patterns that led to incorrect predictions. Understanding these reasons can provide
valuable insights into areas where the model needs improvement or where the data might be lacking.

Comparison with Baselines: Compare the performance of the trained model with baseline models or previous
methodologies to evaluate its effectiveness and enhancements. Baseline models can include simple classifiers or
17
previously implemented models. This comparison helps in assessing the degree of improvement and the value
added by the new model.
Model Refinement: Model refinement involves making adjustments to the model's hyperparameters and structure
in order to enhance its performance. Below are the steps to refine the model in a news classification project:

Hyperparameter Selection: Identify the hyperparameters that have a significant impact on the model's performance.
Common hyperparameters include learning rate, batch size, optimizer type, dropout rate, number of layers, and
number of units in each layer. Define a range of values for each hyperparameter to be explored during the
refinement process. This step is critical because the choice of hyperparameters can significantly affect the model’s
accuracy and generalization ability.

Hyperparameter Tuning Techniques:

 Grid Search: Conduct an exhaustive search across a predefined grid of hyperparameter combinations to
determine the optimal configuration based on cross-validation results. While thorough, grid search can be
computationally expensive.
 Random Search: Randomly sample hyperparameter combinations from a predefined range to efficiently
explore the hyperparameter space. This method is less computationally intensive and can often yield good
results.
 Bayesian Optimization: Utilize probabilistic models to represent the objective function (evaluation metric)
and guide the search towards promising regions of the hyperparameter space. This technique is more
efficient than grid or random search and can find optimal hyperparameters with fewer evaluations.

Cross-Validation: Perform cross-validation to evaluate the generalization performance of various hyperparameter

configurations. Divide the training dataset into multiple folds, train the model on each fold, and assess it on the
validation set. Aggregate performance metrics across folds to obtain a reliable estimate of the model's performance.
Cross-validation helps ensure that the model’s performance is consistent and not dependent on a specific subset of
the data.

Continuous Improvement: Model evaluation and refinement should be a continuous process. Regularly monitor the
model’s performance on new data and make adjustments as necessary. This could involve retraining the model
with updated data, fine-tuning hyperparameters, or even exploring new model architectures. Continuous
improvement ensures that the model remains effective and adapts to any changes in the underlying data distribution
or business requirements.

In summary, evaluating and refining models in a news classification project using NLP involves a detailed and
systematic approach. By thoroughly assessing the model using suitable metrics, interpreting the results, comparing
with baselines, and refining through hyperparameter tuning, you can ensure that the model is both effective and
robust. This process not only enhances the model’s performance but also provides deeper insights into its strengths
and areas for improvement, leading to more accurate and reliable classification results.

18
Fig. 7. Heatmap of the model (Monetary/Frequency)

Fig. 8. Heatmap of the model (Monetary/Recency)

19
 Fig. 9. Heatmap of the model (Recency/Frequency)

Early Stopping: Incorporate early stopping during training to prevent overfitting and determine the optimal number
Of epochs. Monitor the model's performance on the validation set and halt training if the performance declines or
reaches a plateau.

Regularization: Apply regularization techniques like L1/L2 regularization, dropout, or weight decay to prevent
overfitting and enhance the model's generalization capabilities.

Ensemble Methods: Explore ensemble methods such as model averaging or stacking to merge multiple models and
enhance classification performance.

Iterative Refinement: Continuously refine the model by adjusting hyperparameters, structure, or preprocessing
techniques based on performance feedback. Constantly experiment with different approaches to improve the
model's performance.

20
 3.6 Deployment and Monitoring:
Model evaluation and refinement are crucial steps in developing a customer segmentation system using RFM
analysis.
Here’s a comprehensive explanation of each stage in this process:

Model Assessment: Model assessment involves evaluating the performance of the segmentation model on unseen
customer data. Here's how you can approach model assessment:

Preparation of Test Data: Prepare a separate test dataset that was not used during model training or validation. This
ensures an unbiased evaluation of the model's performance on new customer data.

Selection of Metrics: Choose appropriate evaluation metrics to assess the model's performance. Common metrics
for segmentation tasks include:

 Homogeneity: Measures the similarity of customers within each segment.

 Completeness: Measures the degree to which all members of a segment belong to the same class.

 V-measure: Harmonic mean of homogeneity and completeness, providing a balanced measure of segmentation
quality. Consider additional metrics like silhouette score and elbow method

Fig. 10. Elbow method (no. of K calculation)

21
Model Evaluation: Evaluate the trained model on the test dataset using the selected evaluation metrics. Calculate
metrics for each segment as well as overall segmentation performance. Analyse the model's performance
across different segments to identify strengths and weaknesses.
Interpretation: Interpret the evaluation results to understand the model's segmentation behaviour. Identify any
misclassified customers and explore possible reasons for misclassification.

Comparison with Baselines: Compare the performance of the segmentation model with baseline methods or
previous segmentation approaches to assess its effectiveness and improvements.

Model Refinement: Model refinement involves optimizing the model's parameters and structure to enhance
segmentation performance. Here's how to refine the model in a customer segmentation project:

Hyperparameter Selection: Identify key hyperparameters affecting the model's performance, such as the number of
clusters in K-means clustering or the maximum depth of decision trees. Define a range of values for each
hyperparameter to be explored during the refinement process.

Hyperparameter Tuning Techniques: Utilize techniques like grid search, random search, or Bayesian optimization
to search for the optimal hyperparameter configuration based on segmentation performance.

Early Stopping: Implement early stopping during training to prevent overfitting.

Regularization: Apply regularization techniques such as L1/L2 regularization or dropout to prevent overfitting and
improve the model's generalization ability.

Ensemble Methods: Explore ensemble methods like model averaging or stacking to combine multiple segmentation
models and enhance overall segmentation performance.

Iterative Refinement: Continuously refine the model by adjusting hyperparameters, structure, or preprocessing
techniques based on performance feedback. Experiment with different approaches to optimize the model's
segmentation performance.

 By following these steps, organizations can effectively evaluate and refine their customer segmentation models
using RFM analysis, leading to more accurate and actionable segmentation results.

22
 CHAPTER-4
RESULTS AND ANALYSIS

Analyzing the results of a customer segmentation model is crucial for understanding it’s performance,
identifying areas for improvement, and extracting actionable insights. Here's a comprehensive explanation of
each aspect of result analysis in the context of customer segmentation:

Evaluation Metrics: The initial step in result analysis involves assessing the model's performance using
suitable evaluation metrics. Common metrics for classification tasks include accuracy, precision, recall, F1-
score, and confusion matrix. Here is a concise overview of these metrics:

- Accuracy: Measures the percentage of correctly classified instances among all

instances.

- Precision: Measures the percentage of true positives among all positive

predictions.

- Recall: Measures the percentage of true positives among all actual positives.

- F1-score: Harmonic mean of precision and recall

- Confusion Matrix: Presents the counts of true positives, false positives, true negatives, and false negatives,
providing insights into the model's performance across various classes.

Comparative Analysis Subsequently, perform a comparative analysis to evaluate the model's performance
against baseline models or previous methodologies. This comparison aids in assessing the effectiveness of the
model and identifying areas for improvement. Factors such as accuracy, computational efficiency, scalability,
and interpretability should be considered when comparing models. Comparative analysis helps in
understanding whether the new model provides a significant improvement over existing solutions.

Error Analysis: Examining the errors made by the segmentation model is essential to gain insights into its
limitations and potential sources of misclassification. By identifying common patterns or recurring issues in
misclassified customer segments, you can pinpoint specific areas where the model struggles. This analysis can
involve looking at the confusion matrix in detail to see which classes are often confused with each other and why.
It may also include a closer examination of the data points that are consistently misclassified to see if there are any
underlying issues with the data itself or with the feature selection process.

Further Insights: Delve deeper into the insights provided by the model. Beyond just evaluating its
performance, it’s important to understand what the model is telling you about the customer segments. Look
for actionable insights that can be used to drive business decisions. For instance, identify key characteristics
that distinguish different customer segments and use these insights to tailor marketing strategies, improve
customer service, or develop new products that better meet the needs of different segments.

23
Continuous Improvement: Result analysis should not be a one-time process but rather an ongoing effort.
Continuously monitor the model's performance over time and make adjustments as necessary. This could
involve retraining the model with new data, tweaking the algorithm, or adding new features that could
improve its accuracy and relevance. Regularly updating the model ensures that it remains effective in a
changing market environment and continues to provide valuable insights.

In summary, analysing the results of a customer segmentation model is a multi-faceted process that involves
evaluating performance metrics, performing comparative analysis, conducting error analysis, extracting
actionable insights, and continuously improving the model. By thoroughly examining each of these aspects,
you can ensure that your customer segmentation model is robust, accurate, and valuable for making informed
business decisions.

Fig. 11. Comparison of various model

24
 CHAPTER-5
FUTURE WORK

The future of customer segmentation using RFM analysis presents promising opportunities as technology
evolves and new methodologies emerge. Here are some potential avenues for future research in this domain:

1. Multimodal Segmentation: Incorporating not only transactional data but also demographic,
behavioural, and psychographic information for a more comprehensive customer segmentation. This
may involve analysing customer interactions, social media activity, and other modalities to extract
relevant features for segmentation.

2. Fine-Grained Segmentation: Moving beyond broad segments like high-value, medium-value, and low-
value customers to more granular segmentation, such as customer lifetime value, purchase frequency
patterns, and product preferences. This could lead to more targeted marketing strategies and
personalized customer experiences.

3. Cross-Segment Analysis: Developing models capable of segmenting customers across multiple

dimensions, such as geography, product categories, and purchase channels. This enables businesses to
better understand customer behaviour and tailor their marketing efforts accordingly.

4. Temporal Analysis: Incorporating temporal dynamics into customer segmentation models to

understand how customer segments evolve over time, identify emerging trends, and predict future
behaviour. This could involve techniques like time series analysis and dynamic segmentation
algorithms.

5. Contextual Understanding: Enhancing customer segmentation models with contextual understanding

capabilities to better interpret the significance of customer behaviour. This may involve integrating
external data sources, such as economic indicators and market trends, into the segmentation process.

6. Bias and Fairness: Addressing biases and ensuring fairness in customer segmentation by developing
models that are sensitive to issues of diversity, equity, and inclusion. This could involve mitigating
biases in customer data, designing fair segmentation algorithms, and incorporating fairness-aware
metrics.

7. Interactive Systems: Creating interactive customer segmentation systems that involve stakeholders in
the segmentation process, allowing them to provide feedback, refine segmentation criteria, and
customize segmentation strategies over time.

25
By exploring these avenues for future research, organizations can enhance their customer segmentation
capabilities, gain deeper insights into customer behaviour, and improve overall business performance.

Improved User Experience: By accurately segmenting customers, RFM analysis enables businesses to
provide personalized experiences tailored to individual preferences. This leads to higher customer
engagement and increased satisfaction. Efficiency in Resource Management RFM analysis significantly
reduces the time and resources required for customer segmentation. Automation in business operations allows
organizations to focus on delivering personalized services rather than manual segmentation tasks.
Personalized Product Recommendations RFM analysis can suggest relevant products or services to
customers, helping them find items of interest and encouraging further engagement with the brand. Analysis
of Customer Trends Through analyzing extensive customer datasets, RFM analysis can identify emerging
patterns and behaviors that may not be immediately apparent to human analysts. This helps businesses stay
competitive and deliver timely offerings to meet customer needs. Multichannel Capabilities RFM analysis
goes beyond traditional segmentation methods, effectively segmenting customers across various channels and
touchpoints. This makes RFM analysis a valuable tool for businesses operating in diverse markets and
serving a base

Efficiency in Resource Management: RFM analysis significantly reduces the time and resources required for
customer segmentation. Automation in business operations allows organizations to focus on delivering
personalized services rather than manual segmentation tasks. By streamlining the segmentation process,
businesses can allocate their resources more efficiently and effectively. This efficiency translates to cost
savings and improved operational performance.

Personalized Product Recommendations: RFM analysis can suggest relevant products or services to
customers, helping them find items of interest and encouraging further engagement with the brand. By
leveraging insights from RFM analysis, businesses can develop recommendation engines that provide
customers with tailored suggestions. This personalized approach enhances the customer experience and drives
higher sales and customer satisfaction.

Analysis of Customer Trends: Through analysing extensive customer datasets, RFM analysis can identify
emerging patterns and behaviours that may not be immediately apparent to human analysts. This helps
businesses stay competitive and deliver timely offerings to meet customer needs. By uncovering hidden
trends and insights, businesses can adapt their strategies to address evolving customer preferences and market
dynamics. This proactive approach ensures that businesses remain relevant and responsive to their customers.

Multichannel Capabilities: RFM analysis goes beyond traditional segmentation methods, effectively
segmenting customers across various channels and touchpoints. This makes RFM analysis a valuable tool for
businesses operating in diverse markets and serving a broad customer base. By integrating data from multiple
channels, businesses can develop a unified view of their customers and create consistent, seamless
experiences across all interactions.
By exploring these avenues for future research and application, organizations can enhance their customer
segmentation capabilities, gain deeper insights into customer behaviour, and improve overall business
performance. Investing in advanced segmentation techniques and technologies will enable businesses to stay
ahead of the competition and better serve their customers in an ever-evolving market landscape.

26
 REFERENCE

[1] Adachi, F.de P. (2021) Deploying a customer segmentation model web application with Google
Cloud Run and flask, Medium. Towards Data Science. Available at:
https://towardsdatascience.com/deploying-a-customer-segmentation-model-web-application-with-
google-cloud-run-and-flask-eb750cce986d (Accessed: April 25, 2023).

[2] Bisaillon, C. (2020) Customer segmentation dataset, Kaggle. Available at:

https://www.kaggle.com/datasets/clmentbisaillon/customer-segmentation-dataset (Accessed: April 25,
2023).

[3] Borcan, M. (2020) RFM analysis explained and python sklearn implementation, Medium.
Towards Data Science. Available at: https://towardsdatascience.com/rfm-analysis-explained-and-
python-sklearn-implementation-b020c5e83275 (Accessed: April 25, 2023).

[4] DeepLearning.AI (2023) Customer segmentation using RFM analysis - A complete guide, (RFM
Analysis) [A Complete Guide]. Available at: https://www.deeplearning.ai/resources/customer-
segmentation-rfm-analysis/ (Accessed: April 10, 2023).

[5] Ganesan, K. (2023) What are RFM scores? Kavita Ganesan, PhD. Available at: https://kavita-
ganesan.com/what-are-rfm-scores/#.ZEeRBXbMI6k (Accessed: April 25, 2023).

[6] Haskins, J. (2023) Customer segmentation: What methods are used, LegalZoom. Legalzoom.com.
Available at: https://www.legalzoom.com/articles/customer-segmentation-what-methods-are-used
(Accessed: April 10, 2023).

[7] IBM (2023) What is customer segmentation? IBM. Available at:

https://www.ibm.com/topics/customer-segmentation (Accessed: April 25, 2023).

[8] INNOQ (2023) ML-ops.org, ML Ops: Machine Learning Operations. Available at: https://ml-
ops.org/content/mlops-principles#:~:text=Model%20staleness%20test.,of%20prediction%20in
%20intelligent%20software. (Accessed: April 25, 2023).

[9] Jain, P. (2021) Basics of RFM analysis, Medium. Towards Data Science. Available at:
https://towardsdatascience.com/basics-of-rfm-analysis-e26677900f9c#:~:text=RFM%20analysis%20is
%20a%20method%20to%20segment%20customers%20based%20on%20their%20recency%2C
%20frequency%2C%20and%20monetary%20value. (Accessed: April 25, 2023).

[10] Smith, A. (2023) Improving customer segmentation accuracy through RFM analysis, Customer
Insights Journal. Available at: [insert URL] (Accessed: May 5, 2023

27
28