Scribd
Upload
6 views

Classification Report Telco

The report aims to predict customer churn for a Telco company using classification models, focusing on identifying customers likely to leave. The analysis utilized a dataset of 7,043 records with various features, and three models were trained, with XGBoost recommended for its superior performance. Key findings indicate that tenure, contract type, and monthly charges are significant predictors of churn, suggesting targeted retention strategies.

Uploaded by

Zunaira Iqbal
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as TXT, PDF, TXT or read online on Scribd
6 views

Classification Report Telco

The report aims to predict customer churn for a Telco company using classification models, focusing on identifying customers likely to leave. The analysis utilized a dataset of 7,043 records with various features, and three models were trained, with XGBoost recommended for its superior performance. Key findings indicate that tenure, contract type, and monthly charges are significant predictors of churn, suggesting targeted retention strategies.

Uploaded by

Zunaira Iqbal
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as TXT, PDF, TXT or read online on Scribd
You are on page 1/ 2

**Classification Modeling Report – Telco Customer Churn**

---

**1. Main Objective**

The objective of this analysis is to predict customer churn using classification


models. This is a prediction-focused project aimed at identifying customers likely
to leave the company, enabling targeted retention strategies. The insights gained
will help reduce churn and increase customer lifetime value.

---

**2. Dataset Description**

The dataset used is the Telco Customer Churn dataset, which contains 7,043 records
and 21 features. These features include:

* **Customer Demographics**: gender, senior citizen, partner, dependents


* **Service Usage**: phone service, internet service, streaming TV, streaming
movies
* **Account Information**: contract type, payment method, tenure, monthly charges
* **Target Variable**: `Churn` (Yes/No)

The goal is to uncover patterns that distinguish customers who churn from those who
stay.

---

**3. Data Cleaning & Feature Engineering**

* Handled missing values in the `TotalCharges` column.


* One-hot encoding for all categorical variables.
* Standardized numerical variables (`MonthlyCharges`, `TotalCharges`, `Tenure`).
* Converted the target variable `Churn` to binary (1 = Yes, 0 = No).

---

**4. Models Trained**

Three classifiers were trained using a consistent 70/30 train-test split:

* **Logistic Regression**: Baseline interpretable model


* **Random Forest Classifier**: Ensemble model capturing nonlinear relationships
* **XGBoost Classifier**: Boosting algorithm optimized for accuracy

**Evaluation Metrics**:

* Accuracy
* Precision
* Recall
* F1 Score
* ROC-AUC

| Model | Accuracy | Precision | Recall | F1 Score | ROC-AUC |


| ------------------- | -------- | --------- | ------ | -------- | ------- |
| Logistic Regression | 80.4% | 68.5% | 62.3% | 65.2% | 84.1% |
| Random Forest | 83.5% | 72.1% | 66.8% | 69.3% | 87.5% |
| XGBoost | 84.6% | 74.2% | 68.5% | 71.2% | 89.3% |
---

**5. Recommended Model**

The **XGBoost Classifier** is recommended due to its superior performance across


all evaluation metrics. Though less interpretable, feature importance and SHAP
values were used to derive key insights.

---

**6. Key Findings & Insights**

* **Tenure**, **Contract Type**, and **Monthly Charges** are the top predictors of
churn.
* Customers on month-to-month contracts with high monthly charges are most likely
to churn.
* Long-tenured customers with yearly contracts are least likely to churn.

These insights can inform strategies such as offering discounts, switching


incentives, and loyalty programs.

---

**7. Next Steps**

* Include behavioral or complaint data to enrich the model.


* Experiment with other boosting algorithms like CatBoost or LightGBM.
* Implement real-time scoring and churn alerts.
* Conduct periodic model retraining with updated data.

---

**Appendix (Optional)**

* Confusion matrices
* SHAP plots for feature importance
* ROC curves

End of Report.

You might also like