RESQBITES

A Food Surplus Redistribution System

NAME: PAWAN ALANKAR

USN:1BM22CS191
NAME: VARSHA PRASANTH
USN:1BM22CS321
NAME: TANMAYI S BALIJA
USN:1BM22CS359
NAME: RUTH MARY PAUL
USN:1BM22CS360

GUIDE:SOWMYA T
DESIGNATATION: ASSISTANT PROFESSOR

Department of Computer Science & Engineering

B.M.S. College of Engineering
 INTRODUCTION
Food Waste Crisis:
India wastes 67 million tons of food annually.
Millions face hunger and malnutrition (189 million undernourished people).
Key Initiatives:
Organizations like Feeding India, Robin Hood Army, and No Food Waste rescue
surplus food.
Surplus food from restaurants, hotels, and events redistributed to shelters,
orphanages, and low-income areas.
Impact Goals:
Reduce food waste and environmental harm.
Combat hunger and improve nutrition for vulnerable populations.
Challenges:
Logistics and timely food collection before spoilage.
Building infrastructure and raising awareness for large-scale redistribution.
ARTIFICIAL INTELLIGENCE IN FOOD CHEMISTRY: PREDICTIVE MODELING AND ANALYSIS FOR QUALITY

YEAR PUBLISHED: 2022 JOURNAL: IJFANS International Journal of Food and

Nutritional Sciences

PROBLEM STATEMENT: TECHNIQUES USED:

Challenges in predictive modeling for food quality, sensory Predictive modeling with machine learning (SVMs,
analysis, food safety, supply chain optimization, and neural networks, RF).
personalized nutrition. AI-based quality control using image recognition.
AI for personalized nutrition analysis.

PROS OF METHODOLOGY: CONS & LIMITATIONS:

High predictive accuracy. Data quality and availability issues.
Improved efficiency and automation. AI model interpretability challenges.
Versatility in applications (quality control, supply High computational resource needs.
chain, nutrition). Lack of standardization in AI protocols.

DATASET USED: RESULTS OBTAINED:

Large datasets representing ingredients, processing methods, RF achieved the highest accuracy (95%), demonstrated lower
and environmental factors (no specific datasets mentioned). computational cost, and improved quality control and defect
detection.
DEMAND FORECASTING FOR FOOD PRODUCTION USING MACHINE LEARNING ALGORITHMS: A CASE
STUDY OF UNIVERSITY REFECTORY

YEAR PUBLISHED: 2023 JOURNAL: Technical Gazette

PROBLEM STATEMENT: TECHNIQUES USED:

Addressing challenges in predicting food demand for Machine Learning algorithms: ANN, GPR, SVR, Regression
university refectories to reduce food waste. Trees, and Ensemble Decision Trees

PROS OF METHODOLOGY: CONS & LIMITATIONS:

High accuracy in forecasting using diverse machine learning Computationally intensive due to the complexity of multiple
algorithms. models.
Helps reduce food waste by predicting demand without pre- Limited to data from a single university refectory, reducing
booking. generalizability.
Provides a decision support tool for cafeteria management, Challenges in adapting the models to other non-cafeteria
improving daily, weekly, and monthly food planning. environments without additional customization​

DATASET USED: RESULTS OBTAINED:

2241 sales records from Mersin University EDT Boosted model performed best with
refectory MSE=0.51, MAE=0.50, and R=0.96​
 ARTIFICIAL INTELLIGENCE FOR REDUCING FOOD WASTE

YEAR PUBLISHED: 2021 JOURNAL: Fraunhofer Research News

PROBLEM STATEMENT: TECHNIQUES USED:

Reducing food waste in Germany through AI, especially AI and machine learning for sales and production
in production processes planning optimization

PROS OF METHODOLOGY: CONS & LIMITATIONS:

Reduces overproduction and prolongs the shelf life of Limited to industries such as dairy, meat, and bakery,
perishable food items. restricting broader applicability.
AI optimizes processes across the entire food value chain, High initial investment in AI and sensor technology may deter
from farm to retail. smaller businesses.
Uses real-time data to minimize quality-related food loss The complexity of integrating AI systems with existing
during production. production processes​

DATASET USED: RESULTS OBTAINED:

Study-specific dataset, part of a larger project involving Expected reduction in food waste by optimizing
multiple stakeholders production processes using AI​
HARNESSING ARTIFICIAL INTELLIGENCE FOR URBAN FOOD REDISTRIBUTION: A SOCIO-TECHNICAL
ANALYSIS OF THE FEEDING AMERICA INITIATIVE

YEAR PUBLISHED: 2024 JOURNAL: IJRASET

PROBLEM STATEMENT: TECHNIQUES USED:

Using AI and autonomous systems to redistribute AI-driven deep learning, self-driving vehicles,
surplus food to homeless populations and reduce automated vending food hubs
food waste

PROS OF METHODOLOGY: CONS & LIMITATIONS:

Tackles food waste and insecurity with AI-driven High setup costs for AI, vehicles, and hubs.
redistribution. Regulatory and public acceptance hurdles for autonomous
Efficient system using deep learning for prediction and vehicles.
routing. Relies on large datasets, limiting effectiveness in small
Offers tax benefits and CSR incentives for participating areas.
businesses.

DATASET USED: RESULTS OBTAINED:

U.S.-based, uses sales and weather data to predict Using AI and autonomous systems to redistribute surplus food
surplus food availability to homeless populations and reduce food waste
 ENHANCING FOOD INTEGRITY THROUGH ARTIFICIAL INTELLIGENCE AND MACHINE LEARNING: A
COMPREHENSIVE REVIEW​

YEAR PUBLISHED: 2024​ JOURNAL:MDPI Applied Sciences​

PROBLEM STATEMENT: TECHNIQUES USED:

Addressing food safety and integrity using AI and ML applications in quality control, fraud
AI/ML to mitigate risks in the food supply chain​ detection, risk assessment, and traceability​

PROS OF METHODOLOGY: CONS & LIMITATIONS:

• Enhanced food safety • Requires high computational resources
• Improved supply chain resilience • Implementation costs
• Real-time hazard detection​4​ • Potential data privacy issues​

DATASET USED: RESULTS OBTAINED:

Data collected from literature and AI/ML models AI/ML led to improved food safety, public health,
applied in food safety and quality​ and consumer trust in food integrity​
 MOTIVATIONS AND CHALLENGES FOR FOOD COMPANIES IN USING IOT SENSORS FOR
REDUCING FOOD WASTE

YEAR PUBLISHED: 2023​ JOURNAL: MDPI,Sustainability​

PROBLEM STATEMENT: TECHNIQUES USED:

Challenges faced by food companies in using Use of IoT sensors to track food conditions (e.g.,
IoT for food waste reduction and improving temperature, humidity) to prevent food spoilage​
efficiency​
PROS OF METHODOLOGY: CONS & LIMITATIONS:
• Improves food quality • Data security concerns
• Increases reliability • Requires significant investment
• Meets legal compliance​ • Trust issues with new technologies​

DATASET USED: RESULTS OBTAINED:

Data gathered from multiple European food IoT technologies reduced food waste and improved monitoring
companies using IoT sensors​ of food conditions during storage and transport​
 IMPROVING FOOD DISTRIBUTION IN URBAN AREAS WITH IOT AND ML

YEAR PUBLISHED: 2021 JOURNAL: International Journal of Food and Nutritional

Sciences (IJFANS), Volume 10, Issue 2

PROBLEM STATEMENT: TECHNIQUES USED:

The growing urban population and demand for food
IoT Sensors for real-time monitoring of food quality and safety
create challenges in efficient, sustainable food
distribution, requiring new solutions

PROS OF METHODOLOGY: CONS & LIMITATIONS:

Improved food quality control and safety Challenges related to data privacy and security
Efficient inventory management reduces food waste Complexities in integrating technologies
Optimized transportation routes lower costs Need for collaboration with suppliers and stakeholders

DATASET USED: RESULTS OBTAINED:

Historical sales data combined with real-time IoT Improvements in food quality control, operational
sensor data efficiency, reduced food waste
FOOD WASTE REDUCTION THROUGH PREDICTIVE MODELING AND AI-DRIVEN INVENTORY MANAGEMENT

YEAR PUBLISHED: 2022 JOURNAL: International Journal of Food and

Nutritional Sciences

PROBLEM STATEMENT: TECHNIQUES USED:

The paper addresses the significant issue of food waste in Data Collection
the supply chain, contributing to environmental degradation predictive modelling
and economic losses.

PROS OF METHODOLOGY: CONS & LIMITATIONS:

Significant reduction in food waste due to accurate
High initial cost for technology implementation.
demand forecasting. Data quality issues can affect model accuracy.
Efficient inventory management and optimization of Challenges in scaling and integrating AI with existing systems.
supply chain logistics.
Improved transportation efficiency and reduced
spoilage.

DATASET USED: RESULTS OBTAINED:

15-18% reduction in food waste in real-world and simulated scenarios.
Historical sales data, seasonal trends, consumer behavior,
Improved demand forecasting accuracy by 25%.
and real-time IoT sensor data.
 APPLYING ARTIFICIAL INTELLIGENCE TO REDUCE FOOD WASTE IN SMALL GROCERY STORES

YEAR PUBLISHED: 2022 JOURNAL: Information Systems in Latin

America (ISLA 2022) Proceedings

PROBLEM STATEMENT: TECHNIQUES USED:

Small grocery stores, especially those focused on natural and The study used machine learning models (including Random
organic foods, face a challenge in balancing food waste Forest, Decision Trees, and Neural Networks) on sales and
reduction with fulfilling demand. combated using AI model climate data. Random forest was applied

PROS OF METHODOLOGY: CONS & LIMITATIONS:

Limited dataset availability led to the need for manual
Significant reduction in food waste (82%) by accurately
feature engineering and external data augmentation.
forecasting demand using RF.
Increased gross profit and minimized food production The complexity of integrating ML models into small
costs. grocery store operations.

DATASET USED: RESULTS OBTAINED:

Daily sales data from a small grocery store in Brazil for one The RF model achieved 90% accuracy in forecasting
year demand.
Food waste reduced
REDUCING FOOD WASTE USING MACHINE LEARNING MODELS: FORECASTING AND OPTIMIZATION
APPROACHES

YEAR PUBLISHED: 2024 JOURNAL: Journal of Data Science & Artificial

Intelligence

PROBLEM STATEMENT: TECHNIQUES USED:

It highlights the need to address inefficiencies in food
Five models (including Random Forest and decision trees)
demand prediction and inventory management across the
were trained, with Random Forest chosen for its superior
supply chain accuracy
PROS OF METHODOLOGY: CONS & LIMITATIONS:
Improved accuracy in demand forecasting, leading to Challenges in data quality and availability, which can
reduced overstock and waste. degrade model performance.
Efficient management of perishable goods by High integration costs, especially for small and medium-
predicting demand and optimizing inventory. sized enterprises (SMEs).
Ethical issues related to data privacy and employment
impact.

DATASET USED: RESULTS OBTAINED:

The paper does not mention specific datasets but The paper cites several studies and real-world applications,
emphasizes the use of large historical sales data for model showing that ML models can reduce forecast errors by up to
training. 30%, cut food distribution waste by 20%
A COMPREHENSIVE REVIEW ON FOOD WASTE REDUCTION BASED ON IOT AND BIG DATA TECHNOLOGIES

YEAR PUBLISHED: 2023 JOURNAL: MDPI,Sustainability

PROBLEM STATEMENT: TECHNIQUES USED:

The study addresses food waste issues and explores how Use of IoT sensors and Big Data analytics to
IoT and Big Data can optimize food quality control. monitor food conditions and optimize logistics
in real-time.

PROS OF METHODOLOGY: CONS & LIMITATIONS:

- Data quality issues: Integrity, availability, and scalability
- Real-time monitoring reduces spoilage risks.
challenges in cloud computing.
- Data-driven decision-making optimizes logistics.
- Lack of standardization in management systems.
- Scalable across the food supply chain.
- Inadequate communication protocols leading to delays or
data loss.

DATASET USED: RESULTS OBTAINED:

No specific dataset. The review is based on various IoT and Big Data technologies can reduce food waste,
studies using IoT and Big Data for food waste reduction. but challenges like data standardization, security risks,
and energy consumption must be addressed.
SMARTNOSHWASTE: USING BLOCKCHAIN, MACHINE LEARNING, CLOUD COMPUTING AND QR CODE TO
REDUCE FOOD WASTE IN DECENTRALIZED WEB 3.0 ENABLED SMART CITIES

YEAR PUBLISHED: 2022 JOURNAL: MDPI Smart Cities

PROBLEM STATEMENT: TECHNIQUES USED:

The study addresses food waste reduction using a combination The system integrates blockchain for traceability, machine
of advanced technologies like blockchain, machine learning, learning for food quality prediction, QR codes for tracking,
cloud computing, and QR codes in smart cities. and cloud computing for managing data.

PROS OF METHODOLOGY: CONS & LIMITATIONS:

- Decentralization: Enables better food waste management - Blockchain Scalability: High data storage and computational
in Web 3.0-enabled smart cities. demands limit scalability.
- Traceability: QR codes ensure the tracking of food items - QR Code Vulnerability: Physical QR codes can be subject to
throughout the supply chain. man-in-the-middle attacks if tampered with.
- Prediction: Machine learning models predict food spoilage - Reinforcement Learning: Q-learning, used for
and optimize distribution. reinforcement learning, can be slow and

DATASET USED: RESULTS OBTAINED:

The study does not focus on a specific dataset but uses - The integration of blockchain, machine learning, cloud
simulated food waste management data within a smart city computing, and QR codes effectively addresses the food
context. waste issue in smart cities
 FOOD REDISTRIBUTION AS OPTIMIZATION

YEAR PUBLISHED: 2021 JOURNAL: University of Colorado at Boulder Technical

Report, CU-CS-1085-11

PROBLEM STATEMENT: TECHNIQUES USED:

The study addresses food waste and hunger by exploring The authors developed an empirical model and used Monte
food redistribution through optimization, focusing on the Carlo simulations to study food redistribution dynamics.
feasibility of redistributing it to feed the hungry.

PROS OF METHODOLOGY: CONS & LIMITATIONS:

- Focuses on two counties in north central Colorado, limiting
- Demonstrates that food rescue and redistribution is
generalizability to other regions.
feasible on both small and large scales.
- Simplified assumptions, like constant food expiration rates,
- Significant amounts of food can be saved from the waste
may affect accuracy.
stream if sufficient resources are allocated.
- Data from only one food bank, limiting the representation of
other food donation models.
DATASET USED:
The study used food rescue data from Community Food
RESULTS OBTAINED:
-Feasibility: Food rescue and redistribution are feasible on small
Share, which includes donations from 90 distinct donors,
and large scales with sufficient resources.
totaling 2.3 million pounds of food distributed to 304
agencies over a 14-month period.
THE POTENTIAL FOR REDUCING FOOD WASTE THROUGH SHELF-LIFE EXTENSION: ACTIONABLE INSIGHTS
FROM DATA DIGITIZATION

YEAR PUBLISHED: 2024 JOURNAL: MDPI, Sustainability

PROBLEM STATEMENT: TECHNIQUES USED:

The study examines how data digitization and The approach focuses on extending shelf life through the use
shelf-life extension strategies can help reduce of data digitization, which monitors food conditions and
food waste across various supply chains. predicts spoilage, helping to optimize logistics

PROS OF METHODOLOGY: CONS & LIMITATIONS:

Shelf-Life Extension: Extending the shelf life Regional Variability: The effectiveness of shelf-life
of food products can significantly reduce extension may differ by region due to environmental
waste. factors.
Data Digitization: Enhances real-time Limited Applicability: Some products, especially those
monitoring and predictions for better supply with short consumption cycles, may not benefit much from
chain management. this approach.

DATASET USED: RESULTS OBTAINED:

The paper utilizes data from various supply chains, Integrate emerging technologies like AI, blockchain, and IoT
collected via data digitization systems, to analyze shelf-life to enhance predictions
extension strategies.
 AN ENSEMBLE ML MODEL TO PREDICT THE WASTAGE OF FOOD: TOWARDS ACHIEVING THE FOOD
SUSTAINABILITY

YEAR PUBLISHED: 2023 JOURNAL: International Conference on Next-Generation

Computing and Informatics System

PROBLEM STATEMENT: TECHNIQUES USED:

Increasing food wastage globally. Machine Learning models including ensemble learning.
Need for efficient prediction of food wastage to improve Techniques like Random Forest, Decision Trees, etc.
sustainability and reduce wastage.

PROS OF METHODOLOGY:
Higher prediction accuracy. CONS & LIMITATIONS:
Supports better decision-making for reducing food waste. Limited data quality.
Adaptable to various food domains. Model interpretability issues.
High computational cost.

DATASET USED: RESULTS OBTAINED:

Real-world food wastage datasets including Ensemble learning showed a 10% higher accuracy
attributes related to food processing, supply compared to standalone models.
chain, and consumption. Reduced prediction error in food wastage.
 USING ARTIFICIAL INTELLIGENCE TO TACKLE FOOD WASTE AND ENHANCE THE CIRCULAR ECONOMY:
MAXIMISING RESOURCE EFFICIENCY AND MINIMISING ENVIRONMENTAL IMPACT: A REVIEW

YEAR PUBLISHED: 2023 JOURNAL: MDPI, Sustainability, Volume 15

PROBLEM STATEMENT: TECHNIQUES USED:

Food waste is a global issue with significant economic,
AI techniques including machine learning, image recognition,
social, and environmental impacts. AI can help reduce food
NLP, smart agriculture, IoT, blockchain, and reinforcement
waste and enhance circular economy practices, but
learning were used to optimize food production, distribution,
challenges exist in implementation.
and waste reduction.

PROS OF METHODOLOGY: CONS & LIMITATIONS:

Reduces food waste across the supply chain High costs and complexity in implementing AI solutions
Increases resource efficiency Lack of government policies and financial incentives
Supports sustainability through AI-driven optimization Technological barriers in infrastructure and expertise

DATASET USED: RESULTS OBTAINED:

Not specified in detail (review-based article AI technologies demonstrated potential for significant reductions
summarizing various AI applications in the food in food waste, improved resource efficiency, and enhanced
industry) sustainability in food production and distribution systems​
 FOOD WASTE MANAGEMENT USING MACHINE LEARNING

YEAR PUBLISHED: 2022 JOURNAL: International Journal of Creative Research

Thoughts (IJCRT), Volume 10, Issue 4
PROBLEM STATEMENT: TECHNIQUES USED:
Increasing food waste in restaurants, resulting in
Linear regression for predicting food consumption and wastage
environmental and economic concerns. The goal is to predict
in restaurants based on historical data.
food wastage and reduce it through donations to NGOs.

PROS OF METHODOLOGY: CONS & LIMITATIONS:

Reduces food wastage Limited to restaurants
Enables donations to NGOs Requires sufficient historical data for accurate
Reduces costs associated with waste management predictions
Helps in efficient food preparation prediction Dependent on third-party NGO and delivery system

DATASET USED: RESULTS OBTAINED:

Prediction model suggests reducing food preparation from 85 kg
Data collected from restaurants regarding the
to 70 kg daily to minimize waste.
amount of food prepared and wasted daily.
 SMART URBAN FOOD RE-DISTRIBUTION USING WEB APPLICATION

YEAR PUBLISHED: 2024 JOURNAL: International Journal of Advanced Research in

Science, Communication and Technology
(IJARSCT), Volume 4, Issue 1
PROBLEM STATEMENT: TECHNIQUES USED:
The growing challenges in urban food distribution, AI-driven predictive modeling, real-time data analytics, route
including inefficiencies, food waste, and food insecurity, optimization, inventory management, NLP (Natural Language
are addressed through a smart, tech-driven solution. Processing), and machine learning.

PROS OF METHODOLOGY: CONS & LIMITATIONS:

Reduces food waste Dependence on stakeholder collaboration
Optimizes food distribution Requires advanced technology infrastructure
Enhances collaboration between donors and recipients Potential logistical challenges in urban areas
Minimizes environmental impact

DATASET USED: RESULTS OBTAINED:

Not explicitly mentioned, but data from grocery stores, Implemented a platform that connects surplus food donors with
restaurants, and private citizens contributing to surplus food recipients (NGOs), utilizing route optimization to minimize costs
donation. and environmental impact.
 FOOD RECOGNITION AND FOOD WASTE ESTIMATION USING CONVOLUTIONAL NEURAL NETWORK

YEAR PUBLISHED: 2022 JOURNAL: MDPI,Electronics, Volume 11, Article 3746

PROBLEM STATEMENT: TECHNIQUES USED:
Food waste is a global issue with negative environmental, Convolutional Neural Network (CNN) model for food
economic, and social impacts. Accurate food recognition recognition and waste estimation based on before-and-after
and waste estimation are needed to tackle this challenge. meal images, optimizing recognition of 157 food categories.

PROS OF METHODOLOGY: CONS & LIMITATIONS:

High accuracy (98.8%) in food recognition Issues with irregular shapes and cutlery affecting waste
Efficient estimation of food waste estimation
Optimized network with CNN layers for deep learning Deficiencies in recognizing certain food items in non-standard
containers
Overfitting risks

DATASET USED: RESULTS OBTAINED:

Images of 157 food categories, including 50 to 200 images per CNN model achieved 98.8% accuracy in food recognition. The
category. Total of 23,552 images. Data from Serbian students’ average food waste in the dataset was 21.3% based on Serbian
meal images was used for waste evaluation. student meal images.
 RECENT DEVELOPMENTS IN RESEARCH ON FOOD WASTE AND THE CIRCULAR ECONOMY

YEAR PUBLISHED: 2024 JOURNAL: MDPI, Biomass, Volume 4, Article 24, 2024
PROBLEM STATEMENT: TECHNIQUES USED:
Food waste is a global issue with implications for food Bibliometric analysis of 1022 documents.
security and sustainability. The study examines the role of Performance analysis to assess publications, citations, and
the circular economy in reducing food waste. research impact.
Scientific mapping to identify research themes and
collaboration.
PROS OF METHODOLOGY:
Identified key themes like bioenergy, resource CONS & LIMITATIONS:
recovery, and sustainability. Focused on Scopus data, excluding non-English and non-
Highlighted top authors, journals, and countries. journal publications.
Showcased growing interest in food waste Overemphasis on highly cited works, missing emerging
valorization. research.
DATASET USED: RESULTS OBTAINED:
1022 documents from Scopus, published between 2014 and Significant rise in publications over the last decade, with focus
2023. areas shifting to sustainability and waste valorization.
 PAPER SUMMARY

SMART URBAN FOOD RE-DISTRIBUTION USING WEB HARNESSING ARTIFICIAL INTELLIGENCE FOR URBAN
APPLICATION FOOD REDISTRIBUTION: A SOCIO-TECHNICAL ANALYSIS
OF THE FEEDING AMERICA INITIATIVE
SUFD Platform: The Smart Urban Food Distribution (SUFD)
platform addresses urban food redistribution using
advanced technologies. AI Technologies: Deep learning, self-driving vehicles, real-
AI-Driven Predictive Modeling: AI models predict food time data analytics.
supply and demand, optimizing redistribution efforts. Food Redistribution System: Surplus food collected from
Real-Time Data Analytics: The system uses real-time data to restaurants/stores, distributed via vending hubs to
enhance decision-making and ensure timely food delivery. vulnerable communities.
Last-Mile Delivery: Incorporates drones and driverless cars Optimization: AI-driven logistics to reduce food waste and
for efficient last-mile food delivery. enhance access for homeless populations.
Centralized Platform: Connects donors, recipients, and Business Incentives: Tax benefits and corporate social
volunteers, streamlining food distribution and reducing responsibility for participating businesses.
waste. Impact: Social (improved food access), environmental
Key Features: Includes inventory management, route (waste reduction), economic (business incentives).
optimization, and stakeholder collaboration to improve Challenges: Implementation hurdles, regulatory
supply chain resilience. compliance, AI accuracy.
Goals: Aims to enhance food security, reduce food waste,
and minimize environmental impact.
 PAPER SUMMARY

REDUCING FOOD WASTE USING MACHINE LEARNING IMPROVING FOOD DISTRIBUTION IN URBAN AREAS WITH
MODELS: FORECASTING AND OPTIMIZATION APPROACHES IOT AND ML
ML for Food Waste Reduction: Explores how machine learning (ML) can
help tackle food waste through advanced demand forecasting and IoT and ML Integration: Explores how combining IoT and ML can
optimization. transform urban food distribution systems.
Improved Inventory Management: ML analyzes large datasets to Real-Time Monitoring: IoT sensors track food conditions
predict food demand more accurately, improving inventory continuously, ensuring quality control and minimizing waste.
management and reducing waste. Demand Forecasting & Optimization: ML algorithms are used for
ML Techniques: demand forecasting, route optimization, and inventory management,
1. Supervised Learning: For predicting food demand and managing improving operational efficiency.
supply chains. Blockchain for Traceability: Highlights the role of blockchain in
2. Unsupervised Learning: For identifying patterns in food consumption ensuring food traceability and enhancing transparency in the supply
and waste. chain.
3. Reinforcement Learning: For optimizing supply chains and reducing Collaboration: Stresses the need for collaboration among
perishable food waste. stakeholders (donors, distributors, consumers) for an efficient food
4. Real-World Examples: Case studies from companies like Walmart and distribution system.
IBM Food Trust show the practical effectiveness of ML in reducing food Challenges: Addresses issues such as data privacy concerns and
waste. challenges in technology integration.
Challenges: Issues like data quality, integration costs, and ethical Conclusion: IoT and ML are essential for building a more efficient,
considerations are obstacles to wider adoption. sustainable, and customer-focused food distribution system in urban
Conclusion: ML holds transformative potential to reduce food waste areas.
and promote sustainable food systems.
 FINAL PAPER SUMMARY

FOOD REDISTRIBUTION AS OPTIMIZATION

by Caleb Phillips and Colleagues
1. Dual Challenges: The study addresses both food waste and hunger by proposing a solution through food rescue and
redistribution.

2. Empirical Model: The authors use data from a Colorado food bank to develop a model that simulates and optimizes food
recovery and redistribution processes.

3. Food Supply Distribution: The model shows that the food supply follows a heavy-tailed distribution, meaning a small
number of donors provide a large portion of the available food.

4. Feasibility of Food Rescue: Food rescue is achievable if the food is swiftly transported and preserved before expiration.

5. Key Factors: The number of donors and efficient transportation systems are crucial to reducing costs and ensuring
sufficient food is rescued to meet demand.

6. Scalability: With enough resources, the approach can be scaled, making it a viable solution for reducing food waste while
addressing food insecurity.
 PROBLEM STATEMENT

This project aims to develop a machine learning model to forecast food donation volumes and
optimize redistribution logistics, addressing challenges like unpredictable donation patterns and
high transportation costs. By leveraging historical data, the model will minimize costs and food
waste while maximizing deliveries to at-risk populations.

KEY OBJECTIVES

⁠ redict daily food donations from various sources (grocery stores, restaurants, etc.) based on
P
historical data using machine learning techniques such as time-series forecasting.
⁠Optimize the scheduling of pick-up and delivery routes by integrating the predicted donation
volumes with transportation and storage constraints (e.g., perishability of food).
⁠Minimize food wastage by ensuring timely transportation of perishable goods while reducing the
overall operational costs of food redistribution organizations.
 MODELS & ALGORITHM
Algorithm Used:

Monte Carlo Simulation: The model uses Monte Carlo simulations to study the dynamic nature of food rescue and redistribution. The supply
of food is treated as a stochastic (random) process.
Generalized Pareto Distribution (GPD): The food supply follows a heavy-tailed distribution, and the model fits this with a Generalized Pareto
Distribution to handle extreme supply events.
Optimization Problem: The food redistribution problem is framed as an optimization problem where the goal is to minimize the cost of
picking up food from donors while ensuring demand (hunger) is met. The optimization is based on a linear programming (LP) approach to
determine the optimal daily pickup schedules, which minimizes driving costs while meeting food demand.
Geographic Information Systems (GIS): GIS data is used to cluster donors based on their geographical proximity to minimize driving
distances for food pickups.

Why This Algorithm:

The combination of Monte Carlo simulations and optimization is ideal because:

Stochastic Nature: Food supply and demand are unpredictable, so Monte Carlo methods simulate many possible scenarios.
Extreme Events: GPD is well-suited for modeling rare but significant events, such as large food donations.
Optimization: The use of LP minimizes costs in a complex problem where there are multiple constraints (e.g., food demand, donor locations,
and perishable food).
 PROS & CONS

Pros:
1. Scalability: The algorithm can be scaled to larger regions or more donors as it models real-world variability and uncertainty.
2. Flexibility: By using LP, the algorithm can handle a wide variety of constraints, including donor availability and cost-efficiency.
3. Realistic: Monte Carlo simulations offer a realistic picture of the problem by accounting for random events (like large donations).

Cons:
1. Complexity: The heavy-tailed distribution and the Monte Carlo simulation require significant computational resources, especially
when modeling large-scale systems.
2. Assumptions: The algorithm assumes food expiration rates and donor availability, which may not always align with real-world
conditions.

Potential Solutions to Cons:

1. Model Simplification: Using heuristic or metaheuristic methods to approximate solutions in complex, non-linear cases can
reduce computational costs.
2. Improved Forecasting: Enhancing the accuracy of food supply and expiration prediction models can lead to better performance.
3. Parallel Processing: Implementing parallel computing or distributed computing methods can reduce the time complexity of
Monte Carlo simulations.
 Model/Approach Description Advantages Disadvantages Use Case Example

Proven optimization
Optimization technique used May not handle real-time data
technique; effective for Scheduling food deliveries
Linear Programming to minimize waste by matching or complex, non-linear
scheduling and resource from surplus to areas in need.
supply with demand. problems well.
allocation.

Classification algorithm that Easy to interpret, useful for

Can become complex and Classifying food items based
helps to predict surplus and decision-making, and can
Decision Trees prone to overfitting with large on perishability or suitability
demand patterns based on handle both numerical and
datasets. for donation.
historical data. categorical data.

Ensemble method based on Predicting the best

High accuracy and reduces Requires more computational
decision trees, improves destinations for food
Random Forest overfitting compared to power and can become slow
accuracy and handles more redistribution based on
decision trees. with large datasets.
complex data. patterns in large datasets.

Clustering algorithm used to Efficient for identifying Not good with clusters of
Grouping areas with similar
K-Means Clustering group similar sources of patterns in large datasets and varying density; requires
food demand or food donors.
surplus or demand centers. creating groupings. predefined number of clusters.
 Model/Approach Description Advantages Disadvantages Use Case Example

Capable of handling very large Requires large datasets and

Deep learning model to predict Predicting complex supply and
datasets and can model significant computational
Neural Networks food demand and surplus demand shifts in food
complex, non-linear resources; harder to interpret
trends over time. redistribution.
relationships. results.

Used for predicting future Limited in dealing with

Can capture patterns over time Predicting future food surplus
Time Series Forecasting trends in food surplus and external factors or sudden
and forecast future surplus or based on seasonal patterns and
(ARIMA, LSTM) demand based on historical changes (ARIMA); LSTM
demand fluctuations. consumption trends.
data. requires large datasets.

Efficient in optimizing
Algorithms used for route Limited in handling uncertain Optimizing delivery routes
Graph-based Algorithms (e.g., transportation routes; can
optimization in food conditions like traffic or between food donors and
Dijkstra's, A)* reduce fuel costs and time for
distribution networks delays. recipients.
redistribution.
 REFERENCES
https://www.mdpi.com/

https://ieeexplore.ieee.org/Xplore/home.jsp

https://www.researchgate.net/

https://scholar.google.com/
THANK YOU