Meat Freshness Prediction

TL;DR

This paper presents a machine learning approach to assess meat freshness in retail, aiming to improve accuracy and reduce waste compared to traditional methods.

Contribution

It introduces a novel ML-based model that classifies meat freshness with over 90% accuracy, enhancing efficiency and sustainability in food retail.

Findings

Model achieved over 90% accuracy.

High performance in misclassification cost.

Potential to reduce food waste and improve safety.

Abstract

In most retail stores, the number of days since initial processing is used as a proxy for estimating the freshness of perishable foods or freshness is assessed manually by an employee. While the former method can lead to wastage, as some fresh foods might get disposed after a fixed number of days, the latter can be time-consuming, expensive and impractical at scale. This project aims to propose a Machine Learning (ML) based approach that evaluates freshness of food based on live data. For the current scope, it only considers meat as a the subject of analysis and attempts to classify pieces of meat as fresh, half-fresh or spoiled. Finally the model achieved an accuracy of above 90% and relatively high performance in terms of the cost of misclassification. It is expected that the technology will contribute to the optimization of the client's business operation, reducing the risk of selling defective or rotten products that can entail serious monetary, non-monetary and health-based consequences while also achieving higher corporate value as a sustainable company by reducing food wastage through timely sales and disposal.