An Efficient IoT-Based Automated Food Waste Management System with Food Spoilage Detection

Abstract

Recent years have seen a growing concern about food waste, leading to advanced research efforts aimed at addressing its widespread consequences. This concern poses a significant threat to the long-term stability of food supply chains, demand patterns, and production processes. Given the universal importance of nutrition, ensuring the quality and safety of food remains a foremost priority. To address this challenge, a groundbreaking food deterioration monitoring system has been developed. This system uses sensors and actuators to monitor gas emissions, humidity, and temperature in fruits and vegetables. It operates with the Node MCU microcontroller, in conjunction with sensors such as the MQ2/MQ4 methane sensor and the DHT11 humidity/temperature sensor. This system provides real-time assessment capabilities for a variety of food items, including rice, bread, samosas, and dal. The system is anchored by a comprehensive dataset that encompasses a diverse range of food items, locations, temperatures, and humidity levels. This dataset forms the foundation for predictive and analytical endeavors. The analytical process encompasses several crucial stages. Rigorous data preprocessing techniques enhance dataset quality by addressing missing values and outliers. The Recursive Feature Elimination (RFE) method optimizes predictive efficiency by iteratively selecting significant features and mitigating overfitting. The M-SMOTE technique corrects class imbalances by generating synthetic samples to balance model training for underrepresented classes. The Random Forest algorithm combines decision trees to offer robust predictive insights. These analyses empower the system to detect spoilage, provide decision-making insights, and predict remaining shelf life, leading to efficient resource allocation. The proposed results demonstrate an accuracy of 94.76%, underscoring its practicality.