Aditya Gupta, Vibha Jain
ABSTRACT The perishable food cold chain is a vital part of the global food system, mainly ensuring the quality of temperature‐sensitive products such as milk, seafood, fruits, and vegetables. However, this system still faces several challenges related to real‐time monitoring, data transparency, and proactive demand planning, which often lead to spoilage, food safety violations, and disruptions. In this work, we propose a comprehensive end‐to‐end system that includes Internet of Things (IoT)‐based environmental sensors, blockchain‐enabled immutable logging, and a hybrid ARIMA–LSTM model for demand forecasting and spoilage risk detection. The system uses ESP32 microcontrollers with DHT11 sensors at key cold‐chain points to collect temperature and humidity data, streamed to Blynk dashboards for real‐time visualization and anomaly alerts. Anomalies and product metadata are permanently recorded through smart contracts on a private Ethereum blockchain, while payloads are stored off‐chain on IPFS for traceability and auditability. Additionally, historical blockchain logs and external sales data are used to train a hybrid ARIMA–LSTM model that predicts future demand and spoilage risks more accurately. Experimental results show that the proposed model achieves a forecasting accuracy of 90% ( R 2 = 0.90), outperforming baseline approaches on multiple metrics including RMSE, MAE, MAPE, and R 2 . The framework is scalable and data‐driven, aiming to improve supply availability and reduce food waste.