Prediction of cold chain loading environment for agricultural products based on K-medoids-LSTM-XGBoost ensemble model

Abstract

Cold chain loading is a crucial aspect in the process of cold chain transportation, aiming to enhance the quality, reduce energy consumption, and minimize costs associated with cold chain logistics. To achieve these objectives, this study proposes a prediction method based on the combined model of K-medoids-long short-term memory (LSTM) networks—eXtreme Gradient Boosting (XGBoost). This ensemble model accurately predicts the temperature for a specified future time period, providing an appropriate cold chain loading environment for goods. After obtaining temperature data pertaining to the cold chain loading environment, the K-medoids algorithm is initially employed to fuse the data, which is then fed into the constructed ensemble model. The model’s mean absolute error (MAE) is determined to be 2.5343. The experimental results demonstrate that the K-medoids-LSTM-XGBoost combined prediction model outperforms individual models and similar ensemble models in accurately predicting the agricultural product’s cold chain loading environment. This model offers improved monitoring and management capabilities for personnel involved in the cold chain loading process.

Keywords

• K-medoids
• XGBoost
• Ensemble model
• Cold chain loading
• Prediction