Predictive Modeling of Dairy Sales Using Multi-Perspective Fusion Bi-LSTM Integrated with Universal Scale CNN: Insights from the Dairy Supply Chain

Abstract

Sales prediction is a significant task of every industries. A potential prediction may majorly impact the revenue loss, out of stock and excessive stock. Many existing research have been implemented to predict the dairy sales prediction, however there are some limitations. To address this problem, the proposed research uses DL (Deep Learning) based technique to forecast the dairy sales. The proposed research uses dairy supply chain dataset to assess the proposed model CNN (Convolutional Neural Network). The present research uses Universal Scale CNN, specifically 1D-CNN, that is able to acquiring the features in ideal and in effective rates. Followed by, the extracted features are fed as an input to Multi-Perspective based Bi-LSTM (Bidirectional Long Short Term Memory) that is able to acquiring the features in an effective manner in characteristics of reducing the error rates upon the prediction sales rate of dairy based products. The performance of proposed Multi-Perspective Fusion Bi-LSTM with Universal Scale CNN is evaluated by different performance metrics which includes RMSE (Root Mean Squared Error), MAE (Mean Absolute Error), MSE (Mean Square Error) and R² (R Square).   When compare to other models, the proposed Multi-Perspective Fusion Bi-LSTM with Universal Scale CNN outperforms with optimal performance value with high R² value of 0.9824. The performance metrics provides broad analysis in terms of accurate prediction of the proposed model.