Machine Learning Algorithms for Enhancing Predictive Analytics in ERP-Enabled Online Retail Platform

Abstract

Enterprise Resource Planning (ERP) systems are known to be key to online retail management and its resources and operations where proper demand forecasting and sales projections are needed to be effective and competitive. This paper presents a Convolutional Neural Network (CNN) based prediction model for online retail systems that are ERP-enabled. Online Retail II is used to test the model. The methodology starts with the large-scale data preprocessing that involve cleaning, feature engineering, label encoding, and normalization with the help of Standard Scaler and then the further division into testing and training sets entails. The following methods are used to evaluate and train convolutional neural network (CNN) models: root-mean-squared error (RMSE), R2-score, mean absolute percentage error (MAPE), and mean absolute error (MAE). With an R2 of 94, MAE of 2.277, RMSE of 2.814, and MAPE of 13.72%, the experimental findings show that the suggested CNN outperforms the conventional machine learning models. Additional comparative analysis indicates the superiority of the CNN over the models of Decision Tree, Gradient Boosting and Random Forest, which prove its strength in reflecting complex transaction patterns. The results highlight the opportunities of deep learning in enhancing online retail forecasting using ERP which, in turn, enhances business decision-making, operational effectiveness, and customer satisfaction.