Predicting purchase probability of retail items using an ensemble learning approach and historical data

Planning for stocking and supply chain management of retail markets is critical to ensure the high availability of retail items while reducing the risks of oversupply and overstocking. The ability to predict purchase probability of retail items accurately and efficiently is critical to enable such optimized supply chain management. In this paper, we use historical purchase data, carry out pre-processing, analysis, as well as build an ensemble learning-based model to efficiently predict purchase probability of retail items. The proposed ensemble learning model is composed of different segments utilizing Random Forests, Convolution Neural Networks, Extreme Gradient Boosting (XGBoost), and voting mechanism. Detailed evaluation of the proposed solution was carried out by analyzing accuracy, precision, F1 score, sensitivity, specificity, and more. The evaluation further included efficiency analysis, complexity analysis. The proposed solution performed better than the existing solutions, as shown in the evaluation.