Enhancing Sales Demand Prediction for Supply Chain Management: A Dimensionality Reduction Approach at Mukwano Company Limited, Uganda

Abstract

In the dynamic landscape of today's business environment, the accurate prediction of sales demand plays a pivotal role for companies striving to maintain competitiveness. This study focuses on refining demand sales prediction for supply chain management at Mukwano Industries Limited through the application of dimensionality reduction techniques. The research involved a systematic iterative process encompassing problem identification, data preparation, modeling, evaluation, validation, optimization, and documentation. To ensure confidentiality, secondary data from the Mukwano IT department underwent meticulous merging and anonymization. Four dimensionality reduction algorithms, namely PCA, SVD, MDS, and t-SNE, were employed and evaluated using RMSE metrics. The results reveal that MDS and t-SNE exhibited exceptional performance, achieving accuracies of 89% and 88.8%, respectively. PCA and SVD also demonstrated commendable performance with an accuracy of 82.4%. The study underscores the crucial role of dimensionality reduction in enhancing predictive accuracy and optimizing inventory management. Recommendations include the incorporation of feature selection and regularization techniques to address the challenges associated with the curse of dimensionality. In conclusion, this research contributes valuable insights into the effectiveness of diverse dimensionality reduction techniques for demand prediction and inventory management. Additionally, the study highlights the need to address the curse of dimensionality and suggests exploring further research avenues in other aspects of supply chain management. These recommendations are essential for guiding future research efforts in this evolving field. Keywords: Demand sales prediction, Supply chain management, Feature selection, Inventory management, Predictive accuracy