Purchasing Prediction Using Machine Learning Algorithms for Optimizing Inventory Management

Abstract

Effective inventory management is a crucial element in company operations, especially in maintaining a balance between demand and supply. Good inventory management can reduce storage costs, increase product availability, and maximize company profits. However, the challenges that companies often face are the uncertainty of market demand and changes in trends that are difficult to predict. Along with technological developments, traditional methods of inventory management are starting to be replaced by data-based approaches and machine learning algorithms. The use of machine learning is not only limited to predicting purchasing needs, but can also be applied in various other business aspects. This research aims to optimize HP spare parts inventory management at Store X using the Long Short-Term Memory (LSTM) method. By analyzing sales data for 2023 which consists of 96,630 lines, the research applies systematic stages: data acquisition, preprocessing, exploratory data analysis, model building, and evaluation. The LSTM method is used to predict spare parts stock with significant accuracy, demonstrated through evaluation metrics: Mean Absolute Error (MAE) 12%, Mean Squared Error (MSE) 2%, and Root Mean Square Error (RMSE) 15%. The model successfully captured seasonal patterns and trends in sales data, proving its ability to forecast stock requirements. The research results show that the LSTM-based machine learning approach is effective in supporting inventory management decision making, helps reduce the risk of losses due to stock uncertainty, and increases the efficiency of managing HP spare parts inventory.