Implementation of Grid Search Optimization Algorithm and Adaptive Response Rate Exponential Smoothing In Product Sales Prediction

Effective inventory management is one of the keys to a company's success, especially in the retail and distribution sectors that are highly dependent on product availability according to market demand. One common problem faced in inventory management is deadstock, which is a condition where a product is not sold for a long time, causing a buildup of goods and financial losses. This problem is generally caused by inaccuracy in predicting product sales needs. This study aims to overcome this problem by implementing the Adaptive Response Rate Exponential Smoothing (ARRES) algorithm combined with the Grid Search optimization method to improve the accuracy of sales predictions. By utilizing the Sales Data Analysis dataset from Kaggle, the algorithm is implemented in a web-based system using Python and Flask. The results showed that the combination of Grid Search and ARRES was able to significantly increase prediction accuracy, as indicated by a decrease in the MAPE value from 2.845% (ARRES only ) to 0.877% (Grid Search + ARRES). This proves that the proposed method can help companies manage stock more efficiently, reduce the risk of deadstock, and increase the effectiveness of product sales planning