Best Selling Product Sales Prediction Using K-Nearest Neighbors (KNN) Algorithm

Abstract

Purpose: This research aims to evaluate the performance of the K-Nearest Neighbor (KNN) algorithm in predicting product sales based on historical data. The data includes product price information, sales amount in the previous month, product category, promotion, and seasonal factors. This research involves several experiments with various K values and the application of data normalization to optimize prediction accuracy.

Methods/Study design/approach: The KNN algorithm was chosen for its simplicity and ability to handle multivariate data.

Results/Findings: The results show that the value of K=7 is the optimal parameter for this dataset, with a Root Mean Squared Error (RMSE) value of 110. Normalizing the data is proven to improve the model's accuracy, reducing the RMSE value by about 10% compared to the unnormalized data. Product price, previous sales amount, and promotion features significantly contribute to sales prediction.

Novelty/Originality/Value: This research is expected to provide insight for companies that want to use machine learning to predict product sales and support business decision-making.