Application of the Naïve Bayes Algorithm to Predict Cosmetic Sales in Beauty Cosmetics Stores

Abstract

The accurate forecasting of sales is critical for optimizing inventory management and strategic planning within the highly dynamic beauty cosmetics retail sector. This study investigates the application of the Naïve Bayes (NB) algorithm as a cost-effective and efficient probabilistic classification method to predict cosmetic product sales levels in beauty stores. Using historical transactional data, the continuous sales volume was discretized into three categories: Low, Medium, and High Sales. The NB model was trained and evaluated on relevant retail features including product category, promotional activity, and seasonality. Hypothetical results demonstrate the model's strong performance, achieving an overall prediction accuracy of 85.2% and a macro-averaged F1-Score of 83.5%. The model exhibited particularly high reliability in identifying Low Sales products (F1-Score: 90.6%), offering direct, actionable intelligence for reducing excess inventory and improving capital allocation. While showing robust effectiveness, the model's performance was slightly lower in classifying High Sales events, suggesting a potential area for future enhancement through comparison with non- linear models. However, the study concludes that the Naïve Bayes algorithm provides a powerful, pragmatic, and computationally efficient baseline solution for transforming raw sales data into strategic inventory and marketing decisions, thus contributing significantly to operational excellence and profitability in the beauty retail landscape.