AI-Powered Revenue Forecasting for Makeup Products

This chapter presents a tailored approach to predict revenue within the supply chain context for the beauty and cosmetics industry. The analysis draws on actual business data sourced from a rapidly expanding fashion and beauty startup. The dataset covers a broad spectrum of time line factors which impact supply chain results. Various AI models were trained and utilized like linear regression, Random Forest to more complex algorithms, RNNs, and LSTM architectures. Model performance was assessed using established metrics R-squared (R2), MAE, and RMSE. LSTM outperforms well compared to other above mentioned AI models in centralized systems. To further enhance the data privacy the model is integrated to Federated Learning (FL) framework to enhance the data preservation, and employed the FedAvg technique. By testing the results that the federated LSTM achieves high R2 value of 0.952, validating its strength without compromising privacy. These findings highlight the practicality of FL for secure, scalable predictive modeling in the fast forward supply chain landscape of the beauty sector.