Sales forecasting for supply chain using artificial intelligence

Abstract

Supply chain management and logistics are two sectors currently experiencing a transformation thanks to the advent of AI(Artificial Intelligence) technologies. Leveraging predictive analytics powered by AI presents businesses with novel opportunities to streamline their operations effectively. This study utilizes sales forecasting for predictive analysis using three distinct artificial intelligence paradigms: Long Short-term Memory (LSTM), Bayesian Neural Networks (BNN) – both of which belong to the family of deep learning models – and Support Vector Regressors (SVR), a machine learning technique. The empirical data employed for this forecast stems from the historical sales data of Bactiguard, the collaborating company in this study. Subsequent to the essential data manipulation, these models are trained, and their respective results are assessed. The evaluation matrices incorporated in this study include the mean absolute error (MAE), root mean square error (RMSE), and the R2 score. Upon analysis, the LSTM model emerges as the clear frontrunner, exhibiting the lowest error rates and the highest R2 score. The BNN follows closely, demonstrating credible performance, while the SVR lags, presenting suboptimal results. In conclusion, this study highlights the accuracy and efficiency of artificial intelligence models in sales forecasting and underscores their practical, real-world applications.