Machine Learning Models for Enhancing Time Series Demand Forecasting in Business Intelligence

This research explores the efficiency of multiple machine learning models in the context of intelligent demand forecasting in the industry of business intelligence. The research is based on a dataset that presents the sales history of a textile company during a one-year time series. In particular, the research analyses four models: Artificial Neural Network, Long Short-Term Memory, Decision Tree, and Linear Regression to forecast demand based on the sales in the past. Moreover, the analysis applies multiple preprocessing stages to clean the dataset, such as missing value imputation, outlier detection, and feature engineering. The models are trained and tested using a part of the dataset where 70% of data is used for training, and the remaining 30% is utilized for testing. The results reveal that ANN shows the most explicable results with the lowest data loss and the highest accuracy in demand forecasting. Meanwhile, LSTM is competitive but shows slightly worse accuracy results than ANN. The Decision Tree and Linear Regression models provide some insights into the issue but have higher data loss and lower accuracy. The results show that the use of advanced machine learning models, particularly ANN, is vital for the accurate forecasting of demand in the context of business intelligence. Such forecasting helps companies to predict demand more accurately and optimize their inventory management and supply chain in general, which increases the efficiency of their work. Among the potential subject to further exploration are more complex ensembles or hybrid models that can increase the accuracy of demand forecasting in the rapidly changing business landscape. Overall, the research makes a valuable contribution to the understanding and application of machine learning in demand forecasting.