A Scalable Machine Learning Model for Sales Forecasting using PySpark
Talluri Harshitha, Simhadri Tanya, Thadakaluru Jaswanthi, Manju Venugopalan · 2024
Sales forecasting is a useful tool used for estimating future demand and sales. It provides insightful information for strategic decision-making, resource allocation, and inventory control. This research aims to utilize various machine learning algorithms for constructing a robust sales forecasting model, with the help of the Machine Learning library in PySpark. PySpark is chosen for its scalability, parallel processing framework, and seamless integration with machine learning libraries, making it an ideal choice for handling large datasets. MLlib is a powerful machine learning library within PySpark, offering a comprehensive set of scalable and distributed algorithms that enable efficient processing and analysis of large datasets. The research also incorporates time-series analysis. The performance of the research is evaluated using evaluation metrics R-Square(R2) value Root Mean Squared Error (RMSE) and Mean Squared Error (MSE), Mean Absolute Error (MAE). Out of all experimented models, Gradient Boosting Regressor stands out as the best-performing model with R-squared value of 0.93, showcasing its effectiveness in accurate sales predictions.