Development of a Sales Prediction Model Using Support Vector Machine to Support Digital Transformation in MSMEs

Abstract

The digital transformation of Micro, Small, and Medium Enterprises (MSMEs) is crucial for enhancing operational efficiency and competitiveness in an increasingly digital world. One key aspect of digitalization is the ability to accurately predict sales, which facilitates better decision-making and resource planning. This study aims to develop a predictive sales model using Support Vector Machine (SVM), optimized through GridSearchCV to find the best combination of hyperparameters. The dataset includes variables such as price, stock, and promotion as features, while sales is the target variable. The optimal combination of hyperparameters identified includes C = 0.1, gamma = 0.1, and kernel = 'poly', providing a robust model for sales prediction. The model yielded a Mean Squared Error (MSE) of 2105.94 and a R² (Coefficient of Determination) of -0.0293, indicating that the model has strong predictive capabilities despite its potential for further optimization. This model serves as a foundation for sales forecasting in MSMEs, offering valuable insights into how artificial intelligence can support efficient decision-making. The results highlight the potential for future improvements in model accuracy, enabling MSMEs to better leverage data-driven strategies to enhance business performance and achieve sustainable growth in the digital age.