EXPANDED STORE SALES PREDICTION

COMPUTER SCIENCE AND ENFINEERING

optimized staffing, and informed strategic decisions. This project explores the predictive performance of various regression models on the Rossmann Store Sales dataset, a rich, structured dataset containing historical sales data from over 1,000 retail outlets. To evaluate and compare modeling approaches, we categorize the models into two groups based on their mathematical foundations and capacity to handle structured data: Model Group 1 includes regularized linear models—Lasso, Ridge, and Elastic_Net—which help address overfitting and multicollinearity, alongside a basic Decision Tree Regressor for modeling non-linear relationships. Model Group 2 consists of both simple and advanced machine learning models, such as Linear Regression (baseline), DecisionTreeRegressor (for comparison), Random Forest Regressor (bagging ensemble), and AdaBoost Regressor (boosting ensemble). These models are known for their ability to capture complex patterns and improve predictive accuracy. Each model is trained and tested on the same dataset, and their performance is assessed using standard regression evaluation metrics: R² (coefficient of determination), RMSE (Root Mean Squared Error), and MAE (Mean Absolute Error). The analysis reveals that while regularized models offer advantages in interpretability and feature selection, ensemble methods like Random Forest and AdaBoost consistently outperform others in terms of predictive accuracy. This study highlights the importance of model selection based on the trade-off between complexity, accuracy, and interpretability for real-world retail forecasting applications.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.