Dyna ML - A Machine Learning Approach to Sales Forecasting and Product Recommendations

This work explores machine learning models for sales forecasting and product recommendation systems using the ContosoRetailDW dataset. The study aims to improve prediction accuracy and recommendation effectiveness that supports data-driven decision-making. Sales Forecasting: The project uses machine learning algorithms such as Linear Regression, Decision Trees, Random Forests, Support Vector Regressor (SVR), and XGBoost to predict sales volumes. Data preprocessing includes handling missing values, encoding categorical features, and normalizing numerical data. Models are evaluated using standard performance metrics. Preliminary results show that feature integration significantly improves prediction accuracy. Product Recommendations: Two recommendation systems are developed. The first uses collaborative filtering, including Singular Value Decomposition (SVD), to predict user preferences based on previous interactions. Using text clustering analysis, new customers are targeted through product popularity metrics and item-by-item recommendations. The second method leverages the Surprise library, implements algorithms such as SVD and KNN, and evaluates them based on standard accuracy metrics. The comparative analysis highlights the strengths of each approach in different retail scenarios. The technology stack includes SQL for data management, Python for model development, and Power BI for visualization. Expected results include improved accuracy in sales predictions and personalized recommendations and a comprehensive understanding of the impact of various data attributes on sales performance. These results aim to provide strategic insights that can drive more informed business decisions, showcasing the practical applications of machine learning in transforming data into actionable intelligence.