Sales Prediction Using Advertising Expenditures

This project analyzes how advertising investments across TV, Radio, and Newspaper channels affect product sales and builds predictive models to forecast sales given media spend. The work combines thorough exploratory data analysis (EDA), feature engineering, and regression modeling to quantify the return on advertising and recommend optimal allocations. Starting with visualization of spend distributions and pairwise relationships, the project uncovers the relative strength of each medium’s influence on sales.

During EDA, scatter plots and correlation matrices revealed strong associations—TV often shows the largest marginal effect, while radio and newspaper contributions vary by dataset. Data preprocessing included handling outliers, scaling continuous variables, and verifying linearity assumptions. Models developed included simple and multiple linear regression for interpretability, regularized regression (Ridge/Lasso) to control overfitting, and tree-based regressors for non-linear patterns. Model performance was validated using cross-validation and evaluated with RMSE, MAE, and R² metrics.

Final deliverables include a production-ready regression pipeline that predicts sales from specified ad spends, a dashboard-ready set of visuals (spend vs sales scatter, residuals plot, prediction vs actual plot), and business recommendations to re-balance media budgets for maximum ROI. This project demonstrates how data-driven budgeting can increase campaign efficiency and improve marketing ROI for businesses of all sizes.