Modeling process:

• Imported Python libraries and installed missing ones.
• Prepared data by preprocessing, including converting categorical variables to numerical, handling outliers and missing data, and deriving month and day variables. Encoded categorical variables using one-hot and hash encoding.
• Split data into training and testing sets and used XGBoost with hyperparameter optimization.
• Evaluated model performance using metrics like MSE and R². Found an R² value of 0.74 and MSE of 0.03. Used K-Fold Cross Validation with K = 4 for consistent performance.
• Conducted feature importance analysis using Permutation Importance and found that ad_unit, SSP, browser, and creative_adsize are most impactful, while month, device_type, day, and creative_type are least impactful. Other methods like Gain, Weight, and SHAP can be used for detailed analysis.
• Used Learning Curve Analysis to check for overfitting and underfitting and found no signs with improved test performance as data size increased.