I have a public Kaggle dataset with time-stamped observations and several explanatory variables. My objective is not classic forecasting; instead, I want a well-constructed regression workflow that reveals how the predictors interact over time and which of them truly drive the target variable. Please build and document the full pipeline in Python, preferably inside a Kaggle Notebook so the work is fully reproducible. Standard tools such as pandas, NumPy, scikit-learn, statsmodels, XGBoost or LightGBM are all welcome as long as the code stays readable and the findings are clearly explained. Deliverables I need to see • An exploratory section that highlights trends, seasonality and any data quality quirks • Clean, commented preprocessing and feature-engineering steps suitable for time-series data (proper train-test split, lag features, etc.) • At least one baseline model and one tuned model, each evaluated with appropriate regression metrics • Visual and written interpretation of variable importance or similar techniques that make the relationships evident • A brief takeaway section that summarises what drives the target and any limitations of the current approach I will review the notebook directly on Kaggle, so please make sure it runs end-to-end without external dependencies.