Cinema audiance prediction
Budget: ₹600 – ₹1,500 INR
I have a daily-frequency time–series dataset and need a robust numerical regression model that pushes the limits of accuracy without stepping outside the approved toolset. You may rely on NumPy, Pandas, Matplotlib, Scikit-learn, XGBoost, LightGBM, Seaborn, Imblearn, SciPy, Pickle, regex, Plotly and any native Python module; anything beyond that (TensorFlow, PyTorch, NLP-specific packages, etc.) is off-limits.
The raw data are already in a Kaggle notebook environment, so you can jump straight into exploratory analysis, feature engineering, and modeling. I’m looking for a well-documented notebook that:
• walks through preprocessing, handling any imbalance with Imblearn if necessary,
• justifies model choice (Gradient Boosting, Random Forest, LightGBM, XGBoost or any Scikit-learn regressor),
• tunes hyper-parameters rigorously—cross-validation is a must,
• returns the best-performing model saved with Pickle,
• visualises key insights with Seaborn or Plotly, and
• finishes with a concise explanation of performance metrics and potential next steps.
If you can squeeze every last drop of predictive power out of these libraries while keeping the code clean and reproducible, let’s get started.
The raw data are already in a Kaggle notebook environment, so you can jump straight into exploratory analysis, feature engineering, and modeling. I’m looking for a well-documented notebook that:
• walks through preprocessing, handling any imbalance with Imblearn if necessary,
• justifies model choice (Gradient Boosting, Random Forest, LightGBM, XGBoost or any Scikit-learn regressor),
• tunes hyper-parameters rigorously—cross-validation is a must,
• returns the best-performing model saved with Pickle,
• visualises key insights with Seaborn or Plotly, and
• finishes with a concise explanation of performance metrics and potential next steps.
If you can squeeze every last drop of predictive power out of these libraries while keeping the code clean and reproducible, let’s get started.