Classification with Playground Series - S4E7
Kaggle Playground July 2024 - Binary Classification of Insurance Cross Selling
Goal: The objective of this competition is to predict which customers respond positively to an automobile insurance offer.
Evaluation: Submissions are evaluated using area under the ROC curve using the predicted probabilities and the ground truth targets.
Learning:This is my first time applying auto machine learning (Neural Network) with LightAutoML. I learned how to process data for this specific tool and how to config my model to optimize it. I also work on ensembling multiple outputs from other Gradient Boosting Algorithms to improve performance.
Table of Contents
1. Preparation (libraries and data)
2. Exploratory Data Analysis (EDA)
3. Data Preprocessing (Feature encoding and scaling)
5. Machine Learning (XGBoost, LightGBM, CatBoost)
6. Hyperparameter Tuning with Optuna
7. Neural Networks
8. Ensemble Methods
9. Model Evaluation
10. Predict & Submit
Achievement: My ROC AUC score is 0.89399
Check out more details in my Medium blog
