Customer Transaction Prediction
Hi, welcome to my project! Feel free to check out my github page to learn more
https://github.com/vvwzy/SantanderCustomerTransactionPrediction
Throughout this project, I was trying to find out if there is a way to determine which customers will make a specific transaction in the future, irrespective of the amount of money transacted.
The Python Libraries that have been used for this project including Numpy, Pandas, sklearn, scipy, skopt, matplotlib, xgboost, lightgbm, tensorflow and bayes_opt.
After understanding the business background and the problem statement, I started exploring the dataset doing basic EDA. There were 200k observations and 200 features in the training dataset and same size for the testing data. Considering all these 200 features, I needed to find the most relevant features that help to make better predictions.
The next step was feature engineering since the features that went into the model are crucial for training a good prediction model. Then I checked for missing values and repeated values and found that there wass none of them.
After the above, I started to try different models including Logistic Regression, Random Forest, XGBoost and LightGBM. After comparing these models, I decided to stick with LightGBM. Finally after all algorithms have been validated, trained and tested, I ended up using LightGBM model with the help of Bayesian Optimization to pick the best hyperparameters and Stratified K Fold to elevate the AUC score on our validation dataset. After the model being built, by plotting the feature importance among all the features I had, I knew the weighted contribution of each feature to final predictions.