barbking/machine_learning_project_ust_763

Machine learning project for class

Leveraging Customer Data for Targeted Bank Marketing

A Machine Learning Model using Portuguese Bank Customer Data

This was a final project for a graduate school machine learning class at the University of St. Thomas. Our three-member group divided up algorithm analysis and these are the algorithms I analyzed with our data set. Initially, I started with on my own with EDA/feature selection/encoding, but then we decided as a group to have one uniform process for our final comparison. There are two sets of notebooks for some algorithms, one with our group data prep and one where I experimented with other methods.

• Started data exploration/visualization of numerical/categorical features, could remove some features highly correlated to each other (seaborn/matplotlib). Encode categorical data if needed for the algorithm (scikit learn label encoding/one hot encoding).
• One issue with the data set was an imbalanced class label (88% did not buy cd), used the SMOTE algorithm(Synthetic Minority Oversampling Technique) to create synthetic values instead of making copies of data.
• My responsibility for the group was looking into decision tree/random forest/lightGBM algorithms.
• Optimized hyperparameter values using grid search.
• Tried with and without PCA (principal component analysis).
• Results/key features: (economic indicator (Euribar), age, pdays(number of days since last contact), month (dec/march).

Technologies used: python, scikit-learn, pandas, numpy, matplotlib, seaborn, lightGBM