A mock bank dataset to model customer loyalty.
The goal is to find quantifiable reasons that increase bank customers leaving. Immediately, modeling with XGBoost or another machine learning model becomes a useful direction, provided model interpretability is adequate. This is a perfect job for SHAP values. With a few simple lines of code, we've trained a model, measured the accuracy of the model, and visualized the importance of each variable in the model's predictive capability.
A number of variables stick out as important, and when put through a correlational analysis in Tableau, the top SHAP values are confirmed as having higher correlations than other variables.
While the boosting in XGBoost does happen in an ensemble of decision trees, the feature importance and correlation analysis gives us a solid framework to build our own decision tree from scratch.
Lessons Learned:
This cost-benefit analysis is in a context of a fictional, globally operating movie rental business on the brink of expansion. Its primary goal is to employ strategic decisions that result in an augmentation of overall profitability.
Learn more
This is a five-decade examination of the interconnection between the Global Women's Empowerment Index and GDP Per Capita. The central proposition suggests a mutually reinforcing association between economic well-being and Women's Rights. Nevertheless, our unwavering dedication is to adopt a rigorous, data-centric approach, unveiling noteworthy discoveries.
Learn more