This project helps a Germany Financial Institution identify new customers based on existing customers' demographic data and whole population of the Germany. The idea is to train an unsupervised learning model to identify the similarity and difference between the general population and the company’s customers based on geographical features provided. And then build a supervised learning model to identify potential new customers with the training data.
For this project I used Principle Component Analysis for feature deduction, KMeans for clustering and Xgboost for classification. I will talk about the details about Preprocessing and EDA, model selection and training, hyperparameter tuning in the following part.
Keywords: Customer Segmentation, PCA, KMeans, Unsupervised Learning