Customer segmentation

This project is the third of a series of seven projects to be delivered as part of the Udacity DataScience Nanodegree self-study.

Context

In this project, we are given real-life data provided by a german company (note that as those data are confidential they are not provided within the repository).
The data concerns a company that performs mail-order sales in Germany. Their main question of interest is to identify facets of the population that are most likely to be purchasers of their products for a mailout campaign.

Objective

Our job is to use unsupervised learning techniques to organize the general population into clusters, then use those clusters to see which of them comprise the main user base for the company.
But, as there are many features, prior to applying the machine learning methods, we need to assess and clean the data in order to convert the data into a usable form and then apply Principal Component Analysis (PCA) to find the vectors of maximal variability.

Finally, on our transformed data, we can apply clustering techniques to identify groups in the general demographic data. And then we are able to apply the same clustering model to the customers dataset to see how market segments differ between the general population and the mail-order sales company. K-Means method is used to cluster the demographic data into groups.

For more details, please refer to my Github repository for this project.