Creating Customer Segments

Question 1
Selecting Samples

Three separate samples of the data are chosen and their establishment representations are proposed based on the statistical description of the dataset.

Question 2
Feature Relevance

A prediction score for the removed feature is accurately reported. Justification is made for whether the removed feature is relevant.

Question 3
Feature Distributions

Student identifies features that are correlated and compares these features to the predicted feature. Student further discusses the data distribution for those features.

Feature Scaling

Feature scaling for both the data and the sample data has been properly implemented in code.

Question 4
Outlier Detection

Student identifies extreme outliers and discusses whether the outliers should be removed. Justification is made for any data points removed.

Question 5
Principal Component Analysis

The total variance explained for two and four dimensions of the data from PCA is accurately reported. The first four dimensions are interpreted as a representation of customer spending with justification.

Dimensionality Reduction

PCA has been properly implemented and applied to both the scaled data and scaled sample data for the two-dimensional case in code.

Question 6
Clustering Algorithm

The Gaussian Mixture Model and K-Means algorithms have been compared in detail. Student’s choice of algorithm is justified based on the characteristics of the algorithm and data.

Question 7
Creating Clusters

Several silhouette scores are accurately reported, and the optimal number of clusters is chosen based on the best reported score. The cluster visualization provided produces the optimal number of clusters based on the clustering algorithm chosen.

Question 8
Data Recovery

The establishments represented by each customer segment are proposed based on the statistical description of the dataset. The inverse transformation and inverse scaling has been properly implemented and applied to the cluster centers in code.

Question 9
Sample Predictions

Sample points are correctly identified by customer segment, and the predicted cluster for each sample point is discussed.

Question 10
A/B Test

Student correctly identifies how an A/B test can be performed on customers after a change in the wholesale distributor’s service.

Question 11
Predicting Additional Data

Student discusses with justification how the clustering data can be used in a supervised learner for new predictions.

Question 12
Comparing Customer Data

Comparison is made between customer segments and customer ‘Channel’ data. Discussion of customer segments being identified by ‘Channel’ data is provided, including whether this representation is consistent with previous results.