Section III Part 3: Customer Segmentation Prediction (Classification)

[Jen-uis]

We used regression and regressor in the previous example, in this analysis, we are going to use classification to predict customer segments that is listed in train_data. However, this approach cannot be given an accuracy score as the test_data does not provide any segments to verify the accuracy. We are going to create a new column under test_data to see which customers are categorized in to which segments. Lets go:

Use of techniques: Logistic Regression / Random Forest Classifier / Gradient Boosting Classifier (Can possibly use Support Vector Machine, but we will verify how the first three techniques perform)

[Jen-uis]

Completed Customer-Segmentation-Prediction using Three methods:

1. Logistic Regression
2. Random Forest Classifier
3. Gradient Boosting Classifier

For more information, check the latest updated Project-Code.ipynb.

[Jen-uis]

The results are in LogReg_Segmentation RFC_Segmentation GBC_Segmentation A 713 629 718 B 244 568 472 C 783 638 609 D 887 792 828

Seems like the consistency is very low. Looking for ways to test which methods can be trusted.

[Jen-uis]

Update: I used Cross-Validation methods to test which methods can achieve a higher score in predicting the target variable. Results are:	Logistic Regression CV	Random Forest Classifier CV	Gradient Boosting Classifier CV
0.4954	0.4857	0.534

It seems like Gradient Boosting Classifier can return somewhat good results to predict the segmentation of a new customer.

[Jen-uis]

done with final review