IBM-SPSS-Modeler-Create-a-predictive-model-to-predict-customer-churn

Get experience with IBM SPSS Modeler by creating a decision-tree machine-learning model to evaluate the risk that a customer might leave your service. This is a set by step tutorial. Credits to this demo https://www.ibm.com/cloud/garage/demo/try-spss-modeler/

Duration: 15 minutes

In this tutorial demo, you use IBM SPSS Modeler to build a machine-learning model to predict which customers might leave your service.

Task 1: Login to IBM Cloud and create Data Science Experince service

1. Login to your IBM Cloud account, if you don't have one already you can signup here. When you signup, make sure you choose your region as US-South.
2. Open the Catalog, click on Data & Analytics to refine search. Again make sure you choose your region as US-South.
3. Scroll down and click on Data Science Experince service.
4. Click on Create to create an instance of the service.
5. Click on Get started to open the tool.

Task 2: Create project

Once you have logged you you can go ahead and create a project.

1. Scrol down the page and click on (+) New project icon.
2. Name your project 'Predict customer churn'.
3. Scroll down, under Define storage click on Add to create an IBM Cloud Object Storage instance. The service will open up, choose the Lite plan then click Create.
4. Under Spark service click on Add to create an IBM Analytics for Apache Spark instance.The service will open up, choose the Lite plan then click Create.
5. Refresh page to make sure services are added.
6. Click Create, to finish creating your project.

Task 3: Upload data set

1. Download the data set
2. In your project page, nagivate to the Assests tab, drag the data set file you downloaded and drop it in the Load sidebar.
3. Your data set should be loaded successfully shorty!

Task 4: Create SPSS modele

1. Same page as before, scroll down to the Modeler flows.
2. Hit the (+) New flow icon
3. Under the 'New' tab, name your modeler 'Predictive model', make sure you chose IBM SSPS Modeler Runtime. Then click Create.

Task 5: Inspect the data set

You have a data set with customer data and churn data. The data engineer merged both data sets into one set. The data set is waiting for inspection on the canvas. In this task, you inspect the data set by using IBM SPSS Modeler.

1. Drag and drop your data set to the canvas.
2. Click Telco-Customer-Churn.csv. From the menu that opens, click Preview. The first 10 records of the data set are shown.
3. Scroll to inspect the right part of the data set. The last column, CHURN, contains data about whether a customer churned or not.
4. Click OK to close Preview window.

3. In the palette, click the Output tab.
4. Add the Data Audit node to the canvas by clicking Data Audit.
5. Connect the Data Audit node to the Telco-Customer-Churn.csv node by clicking Telco-Customer-Churn.csv. The Data Audit node is automatically renamed to "21 Fields."
6. Click 21 Fields. From the menu that opens, click Run. You can review key statistics and metrics for the data set. When you're finished, click OK to close the window.

Task 6: Prepare the data set

Get the data set ready for machine learning.

1. In the palette, click the Field Ops tab.
2. Add the Type node to the canvas by clicking Type.
3. Connect the Telco-Customer-Churn.csv node to the Type node by clicking Telco-Customer-Churn.csv.
4. Click Type. From the menu that opens, click Open.
5. In the right side bar, expland settings.
6. Click Configure types.
7. Set the measurement level of the columns by clicking Read Values. On the CustomerID row, click Input. From the menu that opens, click Record ID to change the role.
8. Scroll to the CHURN row.
9. Click CHURN and change the role from Input to Target. The CHURN row is used as the target to predict in your machine-learning model.
10. Click Apply and then click OK.

Task 7: Train the model

Train a C&R Tree model with your data set.

1. In the palette, click the Modeling tab.
2. Add the C&R Tree node to the canvas by clicking C&R Tree.
3. Connect the C&R Tree node to the Type node by clicking Type. The C&R Tree node is automatically renamed to "CHURN."
4. Click CHURN. From the menu that opens, click Run. The model is trained and a new CHURN node that looks like a golden nugget is added to the canvas.
5. From the output click on results. Review the model and notice which features are important predictors. When you're finished, click OK to close the window.

Task 8: Evaluate and visualize the model

Evaluate the model performance and visualize the model by using a gain chart.

1. In the palette, click the Output tab.
2. Add a Table node to the canvas by clicking Table.
3. Add an Analysis node to the canvas by clicking Analysis.
4. Connect the CHURN golden nugget node to the Table and Analysis nodes by clicking the CHURN golden nugget node.
5. Click the CHURN golden nugget node again. From the menu that opens, click Run from here.
6. In the side Output window that opens, you should see outputs of both Analysis and Table.
7. Click on the Table output to review results.
8. Scroll to the right and notice that two columns were added to the data set: $R-CHURN and $RC-CHURN. The $R-CHURN column is the prediction column. The $RC-CHURN column is the confidence level column. Click Predictive model to go back to the canvas.
9. Review the model performance, or accuracy, in the Analysis output. Click Predictive model to go back to the canvas.

10. In the palette, click the Graphs tab.
11. Add the EvEvaluationaluation node to the canvas by clicking Evaluation.
12. Connect the Evaluation node to the CHURN golden nugget node by clicking the CHURN golden nugget node.
13. Click R-CHURN. From the menu that opens, click Run. Review the gain chart. When you're finished, Click Predictive model to go back to the canvas.

Summary

You completed the tutorial. Congratulations!

In this tutorial, you completed these tasks:

1. Inspected a data set for customer churn

2. Prepared the data set for machine learning

3. Trained and evaluated a machine-learning model

4. Displayed a gain chart of the model

This tutorial scratches the surface of many of the powerful capabilities of IBM SPSS Modeler.