Adding new features

1. Adding New Features

Adding additional features or data sources is an easy way to improve model performance. Even if the improvement results in a few percentage points, this can add great value when solving a multi-million-dollar business problem such as churn.

2. Additional Data Sources

Customer touch points are important sources of data to incorporate in your churn modeling workflow. These sources provide a 360 degree view of the customer and should be included in your list of features. When searching for touch points, imagining your personal experience as a cell phone customer is a great asset. Data sources that you could incorporate into your model include: customer service call notes, web logs, email campaigns, social networks, transactions, and cell tower signal strength. Including these data sources can greatly improve model performance.

3. Churn Features

In the upcoming exercises you will add additional features to your previous random forest and re-run your churn model. You are doing this to potentially avoid underfitting by providing additional information to train the model with. Features that you will use are: Region Code, Total Charges, Total Minutes, Minutes Per Call, Cost Per Call, and Total Calls. After training your Random Forest that includes these features you will validate the model and compare it to the model you previously built.

4. Model Improvement

Using the knowledge you gained, you will create an ROC Curve that compares both models that includes the calculated AUC. Once you render your ROC Curve, you will be able to compare both models and whether the additional gains are worth the resources to capture them.

5. Benefits | Costs

Working with the business to identify additional data sources provides expertise to the churn problem you are trying to solve. This is also a good time to discuss with them the benefits and costs of incorporating the additional features and sources in your model. Benefits include: improved ROI, decreased costs, and increased performance. These benefits are important that could potentially return millions to the business, however costs need to be weighed as well to ensure the benefits are greater. Increased model complexity, resources, and time to operationalize are some of the costs to consider when deciding to add additional features and data sources.

6. Let's practice!

Now you get to re-fit your model and validate the results using additional features.

Create Your Free Account

By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.