Introduction
Something that we’re yet to explore in CRMA with Workout Wednesday is using Einstein Discovery to create predictive models and this week we’re going to be doing just that. Using a training dataset from a 2020 Kaggle challenge, we’re going to try and predict whether a customer will churn or not from a telecommunications provider. Full details are provided below.
Requirements
- Upload both the train.csv and test.csv datasets to your CRM instance
- Create a model using Einstein Discovery to minimize customer churn. To create the model you should use the train.csv dataset. How you go about creating the model is up to you!
- Consider which fields to be used to train your model. For example, is there any collinearity in our dataset that we should be considering?
- Once you have optimized your model (it might take a few versions!), deploy your model for use on your Salesforce data.
- Use the model to predict churn values in your test.csv dataset. If deployed successfully, you should see that you have run 750 predictions when viewing your model in CRMA.
To validate you have predicted churn in your test.csv dataset, you should see a view that looks similar to below. I have also added a screenshot of the window you will see when evaluating whether your model is ready to be deployed.
Tipps…
Dataset
This week uses the data from the 2020 Kaggle Customer Churn challenge, you can access both the training and test dataset here: https://www.kaggle.com/competitions/customer-churn-prediction-2020/overview
Share
After you finish your workout, share a screenshot of your solutions or interesting insights on Twitter or LinkedIn using the hashtags #WOW2023 and #CRMA and tag @genetis, @PreenzJ, @LaGMills and @JaackParry. (Or you can use this handy link to do that)
Also make sure to fill out the Submission Tracker to track your progress and help us judge the difficulty of our challenges.