Abandonment Prediction, Powtoon

Areej Hatem Eweida

Data Science Fellows June 2020 Cohort

Abstract

in the project we aim to accomplish the following:

1. Identify and visualize which factors contribute to customer abandonment:
2. Build a prediction model that will perform the following:
   • Classify if a customer is going to abandon or not.
   • Preferably and based on model performance, choose a model that will attach a probability to the churn to make it easier for customer service to target low hanging fruits in their efforts to prevent churn.
   • Come up with interesting insights and findings regarding why a customer would abandon to help improve the product.

Challenges (at least two)

1. Working remotely especially at  first when I wasn’t very familiar with the data and the frameworks.
2. Dealing with a large amount of data for the first time.

Achievements (according to KPIs)

1. Analyzed the business case, determined the required variables, and used advanced queries in order to create the database.
2. Experienced with common machine learning algorithms like SVM, random forest and K-means and different neural network architectures.
3. Implemented preprocessing techniques for models to reduce loss and performed grid search for a better choice of hyperparameters. Reached test accuracy of unseen data above 85%.
4. Helped the company develop strategic plans based on predictive modeling and findings. 

Future project development 

1. The current model could be used by the product team for different applications.
2. Further improvements can be done with the predictive model like collecting more features of the users and feed the model with more data.