Capstone Project, Udacity Data Science Nanodegree
Github Repo:https://github.com/cdumen/DataScientist_Capstone
Sparkify is a music streaming service just as Spotify and Pandora. The data provided is the user log of the service, having demographic info, user activities, timestamps and etc. We try to analyze the log and build a model to identify customers who are highly likely to quit using our service, and thus, send marketing offers to them to prevent them from churning. We use F1 score to measure of model performance because we need precision and recall at the same time as we don't want to miss too many customers who are likely to churn whilst we don't want to waste too much on those who are not likely to churn. The model we built has a F1 score of nearly 0.800
Sprakify .ipynb is the main file of the project, it demonstrates the process of using pyspark to explore the data and build the model.
- ETL
- Define customer churn and EDA
- Feature engineering
- Modeling
- Evaluation
- Feature import analysis