Implements a machine learning classifier to predict customer churn in the telco industry based on customer data. Read the full report in ["Capstone Churn Prediction Report Manuel Seeger.pdf"]("Capstone Churn Prediction Report Manuel Seeger.pdf")
Run capstone.ipynb to reproduce the results of the project.
At the beginning of the notebook there are a number of global parameters that control performance intensive tasks. Adjust depending on your host machine.
Required dataset is included in subfolder data in cell2celltrain.csv.
- Python 3.5 or greater
- Required Python libraries are stated in requirements.txt
- Subfolder global_objectives needs to be in the python path of the main notebook to load the custom loss layers. This code is a copy of the Tensorflow reserch module global_objectives.
- utils.py contains convenience functions for the main notebook
- The project uses tensorflow modules that won't be included in TF 2.0. Run with TF 1.x
No further dependencies required to run the Notebook.
requirements.txt
scikit-learn
pandas
numpy
keras
xgboost
tensorflow
matplotlib
imbalanced-learn
scipy