Credit Default Risk Prediction with Anaconda on IBM Z & LinuxONE
Financial organizations around the world face the constant challenge to accurately predict the risk of credit default. This covers all aspects of financial lending, including personal, corporate and public finance. To enhance and automate this process, great interest has been placed on the usage of data science and machine learning to predict the customers most likely to default.
These lending processes mainly run on an institution's core business systems, with IBM Z & LinuxONE providing the industry standard platform for security, resiliency and scalability. With Anaconda on IBM Z & LinuxONE a customer can run their machine learning model co-located to their key processes, without the need to move data off the platform, ensuring security and data currency.
To that end, we have created an example credit default model leveraging python and conda with Linux on Z. We performed data processing and the required transformations to create a Logistic Regression model using scikit-learn.
Model creation is inspired by source : https://github.com/ibmsoe/snap-ml/blob/master/notebooks/credit-default-prediction-example.ipynb
-
Ensure you follow the installation instructions here provided by Anaconda to setup your data science environment as per the architecture you are on.
-
Create a conda environment with the following frameworks installed by running the following commands
conda create -n credit-default --force -y scikit-learn pandas cython jpeg jupyter matplotlib conda activate credit-default -
Start your Jupyter Notebook and launch via web browser using ip address of the IBM Z machine where notebook up and running ie. http://ip-addr:8888/
jupyter notebook --ip "0.0.0.0" --port 8888
NOTE Make sure that port on which jupter notebook running exposed on the IBM Z machine so that it can be accessed via external browser if needed.
- Import the notebook using
uploadoption in the Jupyter Notebook web browser