Real-time Credit Scoring with Feast on AWS

Overview

This tutorial demonstrates the use of Feast as part of a real-time credit scoring application.

• The primary training dataset is a loan table. This table contains historic loan data with accompanying features. The dataset also contains a target variable, namely whether a user has defaulted on their loan.
• Feast is used during training to enrich the loan table with zipcode and credit history features from a S3 files. The S3 files are queried through Redshift.
• Feast is also used to serve the latest zipcode and credit history features for online credit scoring using DynamoDB.

Requirements

• Terraform (v1.0 or later)
• AWS CLI (v2.2 or later)

Setup

Setting up Redshift and S3

First we will set up your data infrastructure to simulate a production environment. We will deploy Redshift, an S3 bucket containing our zipcode and credit history parquet files, IAM roles and policies for Redshift to access S3, and create a Redshift table that can query the parquet files.

Initialize Terraform

cd infra
terraform init

Set Terraform variables (used also in feature store config)

export TF_VAR_region="us-west-2"
export TF_VAR_project_name="your-project-name"
export TF_VAR_admin_password="$(openssl rand -base64 32)"

Make sure the Terraform plan looks good

terraform plan

Deploy your infrastructure

terraform apply

Once your infrastructure is deployed, you should see the following outputs from Terraform

redshift_cluster_identifier = "my-feast-project-redshift-cluster"
redshift_spectrum_arn = "arn:aws:iam::<Account>:role/s3_spectrum_role"
credit_history_table = "credit_history"
zipcode_features_table = "zipcode_features"

To have these outputs in env variables, you can source the env script

(cd .. && source env)

Next we create a mapping from the Redshift cluster to the external catalog

aws redshift-data execute-statement \
    --region "${TF_VAR_region}" \
    --cluster-identifier "${tf_redshift_cluster_identifier}" \
    --db-user admin \
    --database dev \
    --sql "create external schema spectrum from data catalog database 'dev' iam_role '${tf_redshift_spectrum_arn}' create external database if not exists;"

To see whether the command was successful, please run the following command (substitute your statement id)

aws redshift-data describe-statement --id [SET YOUR STATEMENT ID HERE]

You should now be able to query actual zipcode features by executing the following statement

aws redshift-data execute-statement \
    --region "${TF_VAR_region}" \
    --cluster-identifier "${tf_redshift_cluster_identifier}" \
    --db-user admin \
    --database dev \
    --sql "SELECT * from spectrum.zipcode_features LIMIT 1;"

which should print out results by running

aws redshift-data get-statement-result --id [SET YOUR STATEMENT ID HERE]

Return to the root of the credit scoring repository

cd ..

Setting up Feast

Install Feast using pip

pip install feast[aws]

We have already set up a feature repository in feature_repo/. It isn't necessary to create a new feature repository, but it can be done using the following command

feast init -t aws feature_repo # Command only shown for reference.

Since we don't need to init a new repository, all we have to do is configure the feature_store.yaml/ in the feature repository. Please set the fields under offline_store to the configuration you have received when deploying your Redshift cluster and S3 bucket.

Deploy the feature store by running apply from within the feature_repo/ folder

cd feature_repo/
feast apply

Registered entity dob_ssn
Registered entity zipcode
Registered feature view credit_history
Registered feature view zipcode_features
Deploying infrastructure for credit_history
Deploying infrastructure for zipcode_features

Next we load features into the online store using the materialize-incremental command. This command will load the latest feature values from a data source into the online store.

CURRENT_TIME=$(date -u +"%Y-%m-%dT%H:%M:%S")
feast materialize-incremental $CURRENT_TIME

Return to the root of the repository

cd ..

Train and test the model

Finally, we train the model using a combination of loan data from S3 and our zipcode and credit history features from Redshift (which in turn queries S3), and then we test online inference by reading those same features from DynamoDB

python run.py

The script should then output the result of a single loan application

loan rejected!

Interactive demo (using Streamlit)

Once the credit scoring model has been trained it can be used for interactive loan applications using Streamlit:

Simply start the Streamlit application

streamlit run streamlit_app.py

Then navigate to the URL on which Streamlit is being served. You should see a user interface through which loan applications can be made: