Credit Scoring

Determine the riskiness of a credit applicant


How it Works

Click the orange button to run a loan applicant’s credit information through a risk analysis model. The model is written in the statistical programming language Python, while the web application you can interact with is written in Javascript.

In order to include the Python model in the “incompatible” application,” the model is “hosted” on ScienceOps. The website “calls” to the model on ScienceOps to “ask” for its prediction about the riskiness of the loan applicant. ScienceOps can process hundreds of thousands of these predictions per second.


Application Integration with ScienceOps

This model takes users loan application information and creates a credit risk score card!

  • A trained Python model using a logistic regression is deployed to ScienceOps
  • The model is integrated into a credit scoring application using the ScienceOps API
  • Credit applicants data is POST'ed to the model API and predictions are returned

The Integration

curl -X POST --user username:1234567890abcdefg \
  --data '{
    "last_fico_range_low": 605,
    "last_fico_range_high": 632,
    "home_ownership": "MORTGAGE",
    "revolving_line_utilization": 10,
    "credit_inquiries_past_6m": 1
  }' \
  https://sandbox.yhathq.com/demo/models/LendingClub/
 98   glm = LogisticRegression()
 99   glm.fit(df_term[features], df_term.is_bad)
100   glm.predict_log_proba(df_term[features].head())
101   ...
78   # basic GLM
79   my.glm <- glm(I(is_bad==TRUE) ~ last_fico_range_low + last_fico_range_high +  is_rent,
80   data=train, na.action=na.omit, family=binomial())
81   ...
Checkout the rest of the code!
Contact
45 Main St #707,
Brooklyn, NY 11201
info@yhathq.com
+1 718 855 2107
+49 15735983455
Learn More
Connect With Us
Newsletter