Skip to main content
POST
/
credit_score_prediction
POST /credit_score_prediction
curl --request POST \
  --url https://api.example.com/credit_score_prediction \
  --header 'Content-Type: application/json' \
  --data '
{
  "Age": 123,
  "Sex": "<string>",
  "Job": "<string>",
  "Housing": "<string>",
  "Saving accounts": "<string>",
  "Checking account": "<string>",
  "Credit amount": 123,
  "Duration": 123,
  "Purpose": "<string>"
}
'
{
  "200": {},
  "422": {},
  "500": {},
  "prediction": "<string>",
  "probability": 123
}

​
Overview

This endpoint predicts whether a customer’s credit risk is “good” or “bad” based on their demographic and financial information. The prediction is made using a PyTorch deep learning model that has been trained on historical credit data.

​
Endpoint

POST /credit_score_prediction

​
Request Schema

The request body must be a JSON object with the following fields:

​
Required Fields

​
Age
integer
required
Age of the applicant in years. Must be greater than 0.Example: 35
​
Sex
string
required
Gender of the applicant.Allowed values:
  • "male"
  • "female"
Example: "male"
​
Job
string
required
Skill level of the applicant’s job.Allowed values:
  • "unskilled and non-resident"
  • "unskilled and resident"
  • "skilled"
  • "highly skilled"
Example: "skilled"
​
Housing
string
required
Type of housing for the applicant.Allowed values:
  • "own" - Owns housing
  • "rent" - Rents housing
  • "free" - Free housing
Example: "own"
​
Saving accounts
string
required
Status of the savings account. Note the space in the field name.Allowed values:
  • "NA" - No savings account
  • "little" - Little savings
  • "moderate" - Moderate savings
  • "quite rich" - Quite rich savings
  • "rich" - Rich savings
Example: "NA"
​
Checking account
string
required
Status of the checking account. Note the space in the field name.Allowed values:
  • "NA" - No checking account
  • "little" - Little balance
  • "moderate" - Moderate balance
  • "rich" - Rich balance
Example: "little"
​
Credit amount
number
required
Amount of credit requested. Must be greater than 0.Example: 9055.0
​
Duration
integer
required
Duration of the credit in months. Must be greater than 0.Example: 36
​
Purpose
string
required
Purpose of the credit.Allowed values:
  • "car"
  • "furniture/equipment"
  • "radio/TV"
  • "domestic appliances"
  • "repairs"
  • "education"
  • "business"
  • "vacation/others"
Example: "education"

​
Response Schema

The response is a JSON object with the following fields:
​
prediction
string
required
The predicted credit risk category.Possible values:
  • "good" - Good credit risk
  • "bad" - Bad credit risk
​
probability
number
required
The probability that the credit risk is “good”. Value between 0 and 1.Range: 0.0 to 1.0Note: Higher values indicate higher confidence in a “good” prediction.

​
Example Request

{
  "Age": 35,
  "Sex": "male",
  "Job": "unskilled and resident",
  "Housing": "free",
  "Saving accounts": "NA",
  "Checking account": "NA",
  "Credit amount": 9055,
  "Duration": 36,
  "Purpose": "education"
}

​
Example Response

{
  "prediction": "good",
  "probability": 0.8542
}

​
Response Status Codes

​
200
OK
Prediction completed successfully
​
422
Unprocessable Entity
Invalid input data - validation failed
​
500
Internal Server Error
Error during model inference

​
Inference Pipeline

The prediction process follows these steps:
1

Input Validation

The request data is validated against the CreditRiskInput Pydantic schema to ensure all fields are present and have correct types and values.
2

Data Preprocessing

The input is transformed into a pandas DataFrame and processed using a scikit-learn preprocessor that handles:
  • Categorical encoding
  • Feature scaling
  • Feature engineering
3

Model Inference

The preprocessed data is converted to a PyTorch tensor and passed through the deep learning model:
  • The model outputs probabilities for both “bad” and “good” classes
  • The prediction is determined by the class with the highest probability
4

Response Generation

The prediction result and probability are packaged into a CreditRiskOutput response object and returned as JSON.

​
Implementation Details

​
Important Notes

Field Names with Spaces: The fields "Saving accounts" and "Checking account" contain spaces. Ensure you include the space when making requests, or the validation will fail.
Singleton Pattern: The model is loaded once at startup using a singleton pattern. All prediction requests share the same model instance for optimal performance.
CPU Inference: The model runs on CPU by default. For high-throughput production deployments, consider configuring GPU support.

API Overview

Learn about the API architecture

Interactive Docs

Try the API in Swagger UI

Build docs developers (and LLMs) love

Get started for freeTalk to us