Skip to main content
The preprocessor module handles loading and preprocessing credit risk data for model training and inference.

​
Functions

​
load_data

def load_data(filepath: str) -> pd.DataFrame
Loads a dataset from the specified file path.
​
filepath
str
required
Path to the CSV file containing the credit risk dataset
​
return
pd.DataFrame
DataFrame containing the loaded dataset
import pandas as pd
from processing.preprocessor import load_data

# Load the dataset
df = load_data("datasets/credit_score_dataset/german_credit_risk.csv")
print(df.head())

​
preprocess_data

def preprocess_data(
    df: pd.DataFrame,
    target_column: str = "Risk",
    save_path: str = None
)
Preprocesses the credit risk dataset by handling missing values, encoding categorical features, and splitting data into training and test sets.
​
df
pd.DataFrame
required
Input DataFrame containing the credit risk data
​
target_column
str
default:"Risk"
Name of the target column. The function expects values “good” (encoded as 1) and “bad” (encoded as 0)
​
save_path
str
Path to save the fitted preprocessor pipeline. If provided, the preprocessor will be saved as a joblib file
​
return
tuple
Returns a tuple of (X_train, X_test, y_train, y_test) containing the preprocessed and split data:
  • X_train: Training features (80% of data)
  • X_test: Test features (20% of data)
  • y_train: Training labels
  • y_test: Test labels
The function automatically handles:
  • Removing residual index columns (Unnamed: 0)
  • Encoding target variable: good → 1, bad → 0
  • Imputing missing values (mean for numerical, “unknown” for categorical)
  • Standardizing numerical features
  • One-hot encoding categorical features
  • 80/20 train-test split with random_state=42

​
Preprocessing Pipeline

The preprocessing pipeline applies different transformations to numerical and categorical features:

​
Numerical Features

Processed features: Age, Credit amount, Duration
  1. SimpleImputer: Imputes missing values using the mean
  2. StandardScaler: Normalizes features by removing the mean and scaling to unit variance

​
Categorical Features

Processed features: Sex, Job, Housing, Saving accounts, Checking account, Purpose
  1. SimpleImputer: Imputes missing values with “unknown”
  2. OneHotEncoder: Converts categorical variables into binary vectors with handle_unknown="ignore"
import pandas as pd
from processing.preprocessor import load_data, preprocess_data

# Load the dataset
df = load_data("datasets/credit_score_dataset/german_credit_risk.csv")

# Preprocess and split the data
X_train, X_test, y_train, y_test = preprocess_data(
    df,
    target_column="Risk",
    save_path="models/preprocessor.joblib"
)

print(f"Training set size: {X_train.shape}")
print(f"Test set size: {X_test.shape}")

​
Source Code Reference

View the complete implementation at processing/preprocessor.py:22-106

Build docs developers (and LLMs) love

Get started for freeTalk to us