File size: 1,508 Bytes
2a3c5b3 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 | import pandas as pd
import logging
from schemas.patient import Patient
from sklearn.pipeline import Pipeline
from typing import Optional
logger = logging.getLogger(__name__)
def preprocess_input(patient_data: Patient, feature_creator: Optional[Pipeline]=None, preprocessor: Optional[Pipeline]=None):
"""
Apply feature engineering (pipeline) and scaling (preprocessor).
"""
# Convert Pydantic model to DataFrame
input_df = pd.DataFrame([patient_data.model_dump(exclude={'include_shap'})])
# 1. Pipeline (Feature Engineering)
if feature_creator:
try:
# Assuming pipeline transforms dataframe -> dataframe or array
# If pipeline is just a transformer, we call transform
# In some setups, you might have custom steps.
# We'll assume it returns a DataFrame or something compatible with the preprocessor
input_df = feature_creator.transform(input_df)
except Exception as e:
logger.error(f"Feature creator pipeline transformation failed: {e}")
raise e
# 2. Preprocessor (Scaling/Encoding)
if preprocessor:
try:
# Preprocessor usually returns numpy array
processed_data = preprocessor.transform(input_df)
return processed_data
except Exception as e:
logger.error(f"Preprocessor transformation failed: {e}")
raise e
return input_df.values # Fallback if no artifacts
|