predictive-maintenance-pipeline/scripts/07_clean_data.py

#!/usr/bin/env python3
"""Clean and prepare data"""
import os
import logging
import pandas as pd
from huggingface_hub import hf_hub_download

logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

HF_TOKEN = os.getenv("HF_TOKEN")
HF_USERNAME = os.getenv("HF_USERNAME", "SharleyK")
DATASET_NAME = os.getenv("DATASET_NAME", "PredictiveMaintenance")
repo_id = f"{HF_USERNAME}/{DATASET_NAME}"

logger.info("Cleaning data...")

# Download data
file_path = hf_hub_download(repo_id=repo_id, repo_type="dataset",
                             filename="engine_data.csv", token=HF_TOKEN)

df = pd.read_csv(file_path)

logger.info(f"Original shape: {df.shape}")
logger.info(f"Original columns: {df.columns.tolist()}")

# Standardize column names to lowercase with underscores
# This ensures consistency regardless of how they're named in the source
df.columns = df.columns.str.lower().str.replace(' ', '_')
logger.info(f"Standardized columns: {df.columns.tolist()}")

# Verify the target column exists
if 'engine_condition' not in df.columns:
    logger.error(f"Target column 'engine_condition' not found after standardization!")
    logger.error(f"Available columns: {df.columns.tolist()}")
    raise KeyError("Missing expected target column")

# Remove duplicates
initial_rows = df.shape[0]
df = df.drop_duplicates()
logger.info(f"After removing duplicates: {df.shape} (removed {initial_rows - df.shape[0]} rows)")

# Handle missing values (if any)
initial_rows = df.shape[0]
df = df.dropna()
logger.info(f"After dropping NA: {df.shape} (removed {initial_rows - df.shape[0]} rows)")

# Log target distribution
logger.info(f"Target distribution:
{df['engine_condition'].value_counts()}")

# Save cleaned data
os.makedirs("data", exist_ok=True)
df.to_csv("data/cleaned_data.csv", index=False)

logger.info("✓ Data cleaning completed!")
logger.info(f"✓ Final columns: {df.columns.tolist()}")