import pandas as pd
from sklearn.ensemble import IsolationForest

def clean_data(file):
    df = pd.read_csv(file)
    
    # Konversi tanggal
    df['tanggal'] = pd.to_datetime(df['tanggal'])
    
    # Deteksi anomaly
    clf = IsolationForest(contamination=0.05)
    df['anomaly'] = clf.fit_predict(df[['demand', 'supply']])
    
    return df[df['anomaly'] == 1].drop('anomaly', axis=1)