Cloud Cost Modeling and Planning Questions

import pandas as pd, json, logging
from dateutil import parser
def load_csv(path):
    df = pd.read_csv(path, parse_dates=['date'], dtype={'account_id':str,'resource_id':str,'service':str,'amount_usd':float})
    return df
def extract_owner(tag_str):
    if pd.isna(tag_str): return None
    try:
        j=json.loads(tag_str)
        return j.get('owner') or j.get('Owner')
    except Exception:
        return None
# ... apply_fallbacks, normalize_month, aggregate_monthly, compute_mom, detect_top_contributors ...

# libraries: pandas, scipy, scikit-learn, prophet, sqlalchemy, boto3
def ingest(records): pass  # read from S3/CloudStorage, Kinesis/PubSub
def normalize(df): pass    # unify fields, parse tags, IPs, timestamps
def baseline(df):
    # rolling median + Prophet seasonal model per (account, service, resource)
    return baseline_map
def detect(df, baseline_map):
    # rule: absolute > X and pct > Y
    # statistical: z-score > 3 or Prophet forecast residuals > 3 sigma
    return anomalies
def attribute(anoms):
    # heuristics: tags -> account owner, name regex -> team, CIDR -> VPC owner, IAM lookup
    return attributed
def prioritize(attributed):
    # score = impact * confidence * recency
    return sorted_actions