model_script/t.py

import numpy as np
import pandas as pd

from client import ODPSClient

recall_result = "/Users/zhao/Desktop/20241124_recall.csv"
day_oss = "/Users/zhao/Desktop/20241124_day_oss.csv"

odps = ODPSClient.ODPSClient()


def read_day_recall_v2() -> pd.DataFrame:
    df = pd.read_csv(day_oss)
    df['cpm_rank'] = df.groupby("type")['score'].rank(method='first', ascending=False).astype(int)
    df['cpm_rank_2'] = df.groupby("type")['cpm'].rank(method='first', ascending=False).astype(int)

    df['view_rank'] = df.groupby("type")['view_rate'].rank(method='first', ascending=False).astype(int)

    df['day_recall_v2'] = np.where(
        ((df['type'] == '14d') & ((df['cpm_rank'] <= 30) | (df['cpm_rank_2'] <= 20) | (df['view_rank'] <= 30))) |
        ((df['type'] == '3d') & ((df['cpm_rank'] <= 50) | (df['cpm_rank_2'] <= 30) | (df['view_rank'] <= 50))) |
        ((df['type'] == '1d') & ((df['cpm_rank'] <= 80) | (df['cpm_rank_2'] <= 50) | (df['view_rank'] <= 100))),
        True,
        False
    )
    df.to_csv("/Users/zhao/Desktop/3.csv", index=False)

    grouped_df = (
        df.groupby('cid', as_index=False)  # 按 CID 分组
        .agg(day_recall_v2=('day_recall_v2', 'any'))  # 只要有一个为 True，就为 True
    )

    return grouped_df


def read_day_recall() -> pd.DataFrame:
    df = pd.read_csv(day_oss)
    df['cpm_rank'] = df.groupby("type")['score'].rank(method='first', ascending=False).astype(int)

    df['view_rank'] = df.groupby("type")['view_rate'].rank(method='first', ascending=False).astype(int)

    df['day_recall_v1'] = np.where(
        ((df['type'] == '14d') & ((df['cpm_rank'] <= 30) | (df['view_rank'] <= 20))) |
        ((df['type'] == '3d') & ((df['cpm_rank'] <= 50) | (df['view_rank'] <= 30))) |
        ((df['type'] == '1d') & ((df['cpm_rank'] <= 80) | (df['view_rank'] <= 50))),
        True,
        False
    )

    df.to_csv("/Users/zhao/Desktop/2.csv", index=False)

    grouped_df = (
        df.groupby('cid', as_index=False)  # 按 CID 分组
        .agg(day_recall_v1=('day_recall_v1', 'any'))  # 只要有一个为 True，就为 True
    )

    return grouped_df


def _main():
    day_recall = read_day_recall()

    day_recall_v2 = read_day_recall_v2()

    recall = pd.read_csv(recall_result)
    recall['base_diff'] = recall['产品-17-前五组'] - recall['产品-35-前五组']

    recall = (pd.merge(recall, day_recall, on='cid', how='left')
              .merge(day_recall_v2, on='cid', how='left'))

    recall.to_csv("/Users/zhao/Desktop/1.csv", index=False)

    print(recall)


if __name__ == '__main__':
    _main()