jiangcang_uo/data_process.py

import pandas as pd
import numpy as np
import gc
import os


def preprocess_data_simple(df_input, is_train=False):

    print(">>> 开始数据预处理")
    # 城市码映射成数字    
    
    # gid：基于指定字段的分组标记（整数）
    df_input['gid'] = (
        df_input
        .groupby(
            ['citypair', 'flight_numbers', 'from_date'],    # 'baggage_weight' 先不进分组
            sort=False
        )
        .ngroup()
    )

    # 在 gid 与 baggage_weight 内按时间降序
    df_input = df_input.sort_values(
        by=['gid', 'baggage_weight', 'hours_until_departure'],
        ascending=[True, True, False]
    ).reset_index(drop=True)

    df_input = df_input[df_input['hours_until_departure'] <= 480]
    df_input = df_input[df_input['baggage_weight'] == 20]   # 先保留20公斤行李的

    # 在hours_until_departure 的末尾 保留真实的而不是补齐的数据
    if not is_train:
        _tail_filled = df_input.groupby(['gid', 'baggage_weight'])['is_filled'].transform(
            lambda s: s.iloc[::-1].cummin().iloc[::-1]
        )
        df_input = df_input[~((df_input['is_filled'] == 1) & (_tail_filled == 1))]

    # 价格变化最小量阈值
    price_change_amount_threshold = 5
    df_input['_raw_price_diff'] = df_input.groupby(['gid', 'baggage_weight'], group_keys=False)['price_total'].diff()

    # 计算价格变化量
    df_input['price_change_amount'] = (
        df_input['_raw_price_diff']
        .mask(df_input['_raw_price_diff'].abs() < price_change_amount_threshold, 0)
        .replace(0, np.nan)
        .groupby([df_input['gid'], df_input['baggage_weight']], group_keys=False)
        .ffill()
        .fillna(0)
        .round(2)
    )

    # 计算价格变化百分比（相对于上一时间点的变化率）
    df_input['price_change_percent'] = (
        df_input.groupby(['gid', 'baggage_weight'], group_keys=False)['price_total']
        .pct_change()
        .mask(df_input['_raw_price_diff'].abs() < price_change_amount_threshold, 0)
        .replace(0, np.nan)
        .groupby([df_input['gid'], df_input['baggage_weight']], group_keys=False)
        .ffill()
        .fillna(0)
        .round(4)
    )

    # 第一步：标记价格变化段
    df_input['price_change_segment'] = (
        df_input.groupby(['gid', 'baggage_weight'], group_keys=False)['price_change_amount']
        .apply(lambda s: (s != s.shift()).cumsum())
    )

    # 第二步：计算每个变化段内的持续时间
    df_input['price_duration_hours'] = (
        df_input.groupby(['gid', 'baggage_weight', 'price_change_segment'], group_keys=False)
        .cumcount()
        .add(1)
    )

    # 可选：删除临时列
    df_input = df_input.drop(columns=['price_change_segment', '_raw_price_diff'])

    # 训练过程
    if is_train:
        df_target = df_input[(df_input['hours_until_departure'] >= 24) & (df_input['hours_until_departure'] <= 360)].copy()
        df_target = df_target.sort_values(
            by=['gid', 'baggage_weight', 'hours_until_departure'],
            ascending=[True, True, False]
        ).reset_index(drop=True)

        # 对于先升后降的分析
        prev_pct = df_target.groupby(['gid', 'baggage_weight'], group_keys=False)['price_change_percent'].shift(1)
        prev_amo = df_target.groupby(['gid', 'baggage_weight'], group_keys=False)['price_change_amount'].shift(1)
        prev_dur = df_target.groupby(['gid', 'baggage_weight'], group_keys=False)['price_duration_hours'].shift(1)
        prev_price = df_target.groupby(['gid', 'baggage_weight'], group_keys=False)['price_total'].shift(1)
        drop_mask = (prev_pct > 0) & (df_target['price_change_percent'] < 0)

        df_drop_nodes = df_target.loc[drop_mask, ['gid', 'baggage_weight', 'hours_until_departure']].copy()
        df_drop_nodes.rename(columns={'hours_until_departure': 'drop_hours_until_departure'}, inplace=True)
        df_drop_nodes['drop_price_change_percent'] = df_target.loc[drop_mask, 'price_change_percent'].astype(float).round(4).to_numpy()
        df_drop_nodes['drop_price_change_amount'] = df_target.loc[drop_mask, 'price_change_amount'].astype(float).round(2).to_numpy()
        df_drop_nodes['high_price_duration_hours'] = prev_dur.loc[drop_mask].astype(float).to_numpy()
        df_drop_nodes['high_price_change_percent'] = prev_pct.loc[drop_mask].astype(float).round(4).to_numpy()
        df_drop_nodes['high_price_change_amount'] = prev_amo.loc[drop_mask].astype(float).round(2).to_numpy()
        df_drop_nodes['high_price_amount'] = prev_price.loc[drop_mask].astype(float).round(2).to_numpy()
        df_drop_nodes = df_drop_nodes.reset_index(drop=True)

        flight_info_cols = [
            'citypair', 'flight_numbers', 'from_time', 'from_date', 'currency',
        ]
        flight_info_cols = [c for c in flight_info_cols if c in df_target.columns]
        df_gid_info = df_target[['gid', 'baggage_weight'] + flight_info_cols].drop_duplicates(subset=['gid', 'baggage_weight']).reset_index(drop=True)
        df_drop_nodes = df_drop_nodes.merge(df_gid_info, on=['gid', 'baggage_weight'], how='left')

        drop_info_cols = [
            'drop_hours_until_departure', 'drop_price_change_percent', 'drop_price_change_amount',
            'high_price_duration_hours', 'high_price_change_percent', 'high_price_change_amount', 'high_price_amount', 
        ]
        # 按顺序排列 去掉gid
        df_drop_nodes = df_drop_nodes[flight_info_cols + ['baggage_weight'] + drop_info_cols]
        
        # 对于“上涨后再次上涨”的分析（连续两个正向变价段）
        seg_start_mask = df_target['price_duration_hours'].eq(1)
        rise_mask = seg_start_mask & (prev_pct > 0) & (df_target['price_change_percent'] > 0)

        df_rise_nodes = df_target.loc[rise_mask, ['gid', 'baggage_weight', 'hours_until_departure']].copy()
        df_rise_nodes.rename(columns={'hours_until_departure': 'rise_hours_until_departure'}, inplace=True)
        df_rise_nodes['rise_price_change_percent'] = df_target.loc[rise_mask, 'price_change_percent'].astype(float).round(4).to_numpy()
        df_rise_nodes['rise_price_change_amount'] = df_target.loc[rise_mask, 'price_change_amount'].astype(float).round(2).to_numpy()
        df_rise_nodes['prev_rise_duration_hours'] = prev_dur.loc[rise_mask].astype(float).to_numpy()
        df_rise_nodes['prev_rise_change_percent'] = prev_pct.loc[rise_mask].astype(float).round(4).to_numpy()
        df_rise_nodes['prev_rise_change_amount'] = prev_amo.loc[rise_mask].astype(float).round(2).to_numpy()
        df_rise_nodes['prev_rise_amount'] = prev_price.loc[rise_mask].astype(float).round(2).to_numpy()
        df_rise_nodes = df_rise_nodes.reset_index(drop=True)

        df_rise_nodes = df_rise_nodes.merge(df_gid_info, on=['gid', 'baggage_weight'], how='left')
        
        rise_info_cols = [
            'rise_hours_until_departure', 'rise_price_change_percent', 'rise_price_change_amount',
            'prev_rise_duration_hours', 'prev_rise_change_percent', 'prev_rise_change_amount', 'prev_rise_amount',
        ]
        df_rise_nodes = df_rise_nodes[flight_info_cols + ['baggage_weight'] + rise_info_cols]
        
        # 制作历史包络线
        envelope_group = ['citypair', 'flight_numbers', 'from_date', 'baggage_weight']
        idx_peak = df_input.groupby(envelope_group)['price_total'].idxmax()
        df_envelope = df_input.loc[idx_peak, envelope_group + [
            'price_total', 'hours_until_departure'
        ]].rename(columns={
            'price_total': 'peak_price',
            'hours_until_departure': 'peak_hours',
        }).reset_index(drop=True)
        
        del df_gid_info
        del df_target

        return df_input, df_drop_nodes, df_rise_nodes, df_envelope
    
    return df_input, None, None, None