jiangcang_vj/predict.py

import datetime
import torch
import pandas as pd
import numpy as np
import os
from torch.utils.data import DataLoader
from utils import FlightDataset


def predict_future_distribute(model, sequences, group_ids, batch_size=16, target_scaler=None, interval_hours=8, predict_dir=".", pred_time_str=""):
    if not sequences:
        print("没有足够的数据进行预测。")
        return
    
    test_dataset = FlightDataset(sequences, None, group_ids)
    test_loader = DataLoader(test_dataset, batch_size=batch_size)
    model.eval()

    y_preds = []
    group_info = []
    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')

    with torch.no_grad():
        for X_batch, group_ids in test_loader:
            X_batch = X_batch.to(device)
            outputs = model(X_batch)
            y_preds.extend(outputs.cpu().numpy())
            for i in range(len(group_ids[0])):
                group_id = tuple(group_ids_elem[i] for group_ids_elem in group_ids)
                group_info.append(group_id)                
    
    y_preds = np.array(y_preds)
    y_preds_class = y_preds[:, 0]
    y_preds_class_labels = (y_preds_class >= 0.5).astype(int)
    
    results_df = pd.DataFrame({
        'city_pair': [info[0] for info in group_info],
        'flight_day': [info[1] for info in group_info],
        'flight_number_1': [info[2] for info in group_info],
        'flight_number_2': [info[3] for info in group_info],
        'from_date': [info[4] for info in group_info],
        'baggage': [info[5] for info in group_info],
        'price': [info[6] for info in group_info],
        'Hours_until_Departure': [info[7] for info in group_info],
        'update_hour': [info[8] for info in group_info],
        'crawl_date': [info[9] for info in group_info],
        'probability': y_preds_class,
        'Predicted_Will_Price_Drop': y_preds_class_labels,
    })

    # 先转成 datetime
    update_hour_dt = pd.to_datetime(results_df['update_hour'])                # 起飞前36小时对应时间(整点)
    valid_begin_dt = update_hour_dt + pd.Timedelta(hours=interval_hours)      # 起飞前28小时(36-8=28)(32-4=28)(30-2=28)对应时间(整点) 
    valid_end_dt = valid_begin_dt + pd.Timedelta(hours=24)                    # 起飞前4小时(28-24=4) 

    # 统一格式化
    valid_begin_str = valid_begin_dt.dt.strftime('%Y-%m-%d %H:%M:%S')
    valid_end_str = valid_end_dt.dt.strftime('%Y-%m-%d %H:%M:%S')

    # probability 列的位置
    prob_col_idx = results_df.columns.get_loc('probability')
    
    # interval_hours（统一数值）
    results_df.insert(
        loc=prob_col_idx,
        column='interval_hours',
        value=interval_hours
    )

    # valid_begin_hour
    results_df.insert(
        loc=prob_col_idx + 1,   # 原 probability 列的位置 加1
        column='valid_begin_hour',
        value=valid_begin_str
    )

    # valid_end_hour
    results_df.insert(
        loc=prob_col_idx + 2,   # 原 probability 列的位置 加2
        column='valid_end_hour',
        value=valid_end_str
    )
    
    # 数值处理
    threshold = 1e-3
    numeric_columns = ['probability']
    for col in numeric_columns:
        results_df[col] = results_df[col].where(results_df[col].abs() >= threshold, 0)
    
    csv_path1 = os.path.join(predict_dir, f'future_predictions_{pred_time_str}.csv')
    results_df.to_csv(csv_path1, mode='a', index=False, header=not os.path.exists(csv_path1), encoding='utf-8-sig')

    print("预测结果已追加")

    return results_df