ShopTRAINING/run_pharmacy_prediction.py

import os
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from datetime import datetime
import sys

# 导入模块
from pharmacy_predictor import train_product_model, train_product_model_with_mlstm, train_product_model_with_kan, train_product_model_with_transformer
from pharmacy_predictor import load_model_and_predict, load_kan_model_and_predict
from models import ModelManager
from pharmacy_predictor import PharmacyPredictor
from generate_pharmacy_data import generate_pharmacy_data

def clear_screen():
    """清空控制台屏幕"""
    os.system('cls' if os.name == 'nt' else 'clear')

def print_header():
    """打印应用程序标题"""
    print("=" * 80)
    print(" " * 25 + "药店销售预测系统" + " " * 25)
    print("=" * 80)

def main_menu():
    """显示主菜单并获取用户选择"""
    print("\n主菜单:")
    print("1. 训练所有药品的销售预测模型")
    print("2. 训练单个药品的销售预测模型 (Transformer)")
    print("3. 训练单个药品的销售预测模型 (mLSTM)")
    print("4. 训练单个药品的销售预测模型 (KAN)")
    print("5. 训练单个药品的销售预测模型 (优化版KAN)")  # 新增选项
    print("6. 比较原始KAN和优化版KAN模型性能")  # 新增选项
    print("7. 查看已有预测结果")
    print("8. 使用已训练的模型进行预测")
    print("9. 比较不同模型的预测结果")
    print("10. 模型管理")
    print("0. 退出")
    
    choice = input("\n请输入您的选择 (0-10): ")
    return choice

def train_all_products_menu(predictor):
    """训练所有药品的菜单"""
    clear_screen()
    print_header()
    print("\n选择模型类型:")
    print("1. Transformer")
    print("2. mLSTM")
    print("3. KAN")
    print("4. 优化版KAN")  # 新增选项
    print("0. 返回主菜单")
    
    choice = input("\n请输入您的选择 (0-4): ")
    
    if choice == '0':
        return
    
    model_types = {
        '1': 'transformer',
        '2': 'mlstm',
        '3': 'kan',
        '4': 'kan'  # 使用相同的模型类型，但在训练时指定use_optimized=True
    }
    
    if choice in model_types:
        model_type = model_types[choice]
        use_optimized = (choice == '4')  # 如果选择4，则使用优化版KAN
        
        # 获取所有产品ID
        product_ids = predictor.data['product_id'].unique()
        
        for product_id in product_ids:
            print(f"\n正在训练产品 {product_id} 的模型...")
            if model_type == 'kan' and use_optimized:
                predictor.train_optimized_kan_model(product_id)
            else:
                predictor.train_model(product_id, model_type=model_type)
        
        print("\n所有产品的模型训练完成!")
        input("\n按Enter键继续...")
    else:
        print("\n无效的选择!")
        input("\n按Enter键继续...")

def train_single_product_menu(predictor, model_type, use_optimized=False):
    """训练单个药品的菜单"""
    clear_screen()
    print_header()
    
    model_name = model_type
    if model_type == 'kan' and use_optimized:
        model_name = "优化版KAN"
    
    print(f"\n训练单个药品的{model_name}模型")
    
    # 获取所有产品ID
    if predictor.data is not None:
        product_ids = predictor.data['product_id'].unique()
        print("\n可用的产品ID:")
        for i, product_id in enumerate(product_ids):
            print(f"{i+1}. {product_id}")
        
        choice = input("\n请选择产品ID的编号 (或输入0返回): ")
        
        if choice == '0':
            return
        
        try:
            idx = int(choice) - 1
            if 0 <= idx < len(product_ids):
                product_id = product_ids[idx]
                
                # 获取训练参数
                epochs = int(input("\n请输入训练轮数 (默认100): ") or "100")
                batch_size = int(input("请输入批次大小 (默认32): ") or "32")
                learning_rate = float(input("请输入学习率 (默认0.001): ") or "0.001")
                sequence_length = int(input("请输入输入序列长度 (默认30): ") or "30")
                forecast_horizon = int(input("请输入预测天数 (默认7): ") or "7")
                hidden_size = int(input("请输入隐藏层大小 (默认64): ") or "64")
                num_layers = int(input("请输入层数 (默认2): ") or "2")
                dropout = float(input("请输入Dropout比例 (默认0.1): ") or "0.1")
                
                print(f"\n开始训练产品 {product_id} 的{model_name}模型...")
                
                if model_type == 'kan' and use_optimized:
                    # 使用优化版KAN模型
                    metrics = predictor.train_optimized_kan_model(
                        product_id=product_id,
                        epochs=epochs,
                        batch_size=batch_size,
                        learning_rate=learning_rate,
                        sequence_length=sequence_length,
                        forecast_horizon=forecast_horizon,
                        hidden_size=hidden_size,
                        num_layers=num_layers,
                        dropout=dropout
                    )
                else:
                    # 使用普通模型
                    metrics = predictor.train_model(
                        product_id=product_id,
                        model_type=model_type,
                        epochs=epochs,
                        batch_size=batch_size,
                        learning_rate=learning_rate,
                        sequence_length=sequence_length,
                        forecast_horizon=forecast_horizon,
                        hidden_size=hidden_size,
                        num_layers=num_layers,
                        dropout=dropout
                    )
                
                print("\n模型训练完成!")
                print(f"MSE: {metrics['mse']:.4f}")
                print(f"RMSE: {metrics['rmse']:.4f}")
                print(f"MAE: {metrics['mae']:.4f}")
                print(f"R²: {metrics['r2']:.4f}")
                print(f"MAPE: {metrics['mape']:.2f}%")
                print(f"训练时间: {metrics['training_time']:.2f}秒")
            else:
                print("\n无效的选择!")
        except (ValueError, IndexError):
            print("\n无效的输入!")
    else:
        print("\n没有可用的数据。请先生成或加载数据。")
    
    input("\n按Enter键继续...")

def compare_kan_models_menu(predictor):
    """比较原始KAN和优化版KAN模型性能的菜单"""
    clear_screen()
    print_header()
    print("\n比较原始KAN和优化版KAN模型性能")
    
    # 获取所有产品ID
    if predictor.data is not None:
        product_ids = predictor.data['product_id'].unique()
        print("\n可用的产品ID:")
        for i, product_id in enumerate(product_ids):
            print(f"{i+1}. {product_id}")
        
        choice = input("\n请选择产品ID的编号 (或输入0返回): ")
        
        if choice == '0':
            return
        
        try:
            idx = int(choice) - 1
            if 0 <= idx < len(product_ids):
                product_id = product_ids[idx]
                
                # 获取训练参数
                epochs = int(input("\n请输入训练轮数 (默认100): ") or "100")
                batch_size = int(input("请输入批次大小 (默认32): ") or "32")
                learning_rate = float(input("请输入学习率 (默认0.001): ") or "0.001")
                sequence_length = int(input("请输入输入序列长度 (默认30): ") or "30")
                forecast_horizon = int(input("请输入预测天数 (默认7): ") or "7")
                hidden_size = int(input("请输入隐藏层大小 (默认64): ") or "64")
                num_layers = int(input("请输入层数 (默认2): ") or "2")
                dropout = float(input("请输入Dropout比例 (默认0.1): ") or "0.1")
                
                print(f"\n开始比较产品 {product_id} 的原始KAN和优化版KAN模型...")
                
                # 使用比较方法
                comparison = predictor.compare_kan_models(
                    product_id=product_id,
                    epochs=epochs,
                    batch_size=batch_size,
                    learning_rate=learning_rate,
                    sequence_length=sequence_length,
                    forecast_horizon=forecast_horizon,
                    hidden_size=hidden_size,
                    num_layers=num_layers,
                    dropout=dropout
                )
                
                print("\n比较完成!")
            else:
                print("\n无效的选择!")
        except (ValueError, IndexError):
            print("\n无效的输入!")
    else:
        print("\n没有可用的数据。请先生成或加载数据。")
    
    input("\n按Enter键继续...")

def main():
    """主函数"""
    # 检查数据文件是否存在，如果不存在则生成
    data_path = 'pharmacy_sales.xlsx'
    if not os.path.exists(data_path):
        print("生成药店销售数据...")
        generate_pharmacy_data(num_products=10, days=365, output_file=data_path)
        print(f"数据已生成并保存到 {data_path}")
    
    # 创建预测器实例
    predictor = PharmacyPredictor(data_path=data_path)
    
    while True:
        clear_screen()
        print_header()
        choice = main_menu()
        
        if choice == '0':
            print("\n感谢使用药店销售预测系统，再见!")
            break
        elif choice == '1':
            train_all_products_menu(predictor)
        elif choice == '2':
            train_single_product_menu(predictor, model_type='transformer')
        elif choice == '3':
            train_single_product_menu(predictor, model_type='mlstm')
        elif choice == '4':
            train_single_product_menu(predictor, model_type='kan')
        elif choice == '5':
            # 使用优化版KAN模型
            train_single_product_menu(predictor, model_type='kan', use_optimized=True)
        elif choice == '6':
            # 比较原始KAN和优化版KAN模型性能
            compare_kan_models_menu(predictor)
        elif choice == '7':
            # 查看已有预测结果
            pass
        elif choice == '8':
            # 使用已训练的模型进行预测
            pass
        elif choice == '9':
            # 比较不同模型的预测结果
            pass
        elif choice == '10':
            # 模型管理
            pass
        else:
            print("\n无效的选择，请重试!")
            input("\n按Enter键继续...")

if __name__ == "__main__":
    main()