import pandas as pd
import torch
import numpy as np
from sklearn.preprocessing import StandardScaler, MinMaxScaler
# 数据预处理
def process_data(file_path):
    # 读取Excel文件，跳过首行，使用科学计数法解析
    df = pd.read_excel(file_path, header=None,sheet_name="sample1")

    # 提取输入输出数据
    inputs = df.iloc[:, 1:9].values.astype(np.float32)  # 第2-9列
    outputs = df.iloc[:, 10:19].values.astype(np.float32)  # 第11-19列

    # 数据标准化
    input_scaler = StandardScaler()
    output_scaler = StandardScaler()
    inputs = input_scaler.fit_transform(inputs)
    outputs = output_scaler.fit_transform(outputs)
    #
    # input_scaler = MinMaxScaler(feature_range=(-1, 1))
    # output_scaler = MinMaxScaler(feature_range=(-1, 1))
    # inputs = input_scaler.fit_transform(inputs)
    # outputs = output_scaler.fit_transform(outputs)

    return inputs, outputs, input_scaler, output_scaler

# 创建序列数据集 look_back：依据时间序列 pred_step：推测时间序列
def create_sequences(inputs, outputs, look_back=8, pred_step=1):
    X, y = [], []
    for i in range(len(inputs) - look_back - pred_step):
        X.append(inputs[i:i + look_back])
        y.append(outputs[(i+look_back):(i+look_back+pred_step)])
    return torch.FloatTensor(np.array(X)), torch.FloatTensor(np.array(y))