# 基于CNN_A_LSTM的小时天气预测

**Repository Path**: wangp06/CNN_A_LSTM_Weather_Predict

## Basic Information

- **Project Name**: 基于CNN_A_LSTM的小时天气预测
- **Description**: 基于CNN_A_LSTM的小时天气预测
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 3
- **Created**: 2024-07-22
- **Last Updated**: 2024-07-22

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# 基于CNN_A_LSTM的小时天气预测

#### 介绍
基于CNN_A_LSTM的小时天气预测(有问题请联系qq:869114539,vx:15234405680)
| 文件  | 用途 |
|---|---|
|  code目录 |   代码|
| Houston.csv  |  数据集 |
| util.py  |  工具类 |


#### 基于深度学习的小时天气预测(温度单步预测)

基于torch框架完成

| 模型  |  参数 |
|---|---|
|  RNN |  神经元个数=32  步长=24|
|  LSTM | 神经元个数=32  步长=24|
|  GRU |  神经元个数=64 步长=24|
|  CNN_LSTM | 神经元个数=64  步长=24|
|  CNN_A_LSTM | 神经元个数=64    步长=24|

#### 缺失值处理

插值填充
df = df.interpolate()

#### weather数值化

        label_encoder = LabelEncoder()
        df['weather'] = label_encoder.fit_transform(df['weather'])
        category_mapping = dict(zip(label_encoder.classes_, label_encoder.transform(label_encoder.classes_)))
![输入图片说明](1703468248373.png)

#### 数据格式处理
因需要预测未来一小时天气（时间序列单步预测），基于滑动窗口的方式对数据格式进行处理：
![输入图片说明](image.png)


    def create_sliding_window(data, window_size):
        window_data = []
        for i in range(len(data) - window_size):
            window = data[i:i + window_size]
            target = data[i + window_size][2]
            window_data.append((window, target))
        return window_data


#### RNN实现

    class RNN(nn.Module):
        def __init__(self, input_size, hidden_size, output_size):
            super(LSTMModel, self).__init__()
            self.hidden_size = hidden_size
            self.lstm = nn.RNN(input_size, hidden_size)
            self.fc = nn.Linear(hidden_size, output_size)
    
        def forward(self, x):
            lstm_out, _ = self.lstm(x)
            output = self.fc(lstm_out[:, -1, :])
            return output

![RNN训练Loss](rnn_loss.png)![RNN测试集预测结果展示](rnn_true_vs_predict.jpg)

#### LSTM实现
    class LSTMModel(nn.Module):
        def __init__(self, input_size, hidden_size, output_size):
            super(LSTMModel, self).__init__()
            self.hidden_size = hidden_size
            self.lstm = nn.LSTM(input_size, hidden_size)
            self.fc = nn.Linear(hidden_size, output_size)
    
        def forward(self, x):
            lstm_out, _ = self.lstm(x)
            output = self.fc(lstm_out[:, -1, :])
            return output

![LSTM训练Loss](lstm_loss.jpg)![LSTM测试集预测结果展示](lstm_true_vs_predict.jpg)

#### GRU实现
![输入图片说明](gru_loss.jpg)![输入图片说明](gru_true_vs_predict.jpg)

#### CNN_LSTM实现

    class CNNLSTMModel(nn.Module):
        def __init__(self, input_size, hidden_size, output_size):
            super(CNNLSTMModel, self).__init__()
            self.hidden_size = hidden_size
            self.conv1 = nn.Conv1d(input_size, hidden_size, kernel_size=3, padding=1)
            self.lstm = nn.LSTM(hidden_size, hidden_size)
            self.fc = nn.Linear(hidden_size, output_size)
    
        def forward(self, x):
            x = x.permute(0, 2, 1)  # 转换输入的维度，使其适合卷积层的输入
            x = F.relu(self.conv1(x))
            x = x.permute(2, 0, 1)  # 将维度转换回原始维度
            lstm_out, _ = self.lstm(x)
            output = self.fc(lstm_out[-1, :, :])  # 只使用最后一个时间步的输出
            return output
![输入图片说明](cnn_lstm.jpg)
![输入图片说明](cnn_lstm_true_.jpg)


#### CNN_A_LSTM实现
![输入图片说明](cnn_a_lstm_loss.jpg)![输入图片说明](cnn_a_lstm_true_vs_predict.jpg)