003_wz_kaggle之Otto Group Product Classification Challenge多分类问题

一、目的

对kaggle上的Otto Group Product Classification Challenge项目进行多分类，数据集中每个个体提供93个特征，分9类。

二、编程

数据集官网下载即可，https://www.kaggle.com/c/otto-group-product-classification-challenge/data，还可以看到训练集与测试集的细节以及需要提交怎样的预测文件。

使用pytorch框架，本次搭建4层神经网络，使用relu和softmax激活函数。
下面是本次编程使用的模块：

import numpy as np
from torch.utils.data import Dataset, DataLoader
import pandas as pd
import torch

本次编程同titanic项目不同的是，对于数据的处理较为简单，主要是测试集将target转化为one-hot格式存储，训练集使用id存储，方便计算交叉熵损失。

import numpy as np
from torch.utils.data import Dataset, DataLoader
import pandas as pd
import torch

# 将类别标签转换为id标签，方便后面计算交叉熵损失
def lables2id(lables):
    target_id = []
    target_lables = ['Class_1', 'Class_2', 'Class_3', 'Class_4', 'Class_5', 'Class_6', 'Class_7', 'Class_8', 'Class_9']
    for lable in lables:
        # 标签返回索引
        target_id.append(target_lables.index(lable))
    return target_id

# 建立数据集类
class MyDataset(Dataset):
    def __init__(self, filepath):
        data = pd.read_csv(filepath)
        lables = data['target']
        self.len = data.shape[0]
        # 除去id和target列，只包含特征列
        self.data_x = torch.tensor(np.array(data)[:, 1:-1].astype(float))
        # 得到target列的id索引
        self.data_y = lables2id(lables)

    def __getitem__(self, index):
        return self.data_x[index], self.data_y[index]

    def __len__(self):
        return self.len


dataset = MyDataset('./dataset/kaggle/train.csv')
# 建立数据集加载器
train_loader = DataLoader(dataset=dataset, shuffle=True, batch_size=64, num_workers=0)

# 构建网络模型
class Net(torch.nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.l1 = torch.nn.Linear(93, 64)
        self.l2 = torch.nn.Linear(64, 32)
        self.l3 = torch.nn.Linear(32, 16)
        self.l4 = torch.nn.Linear(16, 9)
        self.relu = torch.nn.ReLU()

    def forward(self, x):
        x = self.relu(self.l1(x))
        x = self.relu(self.l2(x))
        x = self.relu(self.l3(x))
        return self.l4(x)

    def predict(self, x):
        with torch.no_grad():
            x = self.relu(self.l1(x))
            x = self.relu(self.l2(x))
            x = self.relu(self.l3(x))
            x = self.relu(self.l4(x))
            # 取出最大概率的索引
            _, predicted = torch.max(x, dim=1)
            # 将预测的类别转为one-hot表示，方便后面保存为csv文件
            y = pd.get_dummies(predicted)
            return y


model = Net()

# 构建损失函数和优化器
criterion = torch.nn.CrossEntropyLoss()
optimizer = torch.optim.SGD(model.parameters(), momentum=0.5, lr=0.01)

# 训练
def train(epoch):
    running_loss = 0.
    for batch_idx, data in enumerate(train_loader, 0):
        # 取出一批数据
        inputs, target = data
        inputs = inputs.float()
        optimizer.zero_grad()
        # 正向传播
        outputs = model(inputs)
        # 计算损失
        loss = criterion(outputs, target)
        # 反向传播
        loss.backward()
        # 参数更新
        optimizer.step()

        running_loss += loss.item()
        if batch_idx % 300 == 299:
            print('[%d %5d] loss: %.3f' % (epoch+1, batch_idx+1, running_loss/300))
            running_loss = 0.


if __name__ == "__main__":
    for epoch in range(10):
        train(epoch)


def predict_sive():
    test_data = pd.read_csv("./dataset/kaggle/test.csv")
    test_inputs = torch.tensor(np.array(test_data)[:, 1:].astype(float))
    out = model.predict(test_inputs.float())
    print(out.shape)
    lables = ['Class_1', 'Class_2', 'Class_3', 'Class_4', 'Class_5', 'Class_6', 'Class_7', 'Class_8', 'Class_9']
    # 添加列标签
    out.columns = lables
    # 插入id行
    out.insert(0, 'id', test_data['id'])
    output = pd.DataFrame(out)
    output.to_csv("my_predict.csv", index=False)


predict_sive()

损失下降如下：

[1   300] loss: 1.714
[1   600] loss: 1.040
[1   900] loss: 0.891
[2   300] loss: 0.769
[2   600] loss: 0.737
[2   900] loss: 0.716
[3   300] loss: 0.691
[3   600] loss: 0.688
[3   900] loss: 0.677
[4   300] loss: 0.650
[4   600] loss: 0.659
[4   900] loss: 0.648
[5   300] loss: 0.628
[5   600] loss: 0.627
[5   900] loss: 0.629
[6   300] loss: 0.611
[6   600] loss: 0.601
[6   900] loss: 0.604
[7   300] loss: 0.585
[7   600] loss: 0.601
[7   900] loss: 0.584
[8   300] loss: 0.579
[8   600] loss: 0.579
[8   900] loss: 0.579
[9   300] loss: 0.573
[9   600] loss: 0.549
[9   900] loss: 0.567
[10   300] loss: 0.546
[10   600] loss: 0.561
[10   900] loss: 0.557

三、参考

《PyTorch深度学习实践》-刘二大人 Otto Group Product Classification作业