《动手学深度学习》之实战Kaggle比赛：房价预测

参考实战Kaggle比赛：房价预测

import torch
import torch.nn as nn
import numpy as np
import pandas as pd
import sys
sys.path.append("..")
import d2lzh_pytorch as d2l


torch.set_default_tensor_type(torch.FloatTensor)

#PART ONE 获取和读取数据集-----------------------------------------------------
train_data = pd.read_csv('./train.csv')
test_data = pd.read_csv('./test.csv')

# 训练集包括1460个样本和80个特征 1个标签,测试数据集包括1459个样本和80个特征
print(train_data.shape)  

# 前4个样本的前4个特征、后2个特征和标签（SalePrice）,0:4指前四行
print(train_data.iloc[0:4, [0, 1, 2, 3, -3, -2, -1]])

# 第一个特征是Id，它能帮助模型记住每个训练样本，
# 但难以推广到测试样本，所以我们不使用它来训练。
# 我们将所有的训练数据和测试数据的79个特征按样本连结。
all_features = pd.concat((train_data.iloc[:, 1:-1], test_data.iloc[:, 1:]))


#PART TWO 预处理数据-----------------------------------------------------------
numeric_features = all_features.dtypes[all_features.dtypes != 'object'].index   
# numeric_features 是  all_features 的索引
# 设该特征在整个数据集上的均值为μ，标准差为σ。
# 那么，我们可以将该特征的每个值先减去μ再除以σ得到标准化后的每个特征值。
all_features[numeric_features] = all_features[numeric_features].apply(
    lambda x: (x - x.mean()) / (x.std()))
# 标准化后，每个数值特征的均值变为0，所以可以直接用0来替换缺失值
all_features[numeric_features] = all_features[numeric_features].fillna(0)

# 将离散数值转成指示特征--
# 举例如下:假设特征MSZoning里面有两个不同的离散值RL和RM，
# 那么这一步转换将去掉MSZoning特征，
# 并新加两个特征MSZoning_RL和MSZoning_RM，其值为0或1。
# 如果一个样本原来在MSZoning里的值为RL，
# 那么有MSZoning_RL=1且MSZoning_RM=0。

# dummy_na=True将缺失值也当作合法的特征值并为其创建指示特征
all_features = pd.get_dummies(all_features, dummy_na=True)
print(all_features.shape) # (2919, 331),这一步转换将特征数从79增加到了331

# 通过values属性得到NumPy格式的数据，并转成Tensor方便后面的训练
n_train = train_data.shape[0]
train_features = torch.tensor(all_features[:n_train].values, dtype=torch.float)
test_features = torch.tensor(all_features[n_train:].values, dtype=torch.float)
train_labels = torch.tensor(train_data.SalePrice.values, dtype=torch.float).view(-1, 1)


# PART THREE  训练模型---------------------------------------------------------
# 使用一个基本的线性回归模型和平方损失函数来训练模型
loss = torch.nn.MSELoss()

def get_net(feature_num):
    net = nn.Linear(feature_num, 1)
    for param in net.parameters():
        nn.init.normal_(param, mean=0, std=0.01)
    return net

# 定义比赛用来评价模型的对数均方根误差
def log_rmse(net, features, labels):
    with torch.no_grad():
        # 将小于1的值设成1，使得取对数时数值更稳定
        clipped_preds = torch.max(net(features), torch.tensor(1.0))
        rmse = torch.sqrt(2 * loss(clipped_preds.log(), labels.log()).mean())
    return rmse.item()


def train(net, train_features, train_labels, test_features, test_labels,
          num_epochs, learning_rate, weight_decay, batch_size):
    train_ls, test_ls = [], []
    dataset = torch.utils.data.TensorDataset(train_features, train_labels)
    train_iter = torch.utils.data.DataLoader(dataset, batch_size, shuffle=True)
    # 这里使用了Adam优化算法
    optimizer = torch.optim.Adam(params=net.parameters(), lr=learning_rate, weight_decay=weight_decay) 
    net = net.float()
    for epoch in range(num_epochs):
        for X, y in train_iter:
            l = loss(net(X.float()), y.float())
            optimizer.zero_grad()
            l.backward()
            optimizer.step()
        train_ls.append(log_rmse(net, train_features, train_labels))
        if test_labels is not None:
            test_ls.append(log_rmse(net, test_features, test_labels))
    return train_ls, test_ls


# PART FOUR K折交叉验证--------------------------------------------------------
# K折交叉验证用来选择模型设计并调节超参数
# 在K折交叉验证中，我们把原始训练数据集分割成K个不重合的子数据集，
# 然后我们做K次模型训练和验证
# 每一次，我们使用一个子数据集验证模型，并使用其他K−1个子数据集来训练模型。
# 在这K次训练和验证中，每次用来验证模型的子数据集都不同。
# 最后，我们对这K次训练误差和验证误差分别求平均。
def get_k_fold_data(k, i, X, y):
    # 返回第i折交叉验证时所需要的训练和验证数据
    assert k > 1   #k>1才执行下面
    fold_size = X.shape[0] // k
    X_train, y_train = None, None
    for j in range(k):
        idx = slice(j * fold_size, (j + 1) * fold_size)          #切片
        X_part, y_part = X[idx, :], y[idx]
        if j == i:
            X_valid, y_valid = X_part, y_part
        elif X_train is None:
            X_train, y_train = X_part, y_part
        else:
            X_train = torch.cat((X_train, X_part), dim=0)  #如果不是第i折就将剩余折连接起来
            y_train = torch.cat((y_train, y_part), dim=0)
    return X_train, y_train, X_valid, y_valid


# 在K折交叉验证中训练K次并返回训练和验证的平均误差。
def k_fold(k, X_train, y_train, num_epochs,
           learning_rate, weight_decay, batch_size):
    train_l_sum, valid_l_sum = 0, 0
    for i in range(k):
        data = get_k_fold_data(k, i, X_train, y_train)
        net = get_net(X_train.shape[1])
        train_ls, valid_ls = train(net, *data, num_epochs, learning_rate,
                                   weight_decay, batch_size)
        train_l_sum += train_ls[-1]
        valid_l_sum += valid_ls[-1]
        if i == 0:
            d2l.semilogy(range(1, num_epochs + 1), train_ls, 'epochs', 'rmse',
                         range(1, num_epochs + 1), valid_ls,
                         ['train', 'valid'])
        print('fold %d, train rmse %f, valid rmse %f' % (i, train_ls[-1], valid_ls[-1]))
    return train_l_sum / k, valid_l_sum / k

k, num_epochs, lr, weight_decay, batch_size = 5, 100, 5, 0, 64
train_l, valid_l = k_fold(k, train_features, train_labels, num_epochs, lr, weight_decay, batch_size)
print('%d-fold validation: avg train rmse(均方根误差) %f, avg valid rmse %f' % (k, train_l, valid_l))


# 定义预测函数
def train_and_pred(train_features, test_features, train_labels, test_data,
                   num_epochs, lr, weight_decay, batch_size):
    net = get_net(train_features.shape[1])
    train_ls, _ = train(net, train_features, train_labels, None, None,
                        num_epochs, lr, weight_decay, batch_size)
    d2l.semilogy(range(1, num_epochs + 1), train_ls, 'epochs', 'rmse')
    print('train rmse %f' % train_ls[-1])
    preds = net(test_features).detach().numpy()
    test_data['SalePrice'] = pd.Series(preds.reshape(1, -1)[0])
    submission = pd.concat([test_data['Id'], test_data['SalePrice']], axis=1)
    submission.to_csv('./submission.csv', index=False)

train_and_pred(train_features, test_features, train_labels, test_data, num_epochs, lr, weight_decay, batch_size)

d2lzh_pytorch