阿里云天池——零基础入门数据挖掘 - 二手车交易价格预测

赛题数据见官网：

零基础入门数据挖掘 - 二手车交易价格预测_学习赛_天池大赛-阿里云天池的赛制 (aliyun.com)

代码展示：

#训练---------------------------------------------------------------------------------
import pandas as pd
import numpy as np
from torch import nn, optim
import torch
import matplotlib.pyplot as plt
config = {
    'epoch': 10,
    'batch_size': 512,
    'learning_rate': 8e-3,
    'device': 'cuda',
    "num_cols": ['regDate', 'creatDate', 'power', 'kilometer', 'v_0', 'v_1', 'v_2', 'v_3', 'v_4', 'v_5', 'v_6', 'v_7', 'v_8', 'v_9', 'v_10',
                 'v_11', 'v_12', 'v_13', 'v_14'],
    "cate_cols": ['model', 'brand', 'bodyType', 'fuelType', 'gearbox', 'seller', 'notRepairedDamage']
}
test_data = pd.read_csv('/gemini/data-1/used_car_testB_20200421.csv', sep=' ')
test_data.shape
# 导入数据
train_data = pd.read_csv('/gemini/data-1/used_car_train_20200313.csv', sep=' ')
train_data.shape
# 合并数据
data = pd.concat([train_data, test_data])
data.shape
# 定义One-Hot编码函数
def oneHotEncode(df, colNames):
    for col in colNames:
        dummies = pd.get_dummies(df[col], prefix=col)
        df = pd.concat([df, dummies],axis=1)
        df.drop([col], axis=1, inplace=True)
    return df
data.columns
data = data.replace('-', '-1')
data.notRepairedDamage = data.notRepairedDamage.astype('float32')
data.loc[data['power']>600,'power'] = 600
# 处理离散数据
for col in config['cate_cols']:
    data[col] = data[col].fillna('-1')
data = oneHotEncode(data, config['cate_cols'])

# 处理连续数据
for col in config['num_cols']:
    data[col] = data[col].fillna(0)
    data[col] = (data[col]-data[col].min()) / (data[col].max()-data[col].min())

# 处理(可能)无关数据 
data.drop(['name', 'regionCode'], axis=1, inplace=True)

data.columns
# 暂存处理后的test数据集
test_data = data[pd.isna(data.price)]
test_data.to_csv('./one_hot_testB.csv')
# 删除test数据(price is nan)
data.reset_index(inplace=True)
train_data = data
train_data = train_data.drop(data[pd.isna(data.price)].index)
train_data.shape
# 删除ID
train_data.drop(['SaleID'], axis=1, inplace=True)
# 打乱
train_data = train_data.sample(frac=1)
train_data.shape
# 分离目标
train_target = train_data['price']
train_data.drop(['price', 'index'], axis=1, inplace=True)
# 分离出验证集，用于观察拟合情况
validation_data = train_data[:10000]
train_data = train_data[10000:]
validation_target = train_target[:10000]
train_target = train_target[10000:]
validation_data.shape, train_data.shape, validation_target.shape, train_target.shape
train_data.columns
# 定义网络结构
class Network(nn.Module):
    def __init__(self, in_dim, hidden_1, hidden_2, hidden_3, hidden_4):
        super().__init__()
        self.layers = nn.Sequential(
            nn.Linear(in_dim, hidden_1),
            nn.BatchNorm1d(hidden_1),
            nn.ReLU(),
            nn.Linear(hidden_1, hidden_2),
            nn.BatchNorm1d(hidden_2),
            nn.ReLU(),
            nn.Linear(hidden_2, hidden_3),
            nn.BatchNorm1d(hidden_3),
            nn.ReLU(),
            nn.Linear(hidden_3, hidden_4),
            nn.BatchNorm1d(hidden_4),
            nn.ReLU(),
            nn.Linear(hidden_4, 1)
        )

    def forward(self, x):
        y = self.layers(x)
        return y
# 定义网络
model = Network(train_data.shape[1], 256, 256, 256, 32)
model.to(config['device'])

# 使用Xavier初始化权重
for line in model.layers:
    if type(line) == nn.Linear:
        print(line)
        nn.init.xavier_uniform_(line.weight)
# 将数据转化为tensor，并移动到cpu或cuda上

train_features = torch.tensor(train_data.values, dtype=torch.float32, device=config['device'])
train_num = train_features.shape[0]
train_labels = torch.tensor(train_target.values, dtype=torch.float32, device=config['device'])

validation_features = torch.tensor(validation_data.values, dtype=torch.float32, device=config['device'])
validation_num = validation_features.shape[0]
validation_labels = torch.tensor(validation_target.values, dtype=torch.float32, device=config['device'])
# 特征长度
train_features[1].shape
# 定义损失函数和优化器
criterion = nn.MSELoss()
criterion.to(config['device'])
optimizer = optim.Adam(model.parameters(), lr=config['learning_rate'])
# 开始训练

mae_list = []

for epoch in range(config['epoch']):
    losses = []
    model.train()
    for i in range(0, train_num, config['batch_size']):
        end = i + config['batch_size']
        if i + config['batch_size'] > train_num-1:
            end = train_num-1
        mini_batch = train_features[i: end]
        mini_batch_label = train_labels[i: end]
        pred = model(mini_batch)
        pred = pred.squeeze()
        loss = criterion(pred, mini_batch_label)

        if torch.isnan(loss):
            break
        mae = torch.abs(mini_batch_label-pred).sum()/(end-i)
        losses.append(mae.item())
        
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()
    
    model.eval()
    pred = model(validation_features)
    validation_mae = torch.abs(validation_labels-pred.squeeze()).sum().item()/validation_num
    
    mae_list.append((sum(losses)/len(losses), validation_mae))
        
    print(f"epoch:{epoch + 1} MAE: {sum(losses)/len(losses)}, Validation_MAE: {validation_mae}")
    torch.save(model, 'model.pth')
x = np.arange(0, config['epoch'])
y1, y2 = zip(*mae_list)
plt.plot(x, y1, label='train')
plt.plot(x, y2, label='valid')
plt.legend()
plt.show()

#预测----------------------------------------------------------------------------------
import pandas as pd
import torch
from torch import nn
from settings import config
class Network(nn.Module):
    def __init__(self, in_dim, hidden_1, hidden_2, hidden_3, hidden_4):
        super().__init__()
        self.layers = nn.Sequential(
            nn.Linear(in_dim, hidden_1),
            nn.BatchNorm1d(hidden_1),
            nn.ReLU(),
            nn.Linear(hidden_1, hidden_2),
            nn.BatchNorm1d(hidden_2),
            nn.ReLU(),
            nn.Linear(hidden_2, hidden_3),
            nn.BatchNorm1d(hidden_3),
            nn.ReLU(),
            nn.Linear(hidden_3, hidden_4),
            nn.BatchNorm1d(hidden_4),
            nn.ReLU(),
            nn.Linear(hidden_4, 1)
        )

    def forward(self, x):
        y = self.layers(x)
        return y
model = torch.load('model.pth', map_location=config['device'])
data = pd.read_csv('./one_hot_testB.csv')
data.shape
data.columns
data = data.drop(columns=['Unnamed: 0', 'price'])
test = data.drop(columns='SaleID')
test.shape
test = torch.tensor(test.values, dtype=torch.float32)
pred = model(test)
price = pd.DataFrame(pred.detach().cpu().numpy(), columns=['price'])
res = pd.concat([data.SaleID, price], axis=1)
res.to_csv('output.csv')

运行结果：见附带的output.csv文件