基于随机森林的房价预测（以国内厦门二手房房价预测为例）

一、数据集

        此数据集来自厦门二手房房源_厦门二手房出售|买卖|交易信息(厦门链家)，爬取的厦门二手房数据，数据下载链接为百度网盘 请输入提取码，提取码为liuj。

二、数据预处理

        根据之前的数据集来看，我们需要对数据进行预处理，包括去除‘['’与 ‘']’，去除unit_price的 ‘元\平’ ,follower 的 ‘人关注’，building_area的 ‘平米’，build_year的 ‘年建’，替换修建时间未知值为空值，删除空值。

u = 0
def data_predict(filepath):
#读取数据集
    import pandas as pd

    global u
    df = pd.read_csv(filepath,encoding='gbk')

# #查看文件前数据
# print(df.head())
#
# #查看数据集的简单描述
# print(df.info())
#
# #查看数据集的简单描述
# print(df.describe())

#改动title、total_price、unit_price、louceng、chaoxiang、zhuangxiu、building_area、build_year、community_name、location_info、location_info_seclisting_day、last_trade
#去除‘['’与 ‘']’
    df.iloc[:,1] = df.iloc[:,1].str[2:-2]
    for i in range(3,5):
        df.iloc[:,i] = df.iloc[:,i].str[2:-2]
    for j in range(7,15):
        df.iloc[:,j] = df.iloc[:,j].str[2:-2]
    for k in range(17, 19):
        df.iloc[:,k] = df.iloc[:,k].str[2:-2]
#去除unit_price的 ‘元\平’ ,follower 的 ‘人关注’，building_area的 ‘平米’，build_year的 ‘年建’
    df.iloc[:,4] = df.iloc[:,4].str[:-3]
    df.iloc[:,5] = df.iloc[:,5].str[:-4]
    df.iloc[:,10] = df.iloc[:,10].str[:-2]
    df.iloc[:,11] = df.iloc[:,11].str[:-2]


#替换修建时间未知值为空值
    mask = df.iloc[:,11].str.startswith("未")
    df.iloc[mask,11] = ''
#   colunmn_title = df['title']
#   df['title'] = colunmn_title.str[2:-2]
#   print(df)


#将df存入csv文件中
    df.to_csv('data/xmhouse_data_clean.csv',index=False,encoding='gbk')

### 删除空值
    df_02 = pd.read_csv('data/xmhouse_data_clean.csv',encoding='gbk')
    print(df_02.info())
#   print(df_02['last_trade'])
#   print(df_02.info())
    df_02 = df_02.dropna(axis=0)
    print(df_02.info())
##更兴第一列索引
    df_02.iloc[:,0] = range(0,len(df_02))
##保存csv文件
    df_02.to_csv('data/xmhouse_data_clean_{}.csv'.format(u),index=False,encoding='gbk')
    u = u + 1
    return 'data/xmhouse_data_clean_{}.csv'.format(u-1)

三、建立模型

        本次实验采用的是随机森林对数据进行训练（训练集占比为80%，测试集占比为20%）

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import LabelEncoder, StandardScaler
from sklearn.ensemble import RandomForestRegressor
from sklearn.metrics import mean_squared_error, r2_score
from data_predict import *

def model(failpath_1,failpath_2):
#   df = pd.read_csv('data/xmhouse_data_clean.csv',encoding='gbk')
#   print(df.describe())
#
#   df.hist(bins=20,figsize=(20,15))
#   plt.show()

    data_csv = pd.read_csv(failpath_1,encoding='gbk')
    data_csv_2 = pd.read_csv(failpath_2,encoding='gbk')
# 对分类变量进行编码
    label_encoders = {}
    for column in ['huxing', 'title','louceng','building_area','last_trade','listing_day','elevator','elevator_house','location_info_sec','zhuangxiu','location_info','build_year', 'chaoxiang', 'community_name']:
        le = LabelEncoder()
        data_csv[column] = le.fit_transform(data_csv[column])
        data_csv_2[column] = le.fit_transform(data_csv[column])
        label_encoders[column] = le
    #   print(data_csv[column].unique())
    #   print(le.classes_)

# 划分特征和目标变量
    X = data_csv[
        ['huxing', 'louceng','title','zhuangxiu','last_trade','listing_day','location_info_sec','elevator_house',
         'elevator','build_year','location_info','building_area', 'chaoxiang', 'community_name']]
    y = data_csv['total_price']


    X2 = data_csv_2[
    ['huxing', 'louceng', 'title', 'zhuangxiu', 'last_trade', 'listing_day', 'location_info_sec', 'elevator_house',
     'elevator', 'build_year', 'location_info', 'building_area', 'chaoxiang', 'community_name']]
    y2 = data_csv_2['total_price']


# 划分训练集和测试集
    X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)
    X_train_2, X_test_2, y_train_2, y_test_2 = train_test_split(X2, y2, test_size=0.999, random_state=42,shuffle=False)


# 特征缩放
    scaler = StandardScaler()
    X_train_scaled = scaler.fit_transform(X_train)
    X_test_scaled = scaler.transform(X_test)
    X_test_scaled_2 = scaler.transform(X_test_2)

# 创建随机森林回归模型
    rf_model = RandomForestRegressor(n_estimators=100, random_state=42)

# 在训练集上训练模型
    rf_model.fit(X_train_scaled, y_train)

# 在测试集上进行预测
    y_pred = rf_model.predict(X_test_scaled)
    y_pred_2 = rf_model.predict(X_test_scaled_2)
# 评估模型性能
    mse = mean_squared_error(y_test, y_pred)
    r2 = r2_score(y_test, y_pred)

    plt.scatter(y_test, y_pred)
    plt.xlabel('Actual Price')
    plt.ylabel('Predicted Price')
    plt.xlim(0, 1000)
    plt.ylim(0, 1000)
    plt.title('Random Forest Regression')
    x_line = np.linspace(0, 4400, 4400)
    y_line = x_line
# 绘制 y=x 线
    plt.plot(x_line, y_line, 'r')
    plt.show()


    pd.set_option('display.max_rows', None)
    # print(pd.DataFrame({'实际房价': y_test, '预测房价': y_pred}))
    print(pd.DataFrame({'实际房价': y_test_2, '预测房价': y_pred_2}))

四、主函数main()

from data_predict import data_predict as D1
from model import model as D2
if __name__ == '__main__':
    failpath_1 = D1('data/xmhouse_data.csv')
    #路径二是对你爬取后来的数据进行预测，本实验并未爬取其他的数据，故使用本身的数据进行测试
    failpath_2 = D1('data/xmhouse_data.csv')
    D2(failpath_1,failpath_2)

五、预测结果

        1.预测结果图如下所示,x轴代表真实价值，y轴代表预测价值。

        2. 房价预测文字展示如下图所示