机器学习 增幅预测

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
%matplotlib inline

data1 = pd.read_csv('lianjia1.csv', encoding='gbk')
data2 = pd.read_csv('lianjia2.csv', encoding='gbk')
data3 = pd.read_csv('lianjia3.csv', encoding='utf-8')
data4 = pd.read_csv('lianjia4.csv', encoding='utf-8')
data5 = pd.read_csv('lianjia5.csv', encoding='utf-8')
data6 = pd.read_csv('lianjia6.csv', encoding='utf-8')
data7 = pd.read_csv('lianjia7.csv', encoding='utf-8')

data = pd.concat([data1, data2, data3, data4, data5,data6,data7])

data.columns

Index(['cjtaoshu', 'mendian', 'cjzongjia', 'zhiwei', 'haoping', 'cjdanjia',
       'cjxiaoqu', 'xingming', 'cjzhouqi', 'biaoqian', 'cjlouceng',
       'cjshijian', 'congyenianxian', 'bankuai'],
      dtype='object')

data = data[['cjzongjia', 'cjdanjia', 'cjxiaoqu', 'cjlouceng', 'cjshijian', 'bankuai', 'mendian', 'xingming', 'congyenianxian', ]]

data = data.dropna()

data.cjdanjia = np.round(data.cjdanjia.str.replace('元/平','').astype(np.float32).map(lambda x: x/10000), 2)

data.head()

data = data[data.cjshijian.str.contains('签约时间：')]

data['cjshijian'] = pd.to_datetime(data.cjshijian.str.replace('签约时间：',''))

data.head(3)

data.set_index('cjshijian', inplace=True)

data.sort_index(inplace=True)

data

data3 = data['2015-01-01' : '2015-09-30']

data4 = data['2015-10-01' : '2015-12-31']

data4

data3 = data3.assign(xiaoqu = data3.cjxiaoqu.map(lambda x : x.split()[0]))

a116 = data3.xiaoqu.value_counts()[data3.xiaoqu.value_counts() > 5].index

data3 = data3[data3.xiaoqu.isin(a116)]

data3_m = data3.groupby('xiaoqu').cjdanjia.mean()

data4 = data4.assign(xiaoqu = data4.cjxiaoqu.map(lambda x : x.split()[0]))

data4 = data4[data4.xiaoqu.isin(a116)]

data4_m = data4.groupby('xiaoqu').cjdanjia.mean()

xq_up = (data4_m - data3_m)/data3_m

(xq_up > 0.1).sum()

598

xq_up = xq_up.dropna()

type(xq_up)

pandas.core.series.Series

xq_up = xq_up.to_frame()
xq_up

xq_up = xq_up.assign(zhangfu = (xq_up.cjdanjia > 0.1).astype('int'))

data3.columns

Index(['cjzongjia', 'cjdanjia', 'cjxiaoqu', 'cjlouceng', 'bankuai', 'mendian',
       'xingming', 'congyenianxian', 'xiaoqu'],
      dtype='object')

data_taoshu = data3.groupby('xiaoqu').count()['cjxiaoqu'].to_frame().rename(columns={'cjxiaoqu':'taoshu'})

data_taoshu

data = pd.merge(xq_up, data_taoshu, left_index=True, right_index=True, how='left')

data.head(3)

del data['cjdanjia']

data_danjia = data3.groupby('xiaoqu').cjdanjia.mean().to_frame()

data = pd.merge(data, data_danjia, left_index=True, right_index=True, how='left')

data_zongjia = data3.groupby('xiaoqu').cjzongjia.mean().to_frame()

data = pd.merge(data, data_zongjia, left_index=True, right_index=True, how='left')

index = np.random.permutation(len(data))

data_train = data.iloc[index[: int(0.7 * len(data))]]

data_test = data.iloc[index[int(0.7 * len(data))]:]

x_train, y_train = data[['taoshu', 'cjdanjia', 'cjzongjia']], data.zhangfu

x_train.head(3)

x_test, y_test = data[['taoshu', 'cjdanjia', 'cjzongjia']], data.zhangfu

from sklearn.linear_model import LogisticRegression

model = LogisticRegression()

model.fit(x_train, y_train)

LogisticRegression(C=1.0, class_weight=None, dual=False, fit_intercept=True,
          intercept_scaling=1, max_iter=100, multi_class='ovr', n_jobs=1,
          penalty='l2', random_state=None, solver='liblinear', tol=0.0001,
          verbose=0, warm_start=False)

model.score(x_test,y_test)

0.6282689912826899

data_jjr_num = data3.groupby('xiaoqu').xingming.unique().map(len).to_frame()

data = pd.merge(data, data_jjr_num, left_index=True, right_index=True, how='left')

data.head(3)