实战五十六：基于机器学习的上海房价预测（完整代码+数据集）

 # coding:utf-8    ctrl+/  注释

    
 import urllib2
    
 from bs4 import BeautifulSoup
    
 import sys
    
 import pymongo
    
 import requests
    
  
    
 def geocode(address):
    
     parameters = {'address': address, 'key': '8ac4f59c23c73f503f350494ff9310d3','city':'上海'}
    
     base = 'http://restapi.amap.com/v3/geocode/geo'
    
     try:
    
     response = requests.get(base, parameters,timeout=1)
    
     except:
    
     return {}
    
     answer = response.json()
    
     return answer
    
  
    
  
    
 reload(sys)
    
 sys.setdefaultencoding("utf-8")
    
  
    
 connection = pymongo.MongoClient()
    
 tdb = connection.program
    
 post_info = tdb.house
    
  
    
  
    
 # 链家网d
    
 def find_data(tmp_url, tmp_district, lists):
    
     count = 0
    
     # 每个区的最大显示页数为100页
    
     for page_Num in range(1, 100):
    
     f_url = tmp_url + tmp_district + "/d" + str(page_Num)
    
     print f_url
    
     # print f_url
    
     f_page = urllib2.urlopen(f_url)
    
     f_soup = BeautifulSoup(f_page, "html.parser")
    
  
    
     page_soup = f_soup.find(class_="m-list")
    
     # print page_soup
    
     ul_soup = page_soup.find('ul')
    
     # print ul_soup
    
     li_list = ul_soup.findAll('li')[0:]
    
     # print page_Num
    
     for tr in li_list:
    
         info = tr.findAll(class_="info-row")[0:]
    
         row1 = info[0].find(class_="info-col row1-text").text
    
  
    
         row1 = row1.strip()
    
         row1 = row1.replace(' ', '')
    
         # print row1
    
         cut_1 = row1.index('|')
    
         # 几室几厅
    
         house_type = row1[0:cut_1]
    
         #print house_type
    
         cut_2 = row1[cut_1 + 1:].index('平')
    
         # 房屋大小
    
         house_size = float(row1[cut_1 + 1:cut_1 + cut_2 + 1])
    
         #print house_size
    
         try:
    
             cut_3 = row1.index('/')
    
         except ValueError:
    
             continue
    
         cut_4 = row1.index('层')
    
         # 建筑总层高
    
         building_height = float(row1[cut_3 + 1:cut_4])
    
         #print building_height
    
         cut_5 = row1.index('区')
    
         # 房屋层高
    
         house_height = row1[cut_5 - 1:cut_5]
    
         #print house_height
    
  
    
         row2 = info[1].find(class_="info-col row2-text")
    
         # 均价
    
         average_price = info[1].find(class_="info-col price-item minor").text.strip()
    
         print average_price
    
         # 位置
    
         location = row2.findAll('a')[0:]
    
  
    
         try:
    
             year_1 = row2.text.index('年建')
    
             year = row2.text[year_1 - 4:year_1]
    
         except ValueError:
    
             continue
    
  
    
         print year
    
         # 小区
    
         housing_estate = location[0].text
    
         #print housing_estate
    
         # 区县
    
         house_district = location[1].text
    
         #print house_district
    
         count = count + 1
    
         # print count
    
         #
    
         #
    
         #整理出room和parlour数量
    
         room_1=house_type.index('室')
    
         room_number=int(house_type[0:room_1])
    
         parlour_1=house_type.index('厅')
    
         parlour_number=int(house_type[room_1+1:parlour_1])
    
  
    
         #判断房子的具体高度
    
         if(house_height=='中'):
    
             house_height_inlist=building_height*0.5
    
         elif(house_height=='高'):
    
             house_height_inlist = building_height * 0.88
    
         elif(house_height == '低'):
    
             house_height_inlist = building_height * 0.23
    
  
    
         #整理出具体房价
    
         price_1=average_price.index('价')
    
         price_2=average_price.index('元')
    
         average_price_inlist=float(average_price[price_1+1:price_2])
    
  
    
         #计算地址经纬度
    
         address = house_district+housing_estate
    
         print address
    
  
    
         house_first_location=geocode(address)
    
  
    
         if "geocodes" not in house_first_location.keys():
    
             print "no geocodes"
    
             continue
    
         house_first_location=house_first_location["geocodes"]
    
  
    
         #print house_first_location
    
         if(house_first_location==[]):
    
             continue
    
  
    
         house_location=house_first_location[0]['location'].encode('utf-8')
    
         print house_location
    
         house_location_1=house_location.index(',')
    
         house_location_longtitude=float(house_location[0:house_location_1])
    
         print house_location_longtitude
    
         house_location_latitude=float(house_location[house_location_1+1:])
    
         print house_location_latitude
    
         #house_location_latitude,house_location_longtitude,
    
         list_use = [room_number,parlour_number, house_size, building_height, house_height_inlist,float(year),house_location_longtitude,house_location_latitude,house_district,address,
    
                     average_price_inlist]
    
  
    
         #print list_use
    
         lists.extend([list_use])
    
  
    
         print "{\"户型\":\"%s\", \"大小\":\"%s\",\"楼高\":\"%s\",\"层高\":\"%s\",\"小区名\":\"%s\",\"市区\":\"%s\",\"均价\":\"%s\"}" % (
    
             house_type, house_size, building_height, house_height, housing_estate, house_district, average_price)
    
         data = {"room_number": room_number, "parlour_number": parlour_number,"house_size":house_size,"building_height": building_height,"house_height_inlist": house_height_inlist,"year": float(year),"house_location_longtitude": house_location_longtitude,"house_location_latitude":house_location_latitude,"house_district":house_district,"address":address,"average_price_inlist": average_price_inlist}
    
         post_info.save(data)
    
         # print "end"
    
         # print "all_end"
    
  
    
  
    
 def use(district):
    
     lists = []
    
     for i in district:
    
     find_data('http://sh.lianjia.com/ershoufang/', i, lists)
    
     return lists
    
  
    
 a = ["pudongxinqu", "minhang", "baoshan", "xuhui", "putuo", "yangpu", "changning", "songjiang", "jiading", "huangpu",
    
      "jingan", "zhabei", "hongkou", "qingpu", "fengxian", "jinshan", "chongming"]
    
  
    
 if __name__ == '__main__':
    
     dataset = use(a)
    
    
    
    
    代码解读

 # coding:utf-8    ctrl+/  注释

    
 import sys
    
 import pymongo
    
 import numpy as np
    
 import price_mongo
    
  
    
 reload(sys)
    
 sys.setdefaultencoding("utf-8")
    
  
    
 connection = pymongo.MongoClient()
    
 tdb = connection.program
    
 post_info = tdb.house
    
  
    
 lists=[]
    
 count=0
    
  
    
 for item in post_info.find():
    
     count+=1
    
  
    
     single_item=[item["room_number"],item["parlour_number"],item["house_size"],item["year"],
    
              item["building_height"],item["house_height_inlist"],item["house_location_longtitude"],
    
              item["house_location_latitude"],item["average_price_inlist"]]
    
     lists.extend([single_item])
    
     #print count
    
  
    
  
    
 from sklearn import metrics
    
 from sklearn import preprocessing
    
 from sklearn.linear_model import LinearRegression
    
 from sklearn.cross_validation import train_test_split
    
  
    
 _array = np.array(lists)
    
 x = _array[:, 0:8]
    
 #normalized_X=preprocessing.normalize(x)
    
 y = _array[:, 8]
    
 #normalized_Y=preprocessing.normalize(y)
    
 X_train, X_test, y_train, y_test = train_test_split(x, y,test_size=0.2)
    
  
    
 print "训练集大小%d"%(X_train.shape[0])
    
 print "测试集大小%d"%(X_test.shape[0])
    
  
    
 #线性回归
    
 clf = LinearRegression()
    
 clf.fit(X_train, y_train)
    
 print "线性回归模型参数:"
    
 print clf.coef_
    
 y_pred = clf.predict(X_test)
    
  
    
 print ("线性回归误差:%d元"%(np.sqrt(metrics.mean_squared_error(y_test, y_pred))))
    
  
    
  
    
 #神经网络
    
 from sklearn.neural_network import MLPRegressor
    
 kmodel = MLPRegressor(learning_rate='adaptive',max_iter=2000).fit(X_train, y_train)
    
 y_kmodel_pred = kmodel.predict(X_test)
    
 print ("神经网络误差:%d元"%np.sqrt(metrics.mean_squared_error(y_test, y_kmodel_pred)))
    
  
    
 #svm模型
    
 from sklearn import svm
    
 smodel=svm.LinearSVR()
    
 smodel.fit(X_train, y_train)
    
 y_smodel_pred=smodel.predict(X_test)
    
  
    
 print "svm误差:%d元"%np.sqrt(metrics.mean_squared_error(y_test, y_smodel_pred))
    
  
    
  
    
 #决策树模型
    
 from sklearn.tree import DecisionTreeRegressor
    
 dmodel=DecisionTreeRegressor()
    
 dmodel.fit(X_train, y_train)
    
  
    
 y_dmodel_pred=dmodel.predict(X_test)
    
 print "非线性决策树误差:%d元"%np.sqrt(metrics.mean_squared_error(y_test, y_dmodel_pred))
    
 print "非线性决策树误差评分:%f"%dmodel.score(X_test,y_test)
    
  
    
 print "各参数权重:"
    
 print dmodel.feature_importances_
    
 # coding:utf-8    ctrl+/  注释
    
  
    
  
    
 import wx
    
 app=wx.App()
    
 win =wx.Frame(None,title="房价查询",size=(420,335))
    
 win.Show()
    
  
    
 btn=wx.Button(win,label='查询',pos=(315,5),size=(80,25))
    
  
    
  
    
 house_size_input=wx.TextCtrl(win,pos=(80,5),size=(210,25))
    
 house_size_input_label=wx.StaticText(win,label='房屋大小',pos=(10,10),size=(60,30))
    
  
    
 room_number_input=wx.TextCtrl(win,pos=(80,45),size=(210,25))
    
 room_number_input_label=wx.StaticText(win,label='房间数',pos=(10,50),size=(60,30))
    
  
    
 parlour_number_input=wx.TextCtrl(win,pos=(80,85),size=(210,25))
    
 parlour_number_input_label=wx.StaticText(win,label='厅数',pos=(10,90),size=(60,30))
    
  
    
 house_height_input=wx.TextCtrl(win,pos=(80,125),size=(210,25))
    
 house_height_input_label=wx.StaticText(win,label='层高',pos=(10,130),size=(60,30))
    
  
    
 building_height_input=wx.TextCtrl(win,pos=(80,165),size=(210,25))
    
 building_height_input_label=wx.StaticText(win,label='楼高',pos=(10,170),size=(60,30))
    
  
    
 year_input=wx.TextCtrl(win,pos=(80,205),size=(210,25))
    
 year_input_label=wx.StaticText(win,label='建造年份',pos=(10,210),size=(60,30))
    
  
    
 address_input=wx.TextCtrl(win,pos=(80,245),size=(210,25))
    
 address_input_label=wx.StaticText(win,label='地址',pos=(10,250),size=(60,30))
    
  
    
  
    
 def search(event):
    
     hz = house_size_input.Value
    
     rn=room_number_input.Value
    
     pn=parlour_number_input.Value
    
     hh=house_height_input.Value
    
     bh=building_height_input.Value
    
     yi=year_input.Value
    
     ai=address_input.Value
    
     location=price_mongo.geocode(ai)["geocodes"][0]['location'].encode('utf-8')
    
     house_location_1=location.index(',')
    
     house_location_longtitude=float(location[0:house_location_1])
    
     house_location_latitude=float(location[house_location_1+1:])
    
     list=[]
    
     list.extend([[hz,rn,pn,hh,bh,yi,house_location_longtitude,house_location_latitude]])
    
     out_price=dmodel.predict(list)
    
     print out_price,float(out_price[0])*float(hz)
    
     #print hz,rn,pn,hh,bh,yi,house_location_longtitude,house_location_latitude
    
  
    
 btn.Bind(wx.EVT_BUTTON,search)
    
  
    
 app.MainLoop()
    
    
    
    
    代码解读