房价预测数据分析

import pandas as pd import matplotlib.pyplot as plt import numpy as np import seaborn as sns from sklearn.model_selection import train_test_split #这里是引用了交叉验证 from sklearn.linear_model import LinearRegression #线性回归 from sklearn.tree import DecisionTreeRegressor#决策树 from sklearn.neighbors import KNeighborsClassifier plt.rcParams['font.sans-serif'] = ['SimHei'] #配置显示中文，否则乱码 plt.rcParams['axes.unicode_minus']=False #用来正常显示负号，如果是plt画图，则将mlp换成plt df=pd.read_csv('date.csv') df=df.sort_values(by='Year')#将数据按年份进行排序q df=df.reset_index()#根据年份值对数据重新排序 delete=df[df['Year']<1944].index#异常值的索引 df=df.drop(delete,axis=0)#删除异常值 df=df.drop(['index'],axis=1)#删除无用的index列 print(df.head()) corr=df.corr() plt.title("各变量之间的相关性") sns.heatmap(corr, annot=True) plt.show() print(corr['Value']) plt.figure() plt.title("Value与各变量之间的相关性") corr['Value'].plot() plt.show() a=df.isnull().sum() for i in a.index: if(a[i]!=0): df[i].fillna(0) print("填充完毕") df['Type']=df['Room']+df['Living']#Room与Living的相关性相似，可以合并这两列构建一列新的变量 X=df[['Area','Type','Year','Floor']] y=df['Value'] length=len(X) X_train=X[:int(length*0.7)] X_test=X[int(length*0.7):] y_train=y[:int(length*0.7)] y_test=y[int(length*0.7):] def train_model(method,name): if(name=='knn'): k=[i+1 for i in range(10)] train_accuracy={} for n_neighbors in k: knn=KNeighborsClassifier(n_neighbors) knn.fit(X_train,y_train) y_predict=knn.predict(X_test) train_accuracy[n_neighbors]=knn.score(X_train,y_train) arg = max(train_accuracy, key=train_accuracy.get) linreg=KNeighborsClassifier(arg) else: linreg= method model=linreg.fit(X_train, y_train) y_pred=linreg.predict(X_test) plt.figure(figsize=(10,6)) if name=='knn': plt.title(name+"预测结果k=%d"%arg) else: plt.title(name+"预测结果") plt.plot(X_test['Year'].sort_values(),y_test,label='测试值') plt.plot(X_test['Year'].sort_values(),y_pred,label='预测值') plt.plot(X_train['Year'].sort_values(),y_train,label='训练数据') plt.legend() plt.show() train_model(LinearRegression(),"线性回归") train_model(DecisionTreeRegressor(),"决策树") train_model(KNeighborsClassifier(),"knn")