机器学习实战1--预测链家租房价格

# coding: utf-8

# ### 导入graphlab

# In[112]:

import graphlab


# In[113]:

graphlab.set_runtime_config('GRAPHLAB_DEFAULT_NUM_PYLAMBDA_WORKERS', 4)


# ### 读取excel文件

# In[114]:

houses = graphlab.SFrame.read_csv('/Users/Redheat/Downloads/lianjia.csv') #读取csv文件


# In[115]:

print houses


# ### 在浏览器打开,设定x和y轴

# In[116]:

graphlab.canvas.set_target('browser')#在浏览器打开
houses.show(view="Scatter Plot", x="size", y="price")


# ### 按百分比区分训练集和测试集,然后创建一个线性回归模型

# In[117]:

train_data,test_data = houses.random_split(.8,seed=0)#按80%分成测试集和训练集


# In[118]:

sqft_model = graphlab.linear_regression.create(train_data, target='price', features=['size'],validation_set=None)#创建一个线性回归模型


# In[119]:

print test_data['price'].mean() #平均值


# In[120]:

print sqft_model.evaluate(test_data)#模型评估函数


# ### 绘制图形

# In[121]:

#绘图
import matplotlib.pyplot as plt 
#在notebook绘图
get_ipython().magic(u'matplotlib inline')


# In[122]:

#分别以.和-绘制图形
plt.plot(test_data['size'],test_data['price'],'.',
        test_data['size'],sqft_model.predict(test_data),'-')


# In[123]:

sqft_model.get('coefficients') #获取权重


# ### 增加新特征

# In[124]:

house_features = ['village', 'room', 'size', 'direction', 'age','area','position','id']


# In[125]:

houses[house_features].show()


# In[99]:

houses.show(view='BoxWhisker Plot', x='area', y='price')


# ### 创建基于新特征的线性回归

# In[110]:

house_features_model = graphlab.linear_regression.create(train_data,target='price',features=house_features,validation_set=None) #更多特征


# In[101]:

print sqft_model.evaluate(test_data) #第一个模型评估
print house_features_model.evaluate(test_data) #多特征模型评估


# ### 价格获取和预测

# In[102]:

house1 = houses[houses['id']=='BJ0004399001']


# In[103]:

print house1['price'] #真实价格


# In[104]:

print sqft_model.predict(house1)#单特征预测价格


# In[111]:

print house_features_model.predict(house1) #多特征预测价格


# <img src="https://image1.ljcdn.com/lianjia-data-sync/ziroom/15289570226877_2440114952_0.jpg.600x450.jpg">

# In[ ]:

代码地址(附作业答案): https://github.com/RedheatWei/aiproject/tree/master/Machine%20Learning%20Specialization/week2