python机器学习预测学习笔记（有点问题）

# -*- coding: utf-8 -*-
"""
Created on Wed May 24 17:19:23 2017
@author: yunjinqi 
E-mail:yunjinqi@qq.com 
Differentiate yourself in the world from anyone else.
"""
import pandas as pd
from sklearn.tree import DecisionTreeRegressor  
from sklearn.ensemble import RandomForestRegressor  
import numpy as np  
data=pd.read_csv('C:/Users/HXWD/Desktop/m9000.csv',encoding='gbk')
data.head()
len(data)
data['ret1']=(data['收盘价']-data['收盘价'].shift(1))/data['收盘价'].shift(1)
data['ret']=data['ret1'].shift(-1)
del data['ret1']
data=data[:len(data)-1]
data['ret']
data=data.fillna(0)
df_train=data.ix[::,list(range(7,31))]
df_target=data.ix[::,31]
value=[]
predict_ret=[]
for i in range(len(df_train)):
    if i>300:
        rf=RandomForestRegressor()   
        train=np.array(df_train[i-300:i])
        #print(train)
        target=np.array(df_target[i-300:i])
        rf.fit(train,target)#进行模型的训练  
        aaa=np.array(df_train.ix[i,::])
        aab=list(df_train.ix[i,::])
        a=rf.predict(aaa)
        predict_ret.append(float(a))
        b=df_target.ix[i,::]
        if float(a)*float(b)>=0:
            value.append(1)
        else:
            value.append(0)
value
count=0
for i in range(len(value)):
    if value[i]==1:
        count=count+1
len(value)  
print(count/len(value) )
data.ret.plot()
data.ret.mean()
predict_ret=pd.Series(predict_ret)
d1=pd.DataFrame(data['ret'],)
d1
d1.to_csv('预测.csv')
predict_ret