【机器学习sklearn】基于sklearn的股票预测

最近了解学习数据统计，了解到了python的sklearn这个库，集成了很多机器学习的模型，感觉很强大，官网
下面通过一个简单的预测的例子来上手sklearn这个库。
根据pythonprogramming上的例子改写。本次实验使用anaconda的集成环境，故不需要下载所需的包，但需要使用conda命令更新sklearn至官网教程所使用的版本（之前因为版本问题纠结了好久）。如果没有使用anaconda，需要先下载scipy，numpy等库作为支持，这里使用python自带的pip install命令即可方便下载。
导入相应的库

<span class="hljs-keyword">import</span> pandas <span class="hljs-keyword">as</span> pd
<span class="hljs-keyword">import</span> numpy <span class="hljs-keyword">as</span> np
<span class="hljs-keyword">import</span> datetime
<span class="hljs-keyword">import</span> pandas.io.data <span class="hljs-keyword">as</span> web
<span class="hljs-keyword">import</span> math
<span class="hljs-keyword">import</span> matplotlib.pyplot <span class="hljs-keyword">as</span> plt
<span class="hljs-keyword">from</span> matplotlib <span class="hljs-keyword">import</span> style
<span class="hljs-keyword">from</span> sklearn.model_selection <span class="hljs-keyword">import</span> cross_val_score
<span class="hljs-keyword">from</span> sklearn <span class="hljs-keyword">import</span> preprocessing, cross_validation, svm
<span class="hljs-keyword">from</span> sklearn.linear_model <span class="hljs-keyword">import</span> LinearRegression

import pandas as pd

import numpy as np

import datetime

import pandas.io.data as web

import math

import matplotlib.pyplot as plt

from matplotlib import style

from sklearn.model_selection import cross_val_score

from sklearn import preprocessing, cross_validation, svm

from sklearn.linear_model import LinearRegression

首先从互联网上获取我们所需要的yahoo的股票数据：

start = datetime.datetime(<span class="hljs-number">2015</span>, <span class="hljs-number">1</span>, <span class="hljs-number">1</span>)
<span class="hljs-function"><span class="hljs-keyword">end</span> = <span class="hljs-title">datetime</span>.<span class="hljs-title">datetime</span>(<span class="hljs-title">2016</span>, <span class="hljs-title">11</span>, <span class="hljs-title">20</span>)</span>
<span class="hljs-comment">#从互联网获取数据</span>
df = web.DataReader(<span class="hljs-string">"XOM"</span>, <span class="hljs-string">"yahoo"</span>, start, <span class="hljs-function"><span class="hljs-keyword">end</span>)</span>
<span class="hljs-comment">#print(df.head())</span>

start = datetime.datetime(2015, 1, 1)

end = datetime.datetime(2016, 11, 20)

#从互联网获取数据

df = web.DataReader("XOM", "yahoo", start, end)

#print(df.head())

构建我们的数据集

df = df<span class="hljs-string">[['Open',  'High',  'Low',  'Close', 'Volume']]</span>
df[<span class="hljs-string">'HL_PCT'</span>] = (df[<span class="hljs-string">'High'</span>] - df[<span class="hljs-string">'Low'</span>]) / df[<span class="hljs-string">'Close'</span>] * <span class="hljs-number">100.0</span>
df[<span class="hljs-string">'PCT_change'</span>] = (df[<span class="hljs-string">'Close'</span>] - df[<span class="hljs-string">'Open'</span>]) / df[<span class="hljs-string">'Open'</span>] * <span class="hljs-number">100.0</span>
df = df<span class="hljs-string">[['Close', 'HL_PCT', 'PCT_change', 'Volume']]</span>
#<span class="hljs-built_in">print</span>(df.head())

df = df[['Open', 'High', 'Low', 'Close', 'Volume']]

df['HL_PCT'] = (df['High'] - df['Low']) / df['Close'] * 100.0

df['PCT_change'] = (df['Close'] - df['Open']) / df['Open'] * 100.0

df = df[['Close', 'HL_PCT', 'PCT_change', 'Volume']]

#print(df.head())

对空数据进行处理，同时对Close的股票value进行预测，forecast_out表示往后预测的天数

forecast_col = <span class="hljs-string">'Close'</span>
df.fillna(<span class="hljs-keyword">value</span>=-<span class="hljs-number">99999</span>, inplace=True)
forecast_out = <span class="hljs-keyword">int</span>(math.ceil(<span class="hljs-number">0.01</span> * len(df)))
<span class="hljs-preprocessor">#预测forecast_out天后的</span>

forecast_col = 'Close'

df.fillna(value=-99999, inplace=True)

forecast_out = int(math.ceil(0.01 * len(df)))

#预测forecast_out天后的

重新构建X和y，X为[‘Close’, ‘HL_PCT’, ‘PCT_change’, ‘Volume’]，y为[‘label’]表示forecast_out天后的股票值，使用preprocessing.scale对数据集进行scaling。
X_lately 表示后forecast_out天的数据集，既对应的y值为NAN


df[<span class="hljs-string">'label'</span>] = df[forecast_col].shift(-forecast_out)

<span class="hljs-keyword">print</span>(df.shape)
<span class="hljs-keyword">print</span>(df.tail())
X = np.<span class="hljs-keyword">array</span>(df.drop([<span class="hljs-string">'label'</span>], <span class="hljs-number">1</span>))


X = preprocessing.scale(X)

X_lately = X[-forecast_out:]
X = X[:-forecast_out]
df.dropna(inplace=<span class="hljs-keyword">True</span>)
<span class="hljs-keyword">print</span>(X)
<span class="hljs-keyword">print</span>(X_lately)
y = np.<span class="hljs-keyword">array</span>(df[<span class="hljs-string">'label'</span>])
<span class="hljs-comment">#print(y)</span>
<span class="hljs-keyword">print</span>(X.shape)
<span class="hljs-keyword">print</span>(y.shape)

df['label'] = df[forecast_col].shift(-forecast_out)

print(df.shape)

print(df.tail())

X = np.array(df.drop(['label'], 1))

X = preprocessing.scale(X)

X_lately = X[-forecast_out:]

X = X[:-forecast_out]

df.dropna(inplace=True)

print(X)

print(X_lately)

y = np.array(df['label'])

#print(y)

print(X.shape)

print(y.shape)

选择数据集80%作为训练集，20%作为测试集
使用sklearn提供的Linear Regression函数进行建模，最后使用测试集进行测试，计算相应的精确度

X_train, X_test, y_train ,y_test = cross_validation<span class="hljs-preprocessor">.train</span>_test_split(<span class="hljs-built_in">X</span>,<span class="hljs-built_in">y</span>,test_size=<span class="hljs-number">0.2</span>)

clf = LinearRegression()
clf<span class="hljs-preprocessor">.fit</span>(X_train,y_train)
accuracy = clf<span class="hljs-preprocessor">.score</span>(X_test,y_test)

print(accuracy)

X_train, X_test, y_train ,y_test = cross_validation.train_test_split(X,y,test_size=0.2)

clf = LinearRegression()

clf.fit(X_train,y_train)

accuracy = clf.score(X_test,y_test)

print(accuracy)

forecast_set是我们通过训练集训练出的模型和我们的最近的数据进行的预测
把我们的预测集放入之前的DataFrame，最后绘图得到股票走势图

forecast_<span class="hljs-keyword">set</span> = clf.predict(X_lately)

print(forecast_<span class="hljs-keyword">set</span>,accuracy,forecast_out)

style.use(<span class="hljs-string">'ggplot'</span>)

df[<span class="hljs-string">'Forecast'</span>]=np.nan

last_date = df.iloc[-<span class="hljs-number">1</span>].name
last_unix = last_date.timestamp()
print(last_date,last_unix)
one_day = <span class="hljs-number">86400</span>
next_unix = last_unix + one_day

<span class="hljs-keyword">for</span> i <span class="hljs-keyword">in</span> forecast_<span class="hljs-keyword">set</span>:
    next_date = datetime.datetime.fromtimestamp(next_unix)
    next_unix += <span class="hljs-number">86400</span>
    df.loc[next_date] = [np.nan <span class="hljs-keyword">for</span> _ <span class="hljs-keyword">in</span> range(len(df.columns)-<span class="hljs-number">1</span>)]+[i]


print(df.tail())

df[<span class="hljs-string">'Close'</span>].plot()
df[<span class="hljs-string">'Forecast'</span>].plot()
plt.show()