Python股票数据分析
最近在学习基于python的股票数据分析,其中主要用到了tushare和seaborn。tushare是一款财经类数据接口包,国内的股票数据还是比较全的
官网地址:http://tushare.waditu.com/index.html#id5。seaborn则是一款绘图库,通过seaborn可以轻松地画出简洁漂亮的图表,而且库本身具有一定的统计功能。
导入的模块:
import matplotlib.pyplot as plt
import seaborn as sns
import seaborn.linearmodels as snsl
from datetime import datetime
import tushare as ts
代码部分:
股票收盘价走势曲线
sns.set_style("whitegrid")
end = datetime.today() #开始时间结束时间,选取最近一年的数据
start = datetime(end.year-1,end.month,end.day)
end = str(end)[0:10]
start = str(start)[0:10]
stock = ts.get_hist_data(\'300104\',start,end)#选取一支股票
stock[\'close\'].plot(legend=True ,figsize=(10,4))
plt.show()
股票日线
同理,可以做出5日均线、10日均线以及20日均线
stock[[\'close\',\'ma5\',\'ma10\',\'ma20\']].plot(legend=True ,figsize=(10,4))
日线、5日均线、10日均线、20日均线
股票每日涨跌幅度
stock[\'Daily Return\'] = stock[\'close\'].pct_change()
stock[\'Daily Return\'].plot(legend=True,figsize=(10,4))
每日涨跌幅
核密度估计
sns.kdeplot(stock[\'Daily Return\'].dropna())
核密度估计
核密度估计+统计柱状图
sns.distplot(stock[\'Daily Return\'].dropna(),bins=100)
核密度+柱状图
两支股票的皮尔森相关系数
sns.jointplot(stock[\'Daily Return\'],stock[\'Daily Return\'],alpha=0.2)
皮尔森相关系数
多只股票相关性计算
stock_lis=[\'300113\',\'300343\',\'300295\',\'300315`] #随便选取了四支互联网相关的股票
df=pd.DataFrame()
for stock in stock_lis: closing_df = ts.get_hist_data(stock,start,end)[\'close\'] df = df.join(pd.DataFrame({stock:closing_df}),how=\'outer\')
tech_rets = df.pct_change()
snsl.corrplot(tech_rets.dropna())
相关性
简单地计算股票的收益与风险,衡量股票收益与风险的数值分别为股票涨跌的平均值以及标准差,平均值为正则说明收益是正的,标准差越大则说明股票波动大,风险也大。
rets = tech_rets.dropna()
plt.scatter(rets.mean(),rets.std())
plt.xlabel(\'Excepted Return\')
plt.ylabel(\'Risk\')
for label,x,y in zip(rets.columns,rets.mean(),rets.std()):#添加标注 plt.annotate( label, xy =(x,y),xytext=(15,15), textcoords = \'offset points\', arrowprops = dict(arrowstyle = \'-\',connectionstyle = \'arc3,rad=-0.3\'))
用Python分析公开数据选出高送转预期股票
根据以往的经验,每年年底都会有一波高送转预期行情。今天,米哥就带大家实践一下如何利用tushare实现高送转预期选股。
本文主要是讲述选股的思路方法,选股条件和参数大家可以根据米哥提供的代码自行修改。
1. 选股原理
一般来说,具备高送转预期的个股,都具有总市值低、每股公积金高、每股收益大,流通股本少的特点。当然,也还有其它的因素,比如当前股价、经营收益变动情况、以及以往分红送股习惯等等。
这里我们暂时只考虑每股公积金、每股收益、流通股本和总市值四个因素,将公积金大于等于5元,每股收益大于等于5毛,流通股本在3亿以下,总市值在100亿以内作为高送转预期目标(这些参数大家可根据自己的经验随意调整)。
2. 数据准备
首先要导入tushare:
import tushare as ts
调取股票基本面数据和行情数据
# 基本面数据
basic = ts.get_stock_basics()
# 行情和市值数据
hq = ts.get_today_all()
3. 数据清洗整理
对获取到的数据进行清洗和整理,只保留需要的字段。(其它字段及含义,请参考 http:// tushare.org 文档)
#当前股价,如果停牌则设置当前价格为上一个交易日股价
hq[\'trade\'] = hq.apply(lambda x:x.settlement if x.trade==0 else x.trade, axis=1)
#分别选取流通股本,总股本,每股公积金,每股收益
basedata = basic[[\'outstanding\', \'totals\', \'reservedPerShare\', \'esp\']]
#选取股票代码,名称,当前价格,总市值,流通市值
hqdata = hq[[\'code\', \'name\', \'trade\', \'mktcap\', \'nmc\']]
#设置行情数据code为index列
hqdata = hqdata.set_index(\'code\')
#合并两个数据表
data = basedata.merge(hqdata, left_index=True, right_index=True)
4. 选股条件
根据上文提到的选股参数和条件,我们对数据进一步处理。
将总市值和流通市值换成亿元单位
data[\'mktcap\'] = data[\'mktcap\'] / 10000
data[\'nmc\'] = data[\'nmc\'] / 10000
设置参数和过滤值(此次各自调整)
#每股公积金>=5
res = data.reservedPerShare >= 5
#流通股本<=3亿
out = data.outstanding <= 30000
#每股收益>=5毛
eps = data.esp >= 0.5
#总市值<100亿
mktcap = data.mktcap <= 100
取并集结果:
allcrit = res & out & eps & mktcap
selected = data[allcrit]
具有高送转预期股票的结果呈现:
以上字段的含义分别为:股票名称、收盘价格、每股公积金、流通股本、每股收益(应该为eps,之前发布笔误)、总市值和流通市值。
https://zhuanlan.zhihu.com/p/23829205
Python 金叉判定
def jincha(context, bar_dict, his): #站上5日线 def zs5(context, bar_dict, his): ma_n = pd.rolling_mean(his, 5) temp = his - ma_n #temp_s包含了前一天站上五日线得股票代码 temp_s = list(temp[temp>0].iloc[-1,:].dropna().index) return temp_s #站上10日线 def zs10(context, bar_dict, his): ma_n = pd.rolling_mean(his, 10) temp = his - ma_n temp_s = list(temp[temp>0].iloc[-1,:].dropna().index) return temp_s #金叉突破 def jc(context, bar_dict, his): mas = pd.rolling_mean(his,5) mal = pd.rolling_mean(his, 10) temp = mas - mal #temp_jc昨天大于0股票代码 #temp_r前天大于0股票代码 temp_jc = list(temp[temp>0].iloc[-1,:].dropna().index) temp_r = list(temp[temp>0].iloc[-2,:].dropna().index) temp = [] for stock in temp_jc: if stock not in temp_r: temp.append(stock) return temp #求三种条件下的股票代码交集 con1 = zs5(context, bar_dict, his) con2 = zs10(context, bar_dict, his) con3 = jc(context, bar_dict, his) tar_list=[con1,con2,con3] tarstock = tar_list[0] for i in tar_list: tarstock = list(set(tarstock).intersection(set(i))) return tarstock
Python 过滤次新股、停牌、涨跌停
#过滤次新股、是否涨跌停、是否停牌等条件 def filcon(context,bar_dict,tar_list): def zdt_trade(stock, context, bar_dict): yesterday = history(2,\'1d\', \'close\')[stock].values[-1] zt = round(1.10 * yesterday,2) dt = round(0.99 * yesterday,2) #last最后交易价 return dt < bar_dict[stock].last < zt filstock = [] for stock in tar_list: con1 = ipo_days(stock,context.now) > 60 con2 = bar_dict[stock].is_trading con3 = zdt_trade(stock,context,bar_dict) if con1 & con2 & con3: filstock.append(stock) return filstock
Python 按平均持仓市值调仓
# 按平均持仓市值调仓 def for_balance(context, bar_dict): #mvalues = context.portfolio.market_value #avalues = context.portfolio.portfolio_value #per = mvalues / avalues hlist = [] for stock in context.portfolio.positions: #获取股票及对应持仓市值 hlist.append([stock,bar_dict[stock].last * context.portfolio.positions[stock].quantity]) if hlist: #按持仓市值由大到小排序 hlist = sorted(hlist,key=lambda x:x[1], reverse=True) temp = 0 for li in hlist: #计算持仓总市值 temp += li[1] for li in hlist: #平均各股持仓市值 if bar_dict[li[0]].is_trading: order_target_value(li[0], temp/len(hlist)) return
Python PCA主成分分析算法
Python主成分分析算法的作用是提取样本的主要特征向量,从而实现数据降维的目的。
# -*- coding: utf-8 -*- """ Created on Sun Feb 28 10:04:26 2016 PCA source code @author: liudiwei """ import numpy as np import pandas as pd import matplotlib.pyplot as plt #计算均值,要求输入数据为numpy的矩阵格式,行表示样本数,列表示特征 def meanX(dataX): return np.mean(dataX,axis=0)#axis=0表示按照列来求均值,如果输入list,则axis=1 #计算方差,传入的是一个numpy的矩阵格式,行表示样本数,列表示特征 def variance(X): m, n = np.shape(X) mu = meanX(X) muAll = np.tile(mu, (m, 1)) X1 = X - muAll variance = 1./m * np.diag(X1.T * X1) return variance #标准化,传入的是一个numpy的矩阵格式,行表示样本数,列表示特征 def normalize(X): m, n = np.shape(X) mu = meanX(X) muAll = np.tile(mu, (m, 1)) X1 = X - muAll X2 = np.tile(np.diag(X.T * X), (m, 1)) XNorm = X1/X2 return XNorm """ 参数: - XMat:传入的是一个numpy的矩阵格式,行表示样本数,列表示特征 - k:表示取前k个特征值对应的特征向量 返回值: - finalData:参数一指的是返回的低维矩阵,对应于输入参数二 - reconData:参数二对应的是移动坐标轴后的矩阵 """ def pca(XMat, k): average = meanX(XMat) m, n = np.shape(XMat) data_adjust = [] avgs = np.tile(average, (m, 1)) data_adjust = XMat - avgs covX = np.cov(data_adjust.T) #计算协方差矩阵 featValue, featVec= np.linalg.eig(covX) #求解协方差矩阵的特征值和特征向量 index = np.argsort(-featValue) #按照featValue进行从大到小排序 finalData = [] if k > n: print "k must lower than feature number" return else: #注意特征向量时列向量,而numpy的二维矩阵(数组)a[m][n]中,a[1]表示第1行值 selectVec = np.matrix(featVec.T[index[:k]]) #所以这里需要进行转置 finalData = data_adjust * selectVec.T reconData = (finalData * selectVec) + average return finalData, reconData def loaddata(datafile): return np.array(pd.read_csv(datafile,sep="\t",header=-1)).astype(np.float) def plotBestFit(data1, data2): dataArr1 = np.array(data1) dataArr2 = np.array(data2) m = np.shape(dataArr1)[0] axis_x1 = [] axis_y1 = [] axis_x2 = [] axis_y2 = [] for i in range(m): axis_x1.append(dataArr1[i,0]) axis_y1.append(dataArr1[i,1]) axis_x2.append(dataArr2[i,0]) axis_y2.append(dataArr2[i,1]) fig = plt.figure() ax = fig.add_subplot(111) ax.scatter(axis_x1, axis_y1, s=50, c=\'red\', marker=\'s\') ax.scatter(axis_x2, axis_y2, s=50, c=\'blue\') plt.xlabel(\'x1\'); plt.ylabel(\'x2\'); plt.savefig("outfile.png") plt.show() #简单测试 #数据来源:http://www.cnblogs.com/jerrylead/archive/2011/04/18/2020209.html def test(): X = [[2.5, 0.5, 2.2, 1.9, 3.1, 2.3, 2, 1, 1.5, 1.1], [2.4, 0.7, 2.9, 2.2, 3.0, 2.7, 1.6, 1.1, 1.6, 0.9]] XMat = np.matrix(X).T k = 2 return pca(XMat, k) #根据数据集data.txt def main(): datafile = "data.txt" XMat = loaddata(datafile) k = 2 return pca(XMat, k) if __name__ == "__main__": finalData, reconMat = main() plotBestFit(finalData, reconMat)
经过主成分降维的数据如红色图案所示,蓝色的是恢复的原始数据。可以看到经过降维的数据样本差异更加明显。
Python KNN最近邻分类算法
# -*- coding: utf-8 -*- """ Created on Mon Feb 22 13:21:22 2016 K-NearestNeighbor """ import numpy as np import operator class KNNClassifier(): """This is a Nearest Neighbor classifier. """ #定义k的值 def __init__(self, k=3): self._k = k #计算新样本与已知分类样本的距离并从小到大排列 def _calEDistance(self, inSample, dataset): m = dataset.shape[0] diffMat = np.tile(inSample, (m,1)) - dataset sqDiffMat = diffMat**2 #每个元素平方 sqDistances = sqDiffMat.sum(axis = 1) #求和 distances = sqDistances ** 0.5 #开根号 return distances.argsort() #按距离的从小到达排列的下标值 def _classify0(self, inX, dataSet, labels): k = self._k dataSetSize = dataSet.shape[0] diffMat = np.tile(inX, (dataSetSize,1)) - dataSet sqDiffMat = diffMat**2 sqDistances = sqDiffMat.sum(axis=1) distances = sqDistances**0.5 sortedDistIndicies = distances.argsort() classCount={} for i in range(k): voteIlabel = labels[sortedDistIndicies[i]] classCount[voteIlabel] = classCount.get(voteIlabel,0) + 1 sortedClassCount = sorted(classCount.iteritems(), key=operator.itemgetter(1), reverse=True) return sortedClassCount[0][0] #对一个样本进行分类 def _classify(self, sample, train_X, train_y): #数据类型检测 if isinstance(sample, np.ndarray) and isinstance(train_X, np.ndarray) \ and isinstance(train_y, np.ndarray): pass else: try: sample = np.array(sample) train_X = np.array(train_X) train_y = np.array(train_y) except: raise TypeError("numpy.ndarray required for train_X and ..") sortedDistances = self._calEDistance(sample, train_X) classCount = {} for i in range(self._k): oneVote = train_y[sortedDistances[i]] #获取最近的第i个点的类别 classCount[oneVote] = classCount.get(oneVote, 0) + 1 sortedClassCount = sorted(classCount.iteritems(),\ key=operator.itemgetter(1), reverse=True) #print "the sample :", sample, "is classified as",sortedClassCount[0][0] return sortedClassCount[0][0] def classify(self, test_X, train_X, train_y): results = [] #数据类型检测 if isinstance(test_X, np.ndarray) and isinstance(train_X, np.ndarray) \ and isinstance(train_y, np.ndarray): pass else: try: test_X = np.array(test_X) train_X = np.array(train_X) train_y = np.array(train_y) except: raise TypeError("numpy.ndarray required for train_X and ..") d = len(np.shape(test_X)) if d == 1: sample = test_X result = self._classify(sample, train_X, train_y) results.append(result) else: for i in range(len(test_X)): sample = test_X[i] result = self._classify(sample, train_X, train_y) results.append(result) return results if __name__=="__main__": train_X = [[1, 2, 0, 1, 0], [0, 1, 1, 0, 1], [1, 0, 0, 0, 1], [2, 1, 1, 0, 1], [1, 1, 0, 1, 1]] train_y = [1, 1, 0, 0, 0] clf = KNNClassifier(k = 3) sample = [[1,2,0,1,0],[1,2,0,1,1]] result = clf.classify(sample, train_X, train_y)
第二部分:KNN测试代码
# -*- coding: utf-8 -*- """ Created on Mon Feb 22 13:21:22 2016 K-NearestNeighbor """ import numpy as np import operator class KNNClassifier(): """This is a Nearest Neighbor classifier. """ #定义k的值 def __init__(self, k=3): self._k = k #计算新样本与已知分类样本的距离并从小到大排列 def _calEDistance(self, inSample, dataset): m = dataset.shape[0] diffMat = np.tile(inSample, (m,1)) - dataset sqDiffMat = diffMat**2 #每个元素平方 sqDistances = sqDiffMat.sum(axis = 1) #求和 distances = sqDistances ** 0.5 #开根号 return distances.argsort() #按距离的从小到达排列的下标值 def _classify0(self, inX, dataSet, labels): k = self._k dataSetSize = dataSet.shape[0] diffMat = np.tile(inX, (dataSetSize,1)) - dataSet sqDiffMat = diffMat**2 sqDistances = sqDiffMat.sum(axis=1) distances = sqDistances**0.5 sortedDistIndicies = distances.argsort() classCount={} for i in range(k): voteIlabel = labels[sortedDistIndicies[i]] classCount[voteIlabel] = classCount.get(voteIlabel,0) + 1 sortedClassCount = sorted(classCount.iteritems(), key=operator.itemgetter(1), reverse=True) return sortedClassCount[0][0] #对一个样本进行分类 def _classify(self, sample, train_X, train_y): #数据类型检测 if isinstance(sample, np.ndarray) and isinstance(train_X, np.ndarray) \ and isinstance(train_y, np.ndarray): pass else: try: sample = np.array(sample) train_X = np.array(train_X) train_y = np.array(train_y) except: raise TypeError("numpy.ndarray required for train_X and ..") sortedDistances = self._calEDistance(sample, train_X) classCount = {} for i in range(self._k): oneVote = train_y[sortedDistances[i]] #获取最近的第i个点的类别 classCount[oneVote] = classCount.get(oneVote, 0) + 1 sortedClassCount = sorted(classCount.iteritems(),\ key=operator.itemgetter(1), reverse=True) #print "the sample :", sample, "is classified as",sortedClassCount[0][0] return sortedClassCount[0][0] def classify(self, test_X, train_X, train_y): results = [] #数据类型检测 if isinstance(test_X, np.ndarray) and isinstance(train_X, np.ndarray) \ and isinstance(train_y, np.ndarray): pass else: try: test_X = np.array(test_X) train_X = np.array(train_X) train_y = np.array(train_y) except: raise TypeError("numpy.ndarray required for train_X and ..") d = len(np.shape(test_X)) if d == 1: sample = test_X result = self._classify(sample, train_X, train_y) results.append(result) else: for i in range(len(test_X)): sample = test_X[i] result = self._classify(sample, train_X, train_y) results.append(result) return results if __name__=="__main__": train_X = [[1, 2, 0, 1, 0], [0, 1, 1, 0, 1], [1, 0, 0, 0, 1], [2, 1, 1, 0, 1], [1, 1, 0, 1, 1]] train_y = [1, 1, 0, 0, 0] clf = KNNClassifier(k = 3) sample = [[1,2,0,1,0],[1,2,0,1,1]] result = clf.classify(sample, train_X, train_y)
Python 决策树算法(ID3 &C4.5)
决策树(Decision Tree)算法:按照样本的属性逐步进行分类,为了能够使分类更快、更有效。每一个新分类属性的选择依据可以是信息增益IG和信息增益率IGR,前者为最基本的ID3算法,后者为改进后的C4.5算法。
以ID3为例,其训练过程的编程思路如下:
(1)输入x、y(x为样本,y为label),行为样本,列为样本特征。
(2)计算信息增益IG,获取使IG最大的特征。
(3)获得删除最佳分类特征后的样本阵列。
(4)按照最佳分类特征的属性值将更新后的样本进行归类。
属性值1(x1,y1) 属性值2(x2,y2) 属性值(x3,y3)
(5)分别对以上类别重复以上操作直至到达叶节点(递归调用)。
叶节点的特征:
(1)所有的标签值y都一样。
(2)没有特征可以继续划分。
测试过程的编程思路如下:
(1)读取训练好的决策树。
(2)从根节点开始递归遍历整个决策树直到到达叶节点为止。
以下为具体代码,训练后的决策树结构为递归套用的字典,其是由特征值组成的索引加上label组成的。
# -*- coding: utf-8 -*- """ Created on Mon Nov 07 09:06:37 2016 @author: yehx """ # -*- coding: utf-8 -*- """ Created on Sun Feb 21 12:17:10 2016 Decision Tree Source Code @author: liudiwei """ import os import numpy as np class DecitionTree(): """This is a decision tree classifier. """ def __init__(self, criteria=\'ID3\'): self._tree = None if criteria == \'ID3\' or criteria == \'C4.5\': self._criteria = criteria else: raise Exception("criterion should be ID3 or C4.5") def _calEntropy(slef, y): \'\'\' 功能:_calEntropy用于计算香农熵 e=-sum(pi*log pi) 参数:其中y为数组array 输出:信息熵entropy \'\'\' n = y.shape[0] labelCounts = {} for label in y: if label not in labelCounts.keys(): labelCounts[label] = 1 else: labelCounts[label] += 1 entropy = 0.0 for key in labelCounts: prob = float(labelCounts[key])/n entropy -= prob * np.log2(prob) return entropy def _splitData(self, X, y, axis, cutoff): """ 参数:X为特征,y为label,axis为某个特征的下标,cutoff是下标为axis特征取值值 输出:返回数据集中特征下标为axis,特征值等于cutoff的子数据集 先将特征列从样本矩阵里除去,然后将属性值为cutoff的数据归为一类 """ ret = [] featVec = X[:,axis] n = X.shape[1] #特征个数 #除去第axis列特征后的样本矩阵 X = X[:,[i for i in range(n) if i!=axis]] for i in range(len(featVec)): if featVec[i] == cutoff: ret.append(i) return X[ret, :], y[ret] def _chooseBestSplit(self, X, y): """ID3 & C4.5 参数:X为特征,y为label 功能:根据信息增益或者信息增益率来获取最好的划分特征 输出:返回最好划分特征的下标 """ numFeat = X.shape[1] baseEntropy = self._calEntropy(y) bestSplit = 0.0 best_idx = -1 for i in range(numFeat): featlist = X[:,i] #得到第i个特征对应的特征列 uniqueVals = set(featlist) curEntropy = 0.0 splitInfo = 0.0 for value in uniqueVals: sub_x, sub_y = self._splitData(X, y, i, value) prob = len(sub_y)/float(len(y)) #计算某个特征的某个值的概率 curEntropy += prob * self._calEntropy(sub_y) #迭代计算条件熵 splitInfo -= prob * np.log2(prob) #分裂信息,用于计算信息增益率 IG = baseEntropy - curEntropy if self._criteria=="ID3": if IG > bestSplit: bestSplit = IG best_idx = i if self._criteria=="C4.5": if splitInfo == 0.0: pass IGR = IG/splitInfo if IGR > bestSplit: bestSplit = IGR best_idx = i return best_idx def _majorityCnt(self, labellist): """ 参数:labellist是类标签,序列类型为list 输出:返回labellist中出现次数最多的label """ labelCount={} for vote in labellist: if vote not in labelCount.keys(): labelCount[vote] = 0 labelCount[vote] += 1 sortedClassCount = sorted(labelCount.iteritems(), key=lambda x:x[1], \ reverse=True) return sortedClassCount[0][0] def _createTree(self, X, y, featureIndex): """ 参数:X为特征,y为label,featureIndex类型是元组,记录X特征在原始数据中的下标 输出:根据当前的featureIndex创建一颗完整的树 """ labelList = list(y) #如果所有的标签都一样(叶节点),直接返回标签 if labelList.count(labelList[0]) == len(labelList): return labelList[0] #如果没有特征可以继续划分,那么将所有的label归为大多数的一类,并返回标签 if len(featureIndex) == 0: return self._majorityCnt(labelList) #返回最佳分类特征的下标 bestFeatIndex = self._chooseBestSplit(X,y) #返回最佳分类特征的索引 bestFeatAxis = featureIndex[bestFeatIndex] featureIndex = list(featureIndex) #获得删除最佳分类特征索引后的列表 featureIndex.remove(bestFeatAxis) featureIndex = tuple(featureIndex) myTree = {bestFeatAxis:{}} featValues = X[:, bestFeatIndex] uniqueVals = set(featValues) for value in uniqueVals: #对每个value递归地创建树 sub_X, sub_y = self._splitData(X,y, bestFeatIndex, value) myTree[bestFeatAxis][value] = self._createTree(sub_X, sub_y, \ featureIndex) return myTree def fit(self, X, y): """ 参数:X是特征,y是类标签 注意事项:对数据X和y进行类型检测,保证其为array 输出:self本身 """ if isinstance(X, np.ndarray) and isinstance(y, np.ndarray): pass else: try: X = np.array(X) y = np.array(y) except: raise TypeError("numpy.ndarray required for X,y") featureIndex = tuple([\'x\'+str(i) for i in range(X.shape[1])]) self._tree = self._createTree(X,y,featureIndex) return self #allow using: clf.fit().predict() def _classify(self, tree, sample): """ 用训练好的模型对输入数据进行分类 注意:决策树的构建是一个递归的过程,用决策树分类也是一个递归的过程 _classify()一次只能对一个样本(sample)分类 """ featIndex = tree.keys()[0] #得到数的根节点值 secondDict = tree[featIndex] #得到以featIndex为划分特征的结果 axis=featIndex[1:] #得到根节点特征在原始数据中的下标 key = sample[int(axis)] #获取待分类样本中下标为axis的值 valueOfKey = secondDict[key] #获取secondDict中keys为key的value值 if type(valueOfKey).__name__==\'dict\': #如果value为dict,则继续递归分类 return self._classify(valueOfKey, sample) else: return valueOfKey def predict(self, X): if self._tree==None: raise NotImplementedError("Estimator not fitted, call `fit` first") #对X的类型进行检测,判断其是否是数组 if isinstance(X, np.ndarray): pass else: try: X = np.array(X) except: raise TypeError("numpy.ndarray required for X") if len(X.shape) == 1: return self._classify(self._tree, X) else: result = [] for i in range(X.shape[0]): value = self._classify(self._tree, X[i]) print str(i+1)+"-th sample is classfied as:", value result.append(value) return np.array(result) def show(self, outpdf): if self._tree==None: pass #plot the tree using matplotlib import treePlotter treePlotter.createPlot(self._tree, outpdf) if __name__=="__main__": trainfile=r"data\train.txt" testfile=r"data\test.txt" import sys sys.path.append(r"F:\CSU\Github\MachineLearning\lib") import dataload as dload train_x, train_y = dload.loadData(trainfile) test_x, test_y = dload.loadData(testfile) clf = DecitionTree(criteria="C4.5") clf.fit(train_x, train_y) result = clf.predict(test_x) outpdf = r"tree.pdf" clf.show(outpdf)
Python K均值聚类
Python K均值聚类是一种无监督的机器学习算法,能够实现自动归类的功能。
算法步骤如下:
(1)随机产生K个分类中心,一般称为质心。
(2)将所有样本划分到距离最近的质心代表的分类中。(距离可以是欧氏距离、曼哈顿距离、夹角余弦等)
(3)计算分类后的质心,可以用同一类中所有样本的平均属性来代表新的质心。
(4)重复(2)(3)两步,直到满足以下其中一个条件:
1)分类结果没有发生改变。
2)最小误差(如平方误差)达到所要求的范围。
3)迭代总数达到设置的最大值。
常见的K均值聚类算法还有2分K均值聚类算法,其步骤如下:
(1)将所有样本作为一类。
(2)按照传统K均值聚类的方法将样本分为两类。
(3)对以上两类分别再分为两类,且分别计算两种情况下误差,仅保留误差更小的分类;即第(2)步产生的两类其中一类保留,另一类进行再次分类。
(4)重复对已有类别分别进行二分类,同理保留误差最小的分类,直到达到所需要的分类数目。
具体Python代码如下:
# -*- coding: utf-8 -*- """ Created on Tue Nov 08 14:01:44 2016 K - means cluster """ import numpy as np class KMeansClassifier(): "this is a k-means classifier" def __init__(self, k=3, initCent=\'random\', max_iter=500 ): self._k = k self._initCent = initCent self._max_iter = max_iter self._clusterAssment = None self._labels = None self._sse = None def _calEDist(self, arrA, arrB): """ 功能:欧拉距离距离计算 输入:两个一维数组 """ return np.math.sqrt(sum(np.power(arrA-arrB, 2))) def _calMDist(self, arrA, arrB): """ 功能:曼哈顿距离距离计算 输入:两个一维数组 """ return sum(np.abs(arrA-arrB)) def _randCent(self, data_X, k): """ 功能:随机选取k个质心 输出:centroids #返回一个m*n的质心矩阵 """ n = data_X.shape[1] #获取特征的维数 centroids = np.empty((k,n)) #使用numpy生成一个k*n的矩阵,用于存储质心 for j in range(n): minJ = min(data_X[:, j]) rangeJ = float(max(data_X[:, j] - minJ)) #使用flatten拉平嵌套列表(nested list) centroids[:, j] = (minJ + rangeJ * np.random.rand(k, 1)).flatten() return centroids def fit(self, data_X): """ 输入:一个m*n维的矩阵 """ if not isinstance(data_X, np.ndarray) or \ isinstance(data_X, np.matrixlib.defmatrix.matrix): try: data_X = np.asarray(data_X) except: raise TypeError("numpy.ndarray resuired for data_X") m = data_X.shape[0] #获取样本的个数 #一个m*2的二维矩阵,矩阵第一列存储样本点所属的族的索引值, #第二列存储该点与所属族的质心的平方误差 self._clusterAssment = np.zeros((m,2)) if self._initCent == \'random\': self._centroids = self._randCent(data_X, self._k) clusterChanged = True for _ in range(self._max_iter): #使用"_"主要是因为后面没有用到这个值 clusterChanged = False for i in range(m): #将每个样本点分配到离它最近的质心所属的族 minDist = np.inf #首先将minDist置为一个无穷大的数 minIndex = -1 #将最近质心的下标置为-1 for j in range(self._k): #次迭代用于寻找最近的质心 arrA = self._centroids[j,:] arrB = data_X[i,:] distJI = self._calEDist(arrA, arrB) #计算误差值 if distJI <</span> minDist: minDist = distJI minIndex = j if self._clusterAssment[i,0] !=minIndex: clusterChanged = True self._clusterAssment[i,:] = minIndex, minDist**2 if not clusterChanged:#若所有样本点所属的族都不改变,则已收敛,结束迭代 break for i in range(self._k):#更新质心,将每个族中的点的均值作为质心 index_all = self._clusterAssment[:,0] #取出样本所属簇的索引值 value = np.nonzero(index_all==i) #取出所有属于第i个簇的索引值 ptsInClust = data_X[value[0]] #取出属于第i个簇的所有样本点 self._centroids[i,:] = np.mean(ptsInClust, axis=0) #计算均值 self._labels = self._clusterAssment[:,0] self._sse = sum(self._clusterAssment[:,1]) def predict(self, X):#根据聚类结果,预测新输入数据所属的族 #类型检查 if not isinstance(X,np.ndarray): try: X = np.asarray(X) except: raise TypeError("numpy.ndarray required for X") m = X.shape[0]#m代表样本数量 preds = np.empty((m,)) for i in range(m):#将每个样本点分配到离它最近的质心所属的族 minDist = np.inf for j in range(self._k): distJI = self._calEDist(self._centroids[j,:], X[i,:]) if distJI <</span> minDist: minDist = distJI preds[i] = j return preds class biKMeansClassifier(): "this is a binary k-means classifier" def __init__(self, k=3): self._k = k self._centroids = None self._clusterAssment = None self._labels = None self._sse = None def _calEDist(self, arrA, arrB): """ 功能:欧拉距离距离计算 输入:两个一维数组 """ return np.math.sqrt(sum(np.power(arrA-arrB, 2))) def fit(self, X): m = X.shape[0] self._clusterAssment = np.zeros((m,2)) centroid0 = np.mean(X, axis=0).tolist() centList =[centroid0] for j in range(m):#计算每个样本点与质心之间初始的平方误差 self._clusterAssment[j,1] = self._calEDist(np.asarray(centroid0), \ X[j,:])**2 while (len(centList) <</span> self._k): lowestSSE = np.inf #尝试划分每一族,选取使得误差最小的那个族进行划分 for i in range(len(centList)): index_all = self._clusterAssment[:,0] #取出样本所属簇的索引值 value = np.nonzero(index_all==i) #取出所有属于第i个簇的索引值 ptsInCurrCluster = X[value[0],:] #取出属于第i个簇的所有样本点 clf = KMeansClassifier(k=2) clf.fit(ptsInCurrCluster) #划分该族后,所得到的质心、分配结果及误差矩阵 centroidMat, splitClustAss = clf._centroids, clf._clusterAssment sseSplit = sum(splitClustAss[:,1]) index_all = self._clusterAssment[:,0] value = np.nonzero(index_all==i) sseNotSplit = sum(self._clusterAssment[value[0],1]) if (sseSplit + sseNotSplit) <</span> lowestSSE: bestCentToSplit = i bestNewCents = centroidMat bestClustAss = splitClustAss.copy() lowestSSE = sseSplit + sseNotSplit #该族被划分成两个子族后,其中一个子族的索引变为原族的索引 #另一个子族的索引变为len(centList),然后存入centList bestClustAss[np.nonzero(bestClustAss[:,0]==1)[0],0]=len(centList) bestClustAss[np.nonzero(bestClustAss[:,0]==0)[0],0]=bestCentToSplit centList[bestCentToSplit] = bestNewCents[0,:].tolist() centList.append(bestNewCents[1,:].tolist()) self._clusterAssment[np.nonzero(self._clusterAssment[:,0] == \ bestCentToSplit)[0],:]= bestClustAss self._labels = self._clusterAssment[:,0] self._sse = sum(self._clusterAssment[:,1]) self._centroids = np.asarray(centList) def predict(self, X):#根据聚类结果,预测新输入数据所属的族 #类型检查 if not isinstance(X,np.ndarray): try: X = np.asarray(X) except: raise TypeError("numpy.ndarray required for X") m = X.shape[0]#m代表样本数量 preds = np.empty((m,)) for i in range(m):#将每个样本点分配到离它最近的质心所属的族 minDist = np.inf for j in range(self._k): distJI = self._calEDist(self._centroids[j,:],X[i,:]) if distJI <</span> minDist: minDist = distJI preds[i] = j return preds
Python股票历史涨跌幅数据获取
股票涨跌幅数据是量化投资学习的基本数据资料之一,下面以Python代码编程为工具,获得所需要的历史数据。主要步骤有:
(1) #按照市值从小到大的顺序活得N支股票的代码;
(2) #分别对这一百只股票进行100支股票操作;
(3) #获取从2016.05.01到2016.11.17的涨跌幅数据;
(4) #选取记录大于40个的数据,去除次新股;
(5) #将文件名名为“股票代码.csv”。
具体代码如下:
# -*- coding: utf-8 -*- """ Created on Thu Nov 17 23:04:33 2016 获取股票的历史涨跌幅,并分别存为csv格式 @author: yehx """ import numpy as np import pandas as pd #按照市值从小到大的顺序活得100支股票的代码 df = get_fundamentals( query(fundamentals.eod_derivative_indicator.market_cap) .order_by(fundamentals.eod_derivative_indicator.market_cap.asc()) .limit(100),\'2016-11-17\', \'1y\' ) #分别对这一百只股票进行100支股票操作 #获取从2016.05.01到2016.11.17的涨跌幅数据 #选取记录大于40个的数据,去除次新股 #将文件名名为“股票代码.csv” for stock in range(100): priceChangeRate = get_price_change_rate(df[\'market_cap\'].columns[stock], \'20160501\', \'20161117\') if priceChangeRate is None: openDays = 0 else: openDays = len(priceChangeRate) if openDays > 40: tempPrice = priceChangeRate[39:(openDays - 1)] for rate in range(len(tempPrice)): tempPrice[rate] = "%.3f" %tempPrice[rate] fileName = \'\' fileName = fileName.join(df[\'market_cap\'].columns[i].split(\'.\')) + \'.csv\' fileName tempPrice.to_csv(fileName)
Python Logistic 回归分类
Logistic回归可以认为是线性回归的延伸,其作用是对二分类样本进行训练,从而对达到预测新样本分类的目的。
假设有一组已知分类的MxN维样本X,M为样本数,N为特征维度,其相应的已知分类标签为Mx1维矩阵Y。那么Logistic回归的实现思路如下:
(1)用一组权重值W(Nx1)对X的特征进行线性变换,得到变换后的样本X’(Mx1),其目标是使属于不同分类的样本X’存在一个明显的一维边界。
(2)然后再对样本X’进一步做函数变换,从而使处于一维边界两测的值变换到相应的范围之内。
(3)训练过程就是通过改变W尽可能使得到的值位于一维边界两侧,并且与已知分类相符。
(4)对于Logistic回归,就是将原样本的边界变换到x=0这个边界。
下面是Logistic回归的典型代码:
# -*- coding: utf-8 -*- """ Created on Wed Nov 09 15:21:48 2016 Logistic回归分类 """ import numpy as np class LogisticRegressionClassifier(): def __init__(self): self._alpha = None #定义一个sigmoid函数 def _sigmoid(self, fx): return 1.0/(1 + np.exp(-fx)) #alpha为步长(学习率);maxCycles最大迭代次数 def _gradDescent(self, featData, labelData, alpha, maxCycles): dataMat = np.mat(featData) #size: m*n labelMat = np.mat(labelData).transpose() #size: m*1 m, n = np.shape(dataMat) weigh = np.ones((n, 1)) for i in range(maxCycles): hx = self._sigmoid(dataMat * weigh) error = labelMat - hx #size:m*1 weigh = weigh + alpha * dataMat.transpose() * error#根据误差修改回归系数 return weigh #使用梯度下降方法训练模型,如果使用其它的寻参方法,此处可以做相应修改 def fit(self, train_x, train_y, alpha=0.01, maxCycles=100): return self._gradDescent(train_x, train_y, alpha, maxCycles) #使用学习得到的参数进行分类 def predict(self, test_X, test_y, weigh): dataMat = np.mat(test_X) labelMat = np.mat(test_y).transpose() #使用transpose()转置 hx = self._sigmoid(dataMat*weigh) #size:m*1 m = len(hx) error = 0.0 for i in range(m): if int(hx[i]) > 0.5: print str(i+1)+\'-th sample \', int(labelMat[i]), \'is classfied as: 1\' if int(labelMat[i]) != 1: error += 1.0 print "classify error." else: print str(i+1)+\'-th sample \', int(labelMat[i]), \'is classfied as: 0\' if int(labelMat[i]) != 0: error += 1.0 print "classify error." error_rate = error/m print "error rate is:", "%.4f" %error_rate return error_rate
Python 朴素贝叶斯(Naive Bayes)分类
Naïve Bayes 分类的核心是计算条件概率P(y|x),其中y为类别,x为特征向量。其意义是在x样本出现时,它被划分为y类的可能性(概率)。通过计算不同分类下的概率,进而把样本划分到概率最大的一类。
根据条件概率的计算公式可以得到:
P(y|x) = P(y)*P(x|y)/P(x)。
由于在计算不同分类概率是等式右边的分母是相同的,所以只需比较分子的大小。并且,如果各个样本特征是独立分布的,那么p(x
|y)等于p(xi|y)相乘。
下面以文本分类来介绍Naïve Bayes分类的应用。其思路如下:
(1)建立词库,即无重复的单词表。
(2)分别计算词库中类别标签出现的概率P(y)。
(3)分别计算各个类别标签下不同单词出现的概率P(xi|y)。
(4)在不同类别下,将待分类样本各个特征出现概率((xi|y)相乘,然后在乘以对应的P(y)。
(5)比较不同类别下(4)中结果,将待分类样本分到取值最大的类别。
下面是Naïve Bayes 文本分类的Python代码,其中为了方便计算,程序中借助log对数函数将乘法转化为了加法。
# -*- coding: utf-8 -*- """ Created on Mon Nov 14 11:15:47 2016 Naive Bayes Clssification """ # -*- coding: utf-8 -*- import numpy as np class NaiveBayes: def __init__(self): self._creteria = "NB" def _createVocabList(self, dataList): """ 创建一个词库向量 """ vocabSet = set([]) for line in dataList: print set(line) vocabSet = vocabSet | set(line) return list(vocabSet) #文档词集模型 def _setOfWords2Vec(self, vocabList, inputSet): """ 功能:根据给定的一行词,将每个词映射到此库向量中,出现则标记为1,不出现则为0 """ outputVec = [0] * len(vocabList) for word in inputSet: if word in vocabList: outputVec[vocabList.index(word)] = 1 else: print "the word:%s is not in my vocabulary!" % word return outputVec # 修改 _setOfWordsVec 文档词袋模型 def _bagOfWords2VecMN(self, vocabList, inputSet): """ 功能:对每行词使用第二种统计策略,统计单个词的个数,然后映射到此库中 输出:一个n维向量,n为词库的长度,每个取值为单词出现的次数 """ returnVec = [0]*len(vocabList) for word in inputSet: if word in vocabList: returnVec[vocabList.index(word)] += 1 # 更新此处代码 return returnVec def _trainNB(self, trainMatrix, trainLabel): """ 输入:训练矩阵和类别标签,格式为numpy矩阵格式 功能:计算条件概率和类标签概率 """ numTrainDocs = len(trainMatrix) #统计样本个数 numWords = len(trainMatrix[0]) #统计特征个数,理论上是词库的长度 pNeg = sum(trainLabel)/float(numTrainDocs) #计算负样本出现的概率 p0Num = np.ones(numWords) #初始样本个数为1,防止条件概率为0,影响结果 p1Num = np.ones(numWords) #作用同上 p0InAll = 2.0 #词库中只有两类,所以此处初始化为2(use laplace) p1InAll = 2.0 # 再单个文档和整个词库中更新正负样本数据 for i in range(numTrainDocs): if trainLabel[i] == 1: p1Num += trainMatrix[i] p1InAll += sum(trainMatrix[i]) else: p0Num += trainMatrix[i] p0InAll += sum(trainMatrix[i]) print p1InAll #计算给定类别的条件下,词汇表中单词出现的概率 #然后取log对数,解决条件概率乘积下溢 p0Vect = np.log(p0Num/p0InAll) #计算类标签为0时的其它属性发生的条件概率 p1Vect = np.log(p1Num/p1InAll) #log函数默认以e为底 #p(ci|w=0) return p0Vect, p1Vect, pNeg def _classifyNB(self, vecSample, p0Vec, p1Vec, pNeg): """ 使用朴素贝叶斯进行分类,返回结果为0/1 """ prob_y0 = sum(vecSample * p0Vec) + np.log(1-pNeg) prob_y1 = sum(vecSample * p1Vec) + np.log(pNeg) #log是以e为底 if prob_y0 <</span> prob_y1: return 1 else: return 0 # 测试NB算法 def testingNB(self, testSample): listOPosts, listClasses = loadDataSet() myVocabList = self._createVocabList(listOPosts) # print myVocabList trainMat=[] for postinDoc in listOPosts: trainMat.append(self._bagOfWords2VecMN(myVocabList, postinDoc)) p0V,p1V,pAb = self._trainNB(np.array(trainMat), np.array(listClasses)) print trainMat thisSample = np.array(self._bagOfWords2VecMN(myVocabList, testSample)) result = self._classifyNB(thisSample, p0V, p1V, pAb) print testSample,\'classified as: \', result return result ############################################################################### def loadDataSet(): wordsList=[[\'my\', \'dog\', \'has\', \'flea\', \'problems\', \'help\', \'please\'], [\'maybe\', \'not\', \'take\', \'him\', \'to\', \'dog\', \'park\', \'stupid\'], [\'my\', \'dalmation\', \'is\', \'so\', \'cute\', \' and\', \'I\', \'love\', \'him\'], [\'stop\', \'posting\', \'stupid\', \'worthless\', \'garbage\'], [\'mr\', \'licks\',\'ate\',\'my\', \'steak\', \'how\', \'to\', \'stop\', \'him\'], [\'quit\', \'buying\', \'worthless\', \'dog\', \'food\', \'stupid\']] classLable = [0,1,0,1,0,1] # 0:good; 1:bad return wordsList, classLable if __name__=="__main__": clf = NaiveBayes() testEntry = [[\'love\', \'my\', \'girl\', \'friend\'], [\'stupid\', \'garbage\'], [\'Haha\', \'I\', \'really\', "Love", "You"], [\'This\', \'is\', "my", "dog"]] clf.testingNB(testEntry[0]) # for item in testEntry: # clf.testingNB(item)
Python股票历史数据预处理(一)
在进行量化投资交易编程时,我们需要股票历史数据作为分析依据,下面介绍如何通过Python获取股票历史数据并且将结果存为DataFrame格式。处理后的股票历史数据下载链接为:http://download.****.net/detail/suiyingy/9688505。
具体步骤如下:
- (1) 建立股票池,这里按照股本大小来作为选择依据。
- (2) 分别读取股票池中所有股票的历史涨跌幅。
- (3) 将各支股票的历史涨跌幅存到DataFrame结构变量中,每一列代表一支股票,对于在指定时间内还没有发行的股票的涨跌幅设置为0。
- (4) 将DataFrame最后一行的数值设置为各支股票对应的交易天数。
- (5) 将DataFrame数据存到csv文件中去。
具体代码如下:
# -*- coding: utf-8 -*- """ Created on Thu Nov 17 23:04:33 2016 获取股票的历史涨跌幅,先合并为DataFrame后存为csv格式 @author: yehx """ import numpy as np import pandas as pd #按照市值从小到大的顺序获得50支股票的代码 df = get_fundamentals( query(fundamentals.eod_derivative_indicator.market_cap) .order_by(fundamentals.eod_derivative_indicator.market_cap.asc()) .limit(50),\'2016-11-17\', \'1y\' ) b1= {} priceChangeRate_300 = get_price_change_rate(\'000300.XSHG\', \'20060101\', \'20161118\') df300 = pd.DataFrame(priceChangeRate_300) lenReference = len(priceChangeRate_300) dfout = df300 dflen = pd.DataFrame() dflen[\'000300.XSHG\'] = [lenReference] #分别对这一百只股票进行50支股票操作 #获取从2006.01.01到2016.11.17的涨跌幅数据 #将数据存到DataFrame中 #DataFrame存为csv文件 for stock in range(50): priceChangeRate = get_price_change_rate(df[\'market_cap\'].columns[stock], \'20150101\', \'20161118\') if priceChangeRate is None: openDays = 0 else: openDays = len(priceChangeRate) dftempPrice = pd.DataFrame(priceChangeRate) tempArr = [] for i in range(lenReference): if df300.index[i] in list(dftempPrice.index): #保存为4位有效数字 tempArr.append( "%.4f" %((dftempPrice.loc[str(df300.index[i])][0]))) pass else: tempArr.append(float(0.0)) fileName = \'\' fileName = fileName.join(df[\'market_cap\'].columns[stock].split(\'.\')) dfout[fileName] = tempArr dflen[fileName] = [len(priceChangeRate)] dfout = dfout.append(dflen) dfout.to_csv(\'00050.csv\')
Python股票历史数据预处理(二)
从网上下载的股票历史数据往往不能直接使用,需要转换为自己所需要的格式。下面以Python代码编程为工具,将csv文件中存储的股票历史数据提取出来并处理。处理的数据结果为是30天涨跌幅子数据库,下载地址为:http://download.****.net/detail/suiyingy/9688605。
主要步骤有(Python csv数据读写):
- #csv文件读取股票历史涨跌幅数据;
- #随机选取30个历史涨跌幅数据;
- #构建自己的数据库;
- #将处理结果保存为新的csv文件。
具体代码如下:
# -*- coding: utf-8 -*- """ Created on Thu Nov 17 23:04:33 2016 csv格式股票历史涨跌幅数据处理 @author: yehx """ import numpy as np import pandas as pd import random import csv import sys reload(sys) sys.setdefaultencoding(\'utf-8\') \'\'\' - 加载csv格式数据 \'\'\' def loadCSVfile1(datafile): filelist = [] with open(datafile) as file: lines = csv.reader(file) for oneline in lines: filelist.append(oneline) filelist = np.array(filelist) return filelist #数据处理 #随机选取30个历史涨跌幅数据 #构建自己的数据库 def dataProcess(dataArr, subLen): totLen, totWid = np.shape(data) print totLen, totWid lenArr = dataArr[totLen-1,2:totWid] columnCnt = 1 dataOut = [] for lenData in lenArr: columnCnt = columnCnt + 1 N60 = int(lenData) / (2 * subLen) print N60 if N60 > 0: randIndex = random.sample(range(totLen-int(lenData)-1,totLen-subLen), N60) for i in randIndex: dataOut.append(dataArr[i:(i+subLen),columnCnt]) dataOut = np.array(dataOut) return dataOut if __name__=="__main__": datafile = "00100 (3).csv" data = loadCSVfile1(datafile) df = pd.DataFrame(data) m, n = np.shape(data) dataOut = dataProcess(data, 30) m, n = np.shape(dataOut) #保存处理结果 csvfile = file(\'csvtest.csv\', \'wb\') writer = csv.writer(csvfile) writer.writerows(dataOut) csvfile.close()
http://blog.sina.com.cn/s/articlelist_6017673753_0_1.html
https://www.cnblogs.com/ttrrpp/