量化交易——传统技术分析布林通道BollingerBands的原理及实现

布林通道

布林通道线是根据统计学的标准差来计算的，其具体可由上中下三条曲线展示。其中上下两线分别代表上升压力线和下降支撑线，故而可以根据K线图是否突破布林曲线来判断较好的买卖节点。三条曲线计算方法如下：

中轨线(MID)=收盘价的M日移动平均线;
上轨线(UPER)=中轨线+N倍的标准差;
下轨线(LOWER)=中轨线-N倍的标准差.、

实现

某些分析过程可以参考前面的博文，量化交易——传统技术分析相对强弱指数RSI的原理及实现，这里不细讲。

import numpy as npimport mathimport randomimport jsonimport matplotlib.pyplot as pltimport syssys.setrecursionlimit(10000)#date|open|high|low|close|volume|adjsuted def get_stock_hist(num):    s_his=np.genfromtxt('C:/Users/Haipeng/Desktop/python/Korea/Korea_{:03d}.csv'.format(num), delimiter=',')    s_hi=s_his[1:][:]    days=s_hi.shape[0]    this_stock = []    for i in range(1,days,1):        this_day = [i]        for k in range(1,7):            this_day.append(s_hi[i][k])        this_stock.append(this_day)    print 'Maximum date is ',len(this_stock)    return this_stockdef get_price(D, p_tpe):    if p_tpe=='close':        pos=4;    elif p_tpe=='open':        pos=1;    elif p_tpe=='high':        pos=2;    elif p_tpe=='low':        pos=3;    else:        pos=5    price=stock_hist[D-1][pos];    return pricedef get_ma(D, N):    p_used=np.zeros(N);    for i in range(1,N+1,1):        p_used[i-1]=stock_hist[(D-1)-(i-1)][4];    ma=np.mean(p_used);    return madef get_mar(fro,to,N):    ma = []    for i in range(fro,to+1):        ma.append(get_ma(i,N))    return ma#Date\Open\High\Low\Closedef get_tuples(fro,to):    res =[]    for d in range(fro,to+1):        tmp = []        tmp.append(d)        tmp.append(get_price(d,'open'))        tmp.append(get_price(d,'high'))        tmp.append(get_price(d,'low'))        tmp.append(get_price(d,'close'))                res.append(tmp)    return resdef get_volume(fro,to):    res = []    for d in range(fro,to+1):        num = 1        try:            if get_price(d,'close')<get_price(d-1,'close'):                num = -1        except:            pass        res.append(num*get_price(d,'volume'))    return res  # BB Band实现def get_BB(D,N):    MD = 0.0    UP = 0.0    DN = 0.0    SD = 0.0    for i in range(N):        MD += get_price(D-i,'close')    MD = MD/N    for i in range(N):        SD += math.pow(MD-get_price(D-i,'close'),2)    SD = math.sqrt(SD/N)    UP = MD + 2*SD    DN = MD - 2*SD    return [UP,MD,DN]def get_bb(fro,to,N):    res = [[],[],[]]    for d in range(fro,to+1):        if d<N:            print 'Date number is not larger than N!'            break        tmp = get_BB(d,N)         res[0].append(tmp[0])        res[1].append(tmp[1])        res[2].append(tmp[2])    return res

绘制k线图及BB指标

绘图代码：

def plot_BB(fro,to):    volume = get_volume(fro,to)    tmp = get_bb(fro,to,20)    up = tmp[0]    md = tmp[1]    dn = tmp[2]    tuples = get_tuples(fro,to)    date = [d for d in range(fro,to+1)]     fig = plt.figure(figsize=(18,10))    p1 = plt.subplot2grid((4,4),(0,0),rowspan=3,colspan=4,axisbg='k')     p1.set_title("Bollinger Bands:(20-Day Moving Average)")    p1.set_ylabel("Price")    p1.plot(date,up,'m')    p1.plot(date,md,'b')    p1.plot(date,dn,'y')    p1.legend(('UP','MD','DN'))    p1.grid(True,color='w')    candlestick_ohlc(p1, tuples, width=0.7,colorup='r',colordown="g")    p2 = plt.subplot2grid((4,4),(3,0),colspan=4,axisbg='c')     p2.set_ylabel("Volume")    colors = []    for i in range(len(volume)):        if volume[i]<0:            colors.append('green')            volume[i] = -volume[i]        else:            colors.append('red')    p2.bar(date,volume,color=colors)    p2.set_xlabel("Dates")    plt.subplots_adjust(hspace=0)    plt.show()# show the plot on the screenstock_hist = get_stock_hist(18)plot_BB(1250,1520)

图示结果1：

缩短时间范围：

#得图2plot_BB(220,320)

我们发现，上图有较大的不合理的跳跃。这是什么原因呢？仔细观察发现图像是在第278天发生了突变，应该就是数据出了问题。现在打开对应股票的CSV数据如下，果然：

可知，当天数据丢失了一位数，这才导致了以上的突变。所以在接下来深入挖掘数据时要考虑减少数据错误带来的影响。