Development environment: python v3.8.2, matplotlib v3.2.1, VSCode
I have started to learn quantification for 2 or 3 weeks, and I am still a novice to python. I have only started to look at the usage of matplotlib in the past 2 days. I searched the Internet for drawing some K-line graphs. Most of them use the K-line graphs drawn by mpl_finance. The effect is average. Yesterday’s whim , While learning matplotlib, while imitating the effect of a straight flush, the mouse interaction has not been done yet (the next time I have time to do it), I am afraid that the code is too long. Originally using the new version of mpl_finance, it does support the K-line chart better, but the code limit is relatively dead, and it is not easy to add sub-graphs later, so some codes are directly applied from it, and the results are posted first. The code is marked in detail, so I won't explain it, just look at the code.
# -*- coding:utf-8 -*-
import tushare as ts # 获取股票数据用
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import matplotlib.ticker as ticker # 用于日期刻度定制
# import mpl_finance as mpf
# 注释了上面一行的旧版，新版是mplfinance，少了一个下划线，运行没警告，新版绘制蜡烛图性能应该更高，但是限得比较死，
# 不好加子图，后面我们把关键代码拿出来自己绘制
from mplfinance import original_flavor as mpf # 用新版库里面的旧版实现(跟mpl_finance一模一样)
from matplotlib import colors as mcolors # 用于颜色转换成渲染时顶点需要的颜色格式
from matplotlib.collections import LineCollection, PolyCollection # 用于绘制直线集合和多边形集合
import tools # 计算MACD和KDJ(没有使用talib，有缺陷)
# 获取数据
df = ts.get_k_data('603986', '2019-01-01') # 获取日K线数据
df.drop('code', axis=1, inplace=True) # 剔除多余的列
tools.calc_macd(df) # 计算MACD值，数据存于DataFrame中
tools.calc_kdj(df) # 计算KDJ值，数据存于DataFrame中
df = df[-150:] # 取最近的N天测试，列顺序为 date,open,close,high,low,volume,dif,dea,bar,k,d,j
# 日期转换成整数序列
date_tickers=df.date.values
df.date = range(0, len(df)) # 日期改变成序号
matix = df.values # 转换成绘制蜡烛图需要的数据格式(date, open, close, high, low, volume)
xdates = matix[:, 0] # X轴数据(这里用的天数索引)
# 设置外观效果
plt.rc('font', family='Microsoft YaHei') # 用中文字体，防止中文显示不出来
plt.rc('figure', fc='k') # 绘图对象背景图
plt.rc('text', c='#800000') # 文本颜色
plt.rc('axes', axisbelow=True, xmargin=0, fc='k', ec='#800000', lw=1.5, labelcolor='#800000', unicode_minus=False) # 坐标轴属性(置底，左边无空隙，背景色，边框色，线宽，文本颜色，中文负号修正)
plt.rc('xtick', c='#d43221') # x轴刻度文字颜色
plt.rc('ytick', c='#d43221') # y轴刻度文字颜色
plt.rc('grid', c='#800000', alpha=0.9, ls=':', lw=0.8) # 网格属性(颜色，透明值，线条样式，线宽)
plt.rc('lines', lw=0.8) # 全局线宽
# 创建绘图对象和4个坐标轴
fig = plt.figure(figsize=(16, 8))
left, width = 0.05, 0.9
ax1 = fig.add_axes([left, 0.6, width, 0.35]) # left, bottom, width, height
ax2 = fig.add_axes([left, 0.45, width, 0.15], sharex=ax1) # 共享ax1轴
ax3 = fig.add_axes([left, 0.25, width, 0.2], sharex=ax1) # 共享ax1轴
ax4 = fig.add_axes([left, 0.05, width, 0.2], sharex=ax1) # 共享ax1轴
plt.setp(ax1.get_xticklabels(), visible=False) # 使x轴刻度文本不可见，因为共享，不需要显示
plt.setp(ax2.get_xticklabels(), visible=False) # 使x轴刻度文本不可见，因为共享，不需要显示
plt.setp(ax3.get_xticklabels(), visible=False) # 使x轴刻度文本不可见，因为共享，不需要显示
# 绘制蜡烛图
def format_date(x, pos=None): return '' if x<0 or x>len(date_tickers)-1 else date_tickers[int(x)] # 日期格式化函数，根据天数索引取出日期值
ax1.xaxis.set_major_formatter(ticker.FuncFormatter(format_date)) # 设置自定义x轴格式化日期函数
ax1.xaxis.set_major_locator(ticker.MultipleLocator(max(int(len(df)/15), 5))) # 横向最多排15个左右的日期，最少5个，防止日期太拥挤
# mpf.candlestick_ochl(ax1, matix, width=0.5, colorup='#ff3232', colordown='#54fcfc')
# # 下面这一段代码，替换了上面注释的这个函数，因为上面的这个函数达不到同花顺的效果
opens, closes, highs, lows = matix[:, 1], matix[:, 2], matix[:, 3], matix[:, 4] # 取出ochl值
avg_dist_between_points = (xdates[-1] - xdates[0]) / float(len(xdates)) # 计算每个日期之间的距离
delta = avg_dist_between_points / 4.0 # 用于K线实体(矩形)的偏移坐标计算
barVerts = [((date - delta, open), (date - delta, close), (date + delta, close), (date + delta, open)) for date, open, close in zip(xdates, opens, closes) ] # 生成K线实体(矩形)的4个顶点坐标
rangeSegLow = [ ((date, low), (date, min(open, close))) for date, low, open, close in zip(xdates, lows, opens, closes) ] # 生成下影线顶点列表
rangeSegHigh = [ ((date, high), (date, max(open, close))) for date, high, open, close in zip(xdates, highs, opens, closes) ] # 生成上影线顶点列表
rangeSegments = rangeSegLow + rangeSegHigh # 上下影线顶点列表
cmap = {True: mcolors.to_rgba('#000000', 1.0), False: mcolors.to_rgba('#54fcfc', 1.0)} # K线实体(矩形)中间的背景色(True是上涨颜色，False是下跌颜色)
inner_colors = [ cmap[opn < cls] for opn, cls in zip(opens, closes) ] # K线实体(矩形)中间的背景色列表
cmap = {True: mcolors.to_rgba('#ff3232', 1.0), False: mcolors.to_rgba('#54fcfc', 1.0)} # K线实体(矩形)边框线颜色(上下影线和后面的成交量颜色也共用)
updown_colors = [ cmap[opn < cls] for opn, cls in zip(opens, closes) ] # K线实体(矩形)边框线颜色(上下影线和后面的成交量颜色也共用)列表
ax1.add_collection(LineCollection(rangeSegments, colors=updown_colors, linewidths=0.5, antialiaseds=False)) # 生成上下影线的顶点数据(颜色，线宽，反锯齿，反锯齿关闭好像没效果)
ax1.add_collection(PolyCollection(barVerts, facecolors=inner_colors, edgecolors=updown_colors, antialiaseds=False, linewidths=0.5)) # 生成多边形(矩形)顶点数据(背景填充色，边框色，反锯齿，线宽)
# 绘制均线
mav_colors = ['#ffffff', '#d4ff07', '#ff80ff', '#00e600', '#02e2f4', '#ffffb9', '#2a6848'] # 均线循环颜色
mav_period = [5, 10, 20, 30, 60, 120, 180] # 定义要绘制的均线周期，可增减
n = len(df)
for i in range(len(mav_period)):
if n >= mav_period[i]:
mav_vals = df['close'].rolling(mav_period[i]).mean().values
ax1.plot(xdates, mav_vals, c=mav_colors[i%len(mav_colors)], label='MA'+str(mav_period[i]))
ax1.set_title('K线图') # 标题
ax1.grid(True) # 画网格
ax1.legend(loc='upper right') # 图例放置于右上角
ax1.xaxis_date() # 好像要不要效果一样？
# 绘制成交量和成交量均线（5日，10日）
# ax2.bar(xdates, matix[:, 5], width= 0.5, color=updown_colors) # 绘制成交量柱状图
barVerts = [((date - delta, 0), (date - delta, vol), (date + delta, vol), (date + delta, 0)) for date, vol in zip(xdates, matix[:,5]) ] # 生成K线实体(矩形)的4个顶点坐标
ax2.add_collection(PolyCollection(barVerts, facecolors=inner_colors, edgecolors=updown_colors, antialiaseds=False, linewidths=0.5)) # 生成多边形(矩形)顶点数据(背景填充色，边框色，反锯齿，线宽)
if n>=5: # 5日均线，作法类似前面的均线
vol5 = df['volume'].rolling(5).mean().values
ax2.plot(xdates, vol5, c='y', label='VOL5')
if n>=10: # 10日均线，作法类似前面的均线
vol10 = df['volume'].rolling(10).mean().values
ax2.plot(xdates, vol10, c='w', label='VOL10')
ax2.yaxis.set_ticks_position('right') # y轴显示在右边
ax2.legend(loc='upper right') # 图例放置于右上角
ax2.grid(True) # 画网格
# ax2.set_ylabel('成交量') # y轴名称
# 绘制MACD
difs, deas, bars = matix[:, 6], matix[:, 7], matix[:, 8] # 取出MACD值
ax3.axhline(0, ls='-', c='g', lw=0.5) # 水平线
ax3.plot(xdates, difs, c='w', label='DIFF') # 绘制DIFF线
ax3.plot(xdates, deas, c='y', label='DEA') # 绘制DEA线
# ax3.bar(xdates, df['bar'], width= 0.05, color=bar_colors) # 绘制成交量柱状图(发现用bar绘制，线的粗细不一致，故使用下面的直线列表)
cmap = {True: mcolors.to_rgba('r', 1.0), False: mcolors.to_rgba('g', 1.0)} # MACD线颜色，大于0为红色，小于0为绿色
bar_colors = [ cmap[bar > 0] for bar in bars ] # MACD线颜色列表
vlines = [ ((date, 0), (date, bars[date])) for date in range(len(bars)) ] # 生成MACD线顶点列表
ax3.add_collection(LineCollection(vlines, colors=bar_colors, linewidths=0.5, antialiaseds=False)) # 生成MACD线的顶点数据(颜色，线宽，反锯齿)
ax3.legend(loc='upper right') # 图例放置于右上角
ax3.grid(True) # 画网格
# 绘制KDJ
K, D, J = matix[:, 9], matix[:, 10], matix[:, 11] # 取出KDJ值
ax4.axhline(0, ls='-', c='g', lw=0.5) # 水平线
ax4.yaxis.set_ticks_position('right') # y轴显示在右边
ax4.plot(xdates, K, c='y', label='K') # 绘制K线
ax4.plot(xdates, D, c='c', label='D') # 绘制D线
ax4.plot(xdates, J, c='m', label='J') # 绘制J线
ax4.legend(loc='upper right') # 图例放置于右上角
ax4.grid(True) # 画网格
# set useblit = True on gtkagg for enhanced performance
from matplotlib.widgets import Cursor # 处理鼠标
cursor = Cursor(ax1, useblit=True, color='w', linewidth=0.5, linestyle='--')
plt.show()
代码中用到了tools库，自己的，我把那2个计算函数贴出来，你们直接替换即可
# 在k线基础上计算MACD，并将结果存储在df上面(dif,dea,bar)
def calc_macd(df, fastperiod=12, slowperiod=26, signalperiod=9):
ewma12 = df['close'].ewm(span=fastperiod,adjust=False).mean()
ewma26 = df['close'].ewm(span=slowperiod,adjust=False).mean()
df['dif'] = ewma12-ewma26
df['dea'] = df['dif'].ewm(span=signalperiod,adjust=False).mean()
df['bar'] = (df['dif']-df['dea'])*2
# df['macd'] = 0
# series = df['dif']>0
# df.loc[series[series == True].index, 'macd'] = 1
return df
# 在k线基础上计算KDF，并将结果存储在df上面(k,d,j)
def calc_kdj(df):
low_list = df['low'].rolling(9, min_periods=9).min()
low_list.fillna(value=df['low'].expanding().min(), inplace=True)
high_list = df['high'].rolling(9, min_periods=9).max()
high_list.fillna(value=df['high'].expanding().max(), inplace=True)
rsv = (df['close'] - low_list) / (high_list - low_list) * 100
df['k'] = pd.DataFrame(rsv).ewm(com=2).mean()
df['d'] = df['k'].ewm(com=2).mean()
df['j'] = 3 * df['k'] - 2 * df['d']
# df['kdj'] = 0
# series = df['k']>df['d']
# df.loc[series[series == True].index, 'kdj'] = 1
# # df.loc[series[(series == True) & (series.shift() == False)].index, 'kdjcross'] = 1
# # df.loc[series[(series == False) & (series.shift() == True)].index, 'kdjcross'] = -1
return df
-
-
-
m0_47901052 : Excuse me , the calculated value of kdj is very close to that of Tongdaxin software. But the three values of macd are quite different from the calculated value of Tongdaxin. What is the reason? Reply 6 months ago
1 -
peakgao
replied: Because the data obtained by tushare is ex-rights data, not pre-restoration, you should change the Tongda letter to ex-rights and look again, it should be the same 3 months ago.-
m0_47901052 : Thank you very much the author! The problem has been solved! Just remove the import tools and add the following two lines of code: calc_macd(df) calc_kdj(df) Reply 6 months ago
-
-
peakgao
reply qxh66: Just copy the calc_macd, calc_kdj code directly, and then remove the tools module, it will be fine. Reply 3 months ago
-
-