tick数据生成K线数据 计算RSI
数据来源大富翁期货数据，文件存放示例
F:\share\zc\20180108\ZC805_20180108.csv F:\share\zc\20180109\ZC805_20180109.csv
以下 代码生成1分钟、5分钟、15分钟k线数据，包括RSI
数据示例：
大富翁数据中心_市场代码,合约代码,时间,最新,持仓,增仓,成交额,成交量,开仓,平仓,成交类型,方向,买一价,卖一价,买一量,卖一量
zc,ZC805,2018-01-02 08:59:00.000,605.0000,235386,10,165770.0000,274.0000,142,132,空开,S,605.0000,605.4000,28,33zc,ZC805,2018-01-02 09:00:00.000,604.2000,235470,84,73810.0000,122.0000,103,19,空开,S,604.2000,605.0000,1,4
zc,ZC805,2018-01-02 09:00:01.000,604.0000,235478,8,44676.0000,74.0000,41,33,空开,S,604.0000,604.4000,2,2
zc,ZC805,2018-01-02 09:00:01.000,603.8000,235508,30,56738.4000,94.0000,62,32,空开,S,603.6000,603.8000,33,10
#coding: utf-8
#!/usr/bin/env pythonimport numpy as np
import pandas as pd
import os
def rsiFunc(prices, n=14):
deltas = np.diff(prices)
seed = deltas[:n+1]
up = (seed[seed>=0].sum())/n+0.000001
down = (-seed[seed<0].sum())/n+0.000001
rs = up/down
rsi = np.zeros_like(prices)
rsi[:n] = 100. - 100./(1.+rs)
for i in range(n, len(prices)):
delta = deltas[i-1] # cause the diff is 1 shorter
if delta>0:
upval = delta
downval = 0.
else:
upval = 0.
downval = -delta
up = (up*(n-1) + upval)/n
down = (down*(n-1) + downval)/n
rs = up/down
rsi[i] = 100. - 100./(1.+rs)
return rsi
def relative_strength(prices, n=14):
deltas = np.diff(prices)
seed = deltas[:n+1]
up = (seed[seed>=0].sum())/n+0.000001
down = (-seed[seed<0].sum())/n+0.000001
rs = up/down
rsi = np.zeros_like(prices)
rsi[:n] = 100. - 100./(1.+rs)
for i in range(n, len(prices)):
delta = deltas[i-1]
if delta>0:
upval = delta
downval = 0.
else:
upval = 0.
downval = -delta
up = (up*(n-1) + upval)/n
down = (down*(n-1) + downval)/n
rs = up/down
rsi[i] = 100. - 100./(1.+rs)
return rsi
def rsi2(price, n=14):
''' rsi indicator '''
gain = (price-price.shift(1)).fillna(0) # calculate price gain with previous day, first row nan is filled with 0
def rsiCalc(p):
# subfunction for calculating rsi for one lookback period
avgGain = p[p>0].sum()/n
avgLoss = -p[p<0].sum()/n
rs = avgGain/avgLoss
return 100 - 100/(1+rs)
# run for all periods with rolling_apply
return pd.rolling_apply(gain,n,rsiCalc)
def kline(infile, period,outfile):
df = pd.read_csv(infile,skiprows=[0],encoding='GBK')
# names=['Symbol', 'Date_Time', 'Price','Volume']
abf=df.iloc[:, [1,2,3,7]]
abf.columns=[ 'Symbol','Date_Time', 'Price','Volume']
abf=abf[abf['Volume']>0]
abf.set_index(['Date_Time'], inplace=True)
abf.index = pd.to_datetime(abf.index)
bars = abf.Price.resample( period).ohlc()
Volumes =abf.Volume.resample(period).sum()
ohlcv = pd.concat([bars, Volumes], axis=1)
ohlcv=ohlcv[ohlcv['Volume']>0].dropna()
ohlcv['Symbol']=abf['Symbol'][1]
return ohlcv
def batch_gen_kline( period,outfile):
basedir="F:\\share\\zc\\"
prefix="ZC805_"
rng = pd.date_range('20180102', freq='D', periods=30).strftime('%Y%m%d')
no=0
for row in np.nditer(rng):
filename=basedir+str(row)+"\\"+prefix+ str(row)+".csv"
if(os.path.isfile(filename)):
print(filename)
data=kline(filename, period,outfile)
no=no+1
if( no==1):
data.to_csv(outfile, mode='a', index_label='Date_Time' ) #header=False,
else:
data.to_csv(outfile, mode='a', header=False, index_label='Date_Time' )
else:
print(" %s is not exisits skip it!" %filename)
period='1T'
outfile= 'ZC805_201801_' + period +'K.csv'
batch_gen_kline( period,outfile)
outfile= 'ZC805_201801_' + period +'K.csv'
batch_gen_kline( period,outfile)
period='10T'
outfile= 'ZC805_201801_' + period +'K.csv'
batch_gen_kline( period,outfile)
period='15T'
outfile= 'ZC805_201801_' + period +'K.csv'
batch_gen_kline( period,outfile)
#rsi= rsiFunc(ohlcv['close'],6)
#rsi= rsi2(ohlcv['close'],6)
#ohlcv['rsi'] = pd.Series(rsi, index=ohlcv.index)