def analysis(k):
# 转DataFrame格式
df = pd.DataFrame(
k, columns=['timestamp', 'open', 'high', 'low', 'close', 'volume'])
t = k[0][0]
# 毫秒级
if len(str(t)) == 13:
df['timestamp'] = df['timestamp'] / 1000
# 分析
stock = StockDataFrame.retype(df)
stock['cci']
df['date'] = df['timestamp'].apply(
lambda x: time.strftime("%Y-%m-%d %H:%M:%S", time.localtime(x)))
df['date'] = pd.to_datetime(df['date'])
df = df[5:]
df['regime'] = np.where((df['cci'] >= 100) & (df['cci'].shift(1) < 100), 1,
0)
df['regime'] = np.where(
((df['cci'] <= 100) & (df['cci'].shift(1) > 100)) |
((df['cci'] >= 100) & (df['stoch_d'] >= df['stoch_k'])), -1,
df['regime'])
# df['regime'] = np.where((df['cci'] <= -100) & (df['cci'].shift(1) >= -100), -1, df['regime'])
# df['regime'] = np.where(((df['cci'] >= -100) & (df['cci'].shift(1) <= -100)) | ((df['cci'] <= -100) & (df['stoch_k'] >= df['stoch_d'])), 1, df['regime'])
# 下单(只下多单)
print(df[['date', 'close', 'cci', 'regime']])
if (df[-1:]['regime'] == 1).bool():
return 'long'
elif (df[-1:]['regime'] == -1).bool():
return 'close_long'
else:
return 'wait'
def cal_fitness(df, pops, pop_size=50, chromosome_length=6):
obj_value = []
async_funcs = []
for i in range(pop_size):
df_ = copy.deepcopy(df)
stock = StockDataFrame.retype(df_)
df_['signal'] = 'wait'
boll_A_ = binary2decimal(pops[0][i], 1.5, 3, chromosome_length)
boll_std_len_ = int(binary2decimal(pops[1][i], 14, 30, chromosome_length))
macd_fast_len_ = int(binary2decimal(pops[2][i], 8, 16, chromosome_length))
macd_slow_len_ = int(binary2decimal(pops[3][i], 20, 30, chromosome_length))
boll_width_threshold_ = binary2decimal(pops[4][i], 0.05, 0.15, chromosome_length)
volatility_ = binary2decimal(pops[5][i], 0.0009, 0.0011, chromosome_length)
stop_limit_ = binary2decimal(pops[6][i], 0.01, 0.1, chromosome_length)
trend_A1_ = binary2decimal(pops[7][i], 0.2, 0.35, chromosome_length)
trend_A2_ = binary2decimal(pops[8][i], 0.01, 0.025, chromosome_length)
trend_B_ = binary2decimal(pops[9][i], 0.02, 0.045, chromosome_length)
trend_C_ = binary2decimal(pops[10][i], 0.03, 0.055, chromosome_length)
stop_trade_times_ = int(binary2decimal(pops[11][i], 6, 12, chromosome_length))
async_funcs.append(cal_someone_fitness(df_, stock, i, boll_A_, boll_std_len_, macd_fast_len_,
macd_slow_len_,
boll_width_threshold_, volatility_, stop_limit_, trend_A1_,
trend_A2_,
trend_B_, trend_C_, stop_trade_times_))
loop = asyncio.get_event_loop()
re = loop.run_until_complete(asyncio.gather(*async_funcs))
n = len(re)
for i in range(n - 1, 0, -1):
for j in range(i):
if re[j][0] > re[j + 1][0]:
re[j], re[j + 1] = re[j + 1], re[j]
for i in range(n):
obj_value.append(re[i][1])
return obj_value
def analysis_(datas):
df = pd.DataFrame(
datas, columns=['timestamp', 'open', 'high', 'low', 'close', 'volume'])
df = df.astype(float)
stock = StockDataFrame.retype(df)
df['date'] = pd.to_datetime(df['timestamp'], unit='s')
df.index = df.date
stock = StockDataFrame.retype(df)
stock['cci']
stock['stoch_rsi']
# 去掉前几个cci指标不准数据
df = df[10:]
df['signal'] = 'wait'
df['signal'] = np.where((df['cci'] >= 100) & (df['cci'].shift(1) < 100),
'long', df['signal'])
df['signal'] = np.where(((df['cci'] <= 100) & (df['cci'].shift(1) > 100)),
'close_long', df['signal'])
df['signal'] = np.where((df['cci'] <= -100) & (df['cci'].shift(1) >= -100),
'short', df['signal'])
df['signal'] = np.where((df['cci'] >= -100) & (df['cci'].shift(1) <= -100),
'close_short', df['signal'])
return df.iloc[-1]
import pandas as pd
import numpy as np
from tools.stockstats import StockDataFrame
import time
import matplotlib.ticker as ticker
from tools import data2df
import matplotlib.pyplot as plt
# 本金
principal = 10000.0
title = 'BTC2017-09-01-now-1D'
df = data2df.csv2df(title + '.csv')
df = df.astype(float)
df['Timestamp'] = df['Timestamp'].astype(int)
stock = StockDataFrame.retype(df)
df['date'] = df['timestamp'].apply(
lambda x: time.strftime("%Y-%m-%d %H:%M:%S", time.localtime(x)))
df['date'] = pd.to_datetime(df['date'])
stock['cci']
# 去掉前几个cci指标不准数据
df = df[5:]
df['regime'] = np.where((df['cci'] >= 100) & (df['cci'].shift(1) < 100), 1, 0)
df['regime'] = np.where((df['cci'] <= 100) & (df['cci'].shift(1) > 100), -1,
df['regime'])
df['regime'] = np.where((df['cci'] <= -100) & (df['cci'].shift(1) >= -100), -1,
df['regime'])
df['regime'] = np.where((df['cci'] >= -100) & (df['cci'].shift(1) <= -100), 1,
def analysis(kline):
df = pd.DataFrame(
kline, columns=['timestamp', 'open', 'high', 'low', 'close', 'volume'])
stock = StockDataFrame.retype(df)
# df['timestamp'] = df['timestamp'] / 1000
df['date'] = pd.to_datetime(df['timestamp'], unit='s')
df.index = df.date
stock['boll']
df['signal'] = 'wait'
# 参数设置
trend_A1 = 0.3
trend_A2 = 0.02
trend_B = 0.04
trend_C = 0.05
# trend_D = 0.07
stop_trade_times = 5
ts = 3600
std_percentage = 0.6
w = 200
volatility = 0.001
# 通道宽度
df['boll_width'] = abs(df['boll_ub'] - df['boll_lb'])
# 单次涨跌幅
df['change'] = abs(
(df['close'] - df['close'].shift(1)) / df['close'].shift(1))
# 两次涨幅
df['change_2'] = abs(
(df['close'] - df['close'].shift(2)) / df['close'].shift(2))
# 趋势判断A：超过通道一定幅度且单次涨幅超过一定幅度
df['signal'] = np.where(
((df['close'] > df['boll_ub'] + df['boll_width'] * trend_A1) |
(df['close'] < df['boll_lb'] - df['boll_width'] * trend_A1)) &
(df['change'] > trend_A2), 'trend', df['signal'])
# 趋势判断B：单次涨幅累计超过一定幅度
df['signal'] = np.where(df['change'] > trend_B, 'trend', df['signal'])
# 趋势判断C：两次涨幅累计超过一定幅度
df['signal'] = np.where(df['change_2'] > trend_C, 'trend', df['signal'])
for i in range(stop_trade_times):
df['signal'] = np.where(
(df['signal'] == 'wait') & (df['signal'].shift(1 + i) == 'trend'),
'can_not_trade', df['signal'])
df['signal'] = np.where(df['signal'] == 'can_not_trade', 'trend',
df['signal'])
df['close_mean'] = df['close'].rolling(center=False, window=w).mean()
df['close_std'] = df['close'].rolling(center=False, window=w).std()
df['close_std'] = df['close_std'] * std_percentage
df['rate'] = df['close_std'] / df['boll_width']
# df['close_std'] = np.where(df['rate'] > 0.1, df['boll_width'] * 0.1, df['close_std'])
# 策略1
df['signal'] = np.where(
(df['signal'] == 'wait') & (df['change'] > volatility) &
(df['close'] > df['boll_ub'] - df['close_std'] / 4) &
(df['close'] < df['boll_ub'] + df['close_std'] / 2), 'short',
df['signal'])
df = public_func(df, 'short', 'signal')
df['signal'] = np.where(
(df['signal'] == 'wait') &
(df['high'] >= df['signal_boll_ub'] + df['close_std'] / 4),
'close_short', df['signal'])
# 策略2
df['signal'] = np.where(
(df['signal'] == 'wait') & (df['change'] > volatility) &
(df['close'] < df['boll_lb'] + df['close_std'] / 4) &
(df['close'] > df['boll_lb'] - df['close_std'] / 2), 'long',
df['signal'])
df = public_func(df, 'long', 'signal')
df['signal'] = np.where(
(df['signal'] == 'wait') &
(df['low'] <= df['signal_boll_lb'] - df['close_std'] / 4),
'close_long', df['signal'])
# 策略3
df['signal'] = np.where(
(df['signal'] == 'wait') & (df['change'] > volatility) &
(df['close'].shift(1) < df['boll'].shift(1) + df['close_std'].shift(1))
& (df['close'] > df['boll'] + df['close_std']), 'long', df['signal'])
df['signal'] = np.where(
(df['signal'] == 'wait') &
(df['close'].shift(1) >
df['boll'].shift(1) + df['close_std'].shift(1) / 2) &
(df['close'] < df['boll'] + df['close_std'] / 2), 'close_long',
df['signal'])
# 策略4
df['signal'] = np.where(
(df['signal'] == 'wait') & (df['change'] > volatility) &
(df['close'].shift(1) > df['boll'].shift(1) - df['close_std'].shift(1))
& (df['close'] < df['boll'] - df['close_std']), 'short', df['signal'])
df['signal'] = np.where(
(df['signal'] == 'wait') &
(df['close'].shift(1) <
df['boll'].shift(1) - df['close_std'].shift(1) / 2) &
(df['close'] > df['boll'] - df['close_std'] / 2), 'close_short',
df['signal'])
return df.iloc[-1]
def k_analysis(k):
# 转DataFrame格式
k_data = pd.DataFrame(
k, columns=['timestamp', 'open', 'high', 'low', 'close', 'volume'])
t = k[0][0]
# 毫秒级
if len(str(t)) == 13:
k_data['timestamp'] = k_data['timestamp'] / 1000
stock = StockDataFrame.retype(k_data)
stock['middle_hl']
k_data['date'] = k_data['timestamp'].apply(
lambda x: time.strftime("%Y-%m-%d %H:%M:%S", time.localtime(x)))
k_data['date'] = pd.to_datetime(k_data['date'])
after_fenxing = pd.DataFrame()
temp_data = k_data.iloc[0]
zoushi = [] # 0:持平 -1:向下 1:向上
for i, row in k_data.iterrows():
# 第一根包含第二根
case1_1 = temp_data['high'] > row['high'] and temp_data['low'] < row[
'low']
case1_2 = temp_data['high'] > row['high'] and temp_data['low'] == row[
'low']
case1_3 = temp_data['high'] == row[
'high'] and temp_data['low'] < row['low']
# 第二根包含第一根
case2_1 = temp_data['high'] < row['high'] and temp_data['low'] > row[
'low']
case2_2 = temp_data['high'] < row['high'] and temp_data['low'] == row[
'low']
case2_3 = temp_data['high'] == row[
'high'] and temp_data['low'] > row['low']
# 第一根等于第二根
case3 = temp_data['high'] == row['high'] and temp_data['low'] == row[
'low']
# 向下趋势
case4 = temp_data['high'] > row['high'] and temp_data['low'] > row[
'low']
# 向上趋势
case5 = temp_data['high'] < row['high'] and temp_data['low'] < row[
'low']
if case1_1 or case1_2 or case1_3:
if zoushi[-1] == -1:
temp_data['high'] = row['high']
else:
temp_data['low'] = row['low']
elif case2_1 or case2_2 or case2_3:
temp_temp = temp_data
temp_data = row
if zoushi[-1] == -1:
temp_data['high'] = temp_temp['high']
else:
temp_data['low'] = temp_temp['low']
elif case3:
zoushi.append(0)
pass
elif case4:
zoushi.append(-1)
# 使用默认index: ignore_index=True:
after_fenxing = pd.concat(
[after_fenxing, temp_data.to_frame().T], ignore_index=True)
temp_data = row
elif case5:
zoushi.append(1)
after_fenxing = pd.concat(
[after_fenxing, temp_data.to_frame().T], ignore_index=True)
temp_data = row
# 因为使用candlestick2函数，要求输入open、close、high、low。为了美观，处理k线的最大最小值、开盘收盘价，之后k线不显示影线。
for i, row in after_fenxing.iterrows():
if row['open'] > row['close']:
row['open'] = row['high']
row['close'] = row['low']
else:
row['open'] = row['low']
row['close'] = row['high']
# 找出顶和底
temp_num = 0 # 上一个顶或底的位置
temp_high = 0 # 上一个顶的high值
temp_low = 0 # 上一个底的low值
temp_type = 0 # 上一个记录位置的类型 1-顶分型， 2-底分型
i = 1
fenxing_type = [] # 记录分型点的类型，1为顶分型，-1为底分型
fenxing_time = [] # 记录分型点的时间
fenxing_plot = [] # 记录点的数值，为顶分型去high值，为底分型去low值
fenxing_data = pd.DataFrame() # 分型点的DataFrame值
interval = 4
while (i < len(after_fenxing) - 1):
# 顶分型
case1 = after_fenxing.high[i - 1] < after_fenxing.high[
i] and after_fenxing.high[i] > after_fenxing.high[i + 1]
# 底分型
case2 = after_fenxing.low[i - 1] > after_fenxing.low[
i] and after_fenxing.low[i] < after_fenxing.low[i + 1]
if case1:
# 如果上一个分型为顶分型，则进行比较，选取更高的分型
if temp_type == 1:
if after_fenxing.high[i] <= temp_high:
i += 1
continue
else:
temp_high = after_fenxing.high[i]
temp_num = i
temp_type = 1
i += interval
# 如果上一个分型为底分型，则记录上一个分型，用当前分型与后面的分型比较，选取通向更极端的分型
elif temp_type == 2:
if temp_low >= after_fenxing.high[i]:
i += 1
else:
fenxing_type.append(-1)
fenxing_time.append(after_fenxing.date[temp_num])
fenxing_data = pd.concat(
[fenxing_data, after_fenxing[temp_num:temp_num + 1]],
axis=0)
fenxing_plot.append(after_fenxing.high[i])
temp_high = after_fenxing.high[i]
temp_num = i
temp_type = 1
i += interval
else:
temp_high = after_fenxing.high[i]
temp_num = i
temp_type = 1
i += interval
elif case2:
# 如果上一个分型为底分型，则进行比较，选取低点更低的分型
if temp_type == 2:
if after_fenxing.low[i] >= temp_low:
i += 1
continue
else:
temp_low = after_fenxing.low[i]
temp_num = i
temp_type = 2
i += interval
# 如果上一个分型为顶分型，则记录上一个分型，用当前分型与后面的分型比较，选取通向更极端的分型
elif temp_type == 1:
# 如果上一个顶分型的底比当前底分型的底低，则跳过当前的底分型
if temp_high <= after_fenxing.low[i]:
i += 1
else:
fenxing_type.append(1)
fenxing_time.append(after_fenxing.date[temp_num])
fenxing_data = pd.concat(
[fenxing_data, after_fenxing[temp_num:temp_num + 1]],
axis=0)
fenxing_plot.append(after_fenxing.low[i])
temp_low = after_fenxing.low[i]
temp_num = i
temp_type = 2
i += interval
else:
temp_low = after_fenxing.low[i]
temp_num = i
temp_type = 2
i += interval
else:
i += 1
return fenxing_type, fenxing_time, fenxing_plot, fenxing_data
'''
import pandas as pd
import numpy as np
from tools import data2df
import time
from tools.stockstats import StockDataFrame
import matplotlib.ticker as ticker
import matplotlib.pyplot as plt
filename = 'BTC2017-09-01-now-1H'
k_data = data2df.csv2df(filename + '.csv')
k_data = k_data.astype(float)
k_data['Timestamp'] = k_data['Timestamp'].astype(int)
stock = StockDataFrame.retype(k_data)
k_data['date'] = k_data['timestamp'].apply(
lambda x: time.strftime("%Y-%m-%d %H:%M:%S", time.localtime(x)))
k_data['date'] = pd.to_datetime(k_data['date'])
stock['middle_hl']
after_fenxing = pd.DataFrame()
temp_data = k_data.iloc[0]
zoushi = [] # 0:持平 -1:向下 1:向上
for i, row in k_data.iterrows():
# 第一根包含第二根
case1_1 = temp_data['high'] > row['high'] and temp_data['low'] < row['low']
case1_2 = temp_data['high'] > row['high'] and temp_data['low'] == row['low']