adelorenzo/stable
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Merge pull request #930 from freqtrade/skopt

Browse Source 
Replace Hyperopt with scikit-optimize
This commit is contained in:
Samuel Husso
2018-07-04 13:51:14 -05:00
committed by GitHub
parent 64c68d93c3 96bb2efe69
commit d8d0579c5a
5 changed files with 296 additions and 751 deletions
Download Patch File Download Diff File Expand all files Collapse all files

463
freqtrade/optimize/hyperopt.py
View File

@@ -4,22 +4,21 @@
This module contains the hyperopt logic
"""
import json
import logging
import multiprocessing
import os
import pickle
import signal
import sys
from argparse import Namespace
from functools import reduce
from math import exp
from operator import itemgetter
from typing import Dict, Any, Callable, Optional
from typing import Any, Callable, Dict, List
import numpy
import talib.abstract as ta
from hyperopt import STATUS_FAIL, STATUS_OK, Trials, fmin, hp, space_eval, tpe
from pandas import DataFrame
from sklearn.externals.joblib import Parallel, delayed, dump, load
from skopt import Optimizer
from skopt.space import Categorical, Dimension, Integer, Real
import freqtrade.vendor.qtpylib.indicators as qtpylib
from freqtrade.arguments import Arguments
@@ -29,6 +28,9 @@ from freqtrade.optimize.backtesting import Backtesting
logger = logging.getLogger(__name__)
MAX_LOSS = 100000 # just a big enough number to be bad result in loss optimization
TICKERDATA_PICKLE = os.path.join('user_data', 'hyperopt_tickerdata.pkl')
class Hyperopt(Backtesting):
"""
@@ -44,7 +46,6 @@ class Hyperopt(Backtesting):
# to the number of days
self.target_trades = 600
self.total_tries = config.get('epochs', 0)
self.current_tries = 0
self.current_best_loss = 100
# max average trade duration in minutes
@@ -56,130 +57,38 @@ class Hyperopt(Backtesting):
# check that the reported Σ% values do not exceed this!
self.expected_max_profit = 3.0
# Configuration and data used by hyperopt
self.processed: Optional[Dict[str, Any]] = None
# Previous evaluations
self.trials_file = os.path.join('user_data', 'hyperopt_results.pickle')
self.trials: List = []
# Hyperopt Trials
self.trials_file = os.path.join('user_data', 'hyperopt_trials.pickle')
self.trials = Trials()
def get_args(self, params):
dimensions = self.hyperopt_space()
# Ensure the number of dimensions match
# the number of parameters in the list x.
if len(params) != len(dimensions):
raise ValueError('Mismatch in number of search-space dimensions. '
f'len(dimensions)=={len(dimensions)} and len(x)=={len(params)}')
# Create a dict where the keys are the names of the dimensions
# and the values are taken from the list of parameters x.
arg_dict = {dim.name: value for dim, value in zip(dimensions, params)}
return arg_dict
@staticmethod
def populate_indicators(dataframe: DataFrame) -> DataFrame:
"""
Adds several different TA indicators to the given DataFrame
"""
dataframe['adx'] = ta.ADX(dataframe)
dataframe['ao'] = qtpylib.awesome_oscillator(dataframe)
dataframe['cci'] = ta.CCI(dataframe)
macd = ta.MACD(dataframe)
dataframe['macd'] = macd['macd']
dataframe['macdsignal'] = macd['macdsignal']
dataframe['macdhist'] = macd['macdhist']
dataframe['mfi'] = ta.MFI(dataframe)
dataframe['minus_dm'] = ta.MINUS_DM(dataframe)
dataframe['minus_di'] = ta.MINUS_DI(dataframe)
dataframe['plus_dm'] = ta.PLUS_DM(dataframe)
dataframe['plus_di'] = ta.PLUS_DI(dataframe)
dataframe['roc'] = ta.ROC(dataframe)
dataframe['rsi'] = ta.RSI(dataframe)
# Inverse Fisher transform on RSI, values [-1.0, 1.0] (https://goo.gl/2JGGoy)
rsi = 0.1 * (dataframe['rsi'] - 50)
dataframe['fisher_rsi'] = (numpy.exp(2 * rsi) - 1) / (numpy.exp(2 * rsi) + 1)
# Inverse Fisher transform on RSI normalized, value [0.0, 100.0] (https://goo.gl/2JGGoy)
dataframe['fisher_rsi_norma'] = 50 * (dataframe['fisher_rsi'] + 1)
# Stoch
stoch = ta.STOCH(dataframe)
dataframe['slowd'] = stoch['slowd']
dataframe['slowk'] = stoch['slowk']
# Stoch fast
stoch_fast = ta.STOCHF(dataframe)
dataframe['fastd'] = stoch_fast['fastd']
dataframe['fastk'] = stoch_fast['fastk']
# Stoch RSI
stoch_rsi = ta.STOCHRSI(dataframe)
dataframe['fastd_rsi'] = stoch_rsi['fastd']
dataframe['fastk_rsi'] = stoch_rsi['fastk']
dataframe['minus_di'] = ta.MINUS_DI(dataframe)
# Bollinger bands
bollinger = qtpylib.bollinger_bands(qtpylib.typical_price(dataframe), window=20, stds=2)
dataframe['bb_lowerband'] = bollinger['lower']
dataframe['bb_middleband'] = bollinger['mid']
dataframe['bb_upperband'] = bollinger['upper']
# EMA - Exponential Moving Average
dataframe['ema3'] = ta.EMA(dataframe, timeperiod=3)
dataframe['ema5'] = ta.EMA(dataframe, timeperiod=5)
dataframe['ema10'] = ta.EMA(dataframe, timeperiod=10)
dataframe['ema50'] = ta.EMA(dataframe, timeperiod=50)
dataframe['ema100'] = ta.EMA(dataframe, timeperiod=100)
# SAR Parabolic
dataframe['sar'] = ta.SAR(dataframe)
# SMA - Simple Moving Average
dataframe['sma'] = ta.SMA(dataframe, timeperiod=40)
# TEMA - Triple Exponential Moving Average
dataframe['tema'] = ta.TEMA(dataframe, timeperiod=9)
# Hilbert Transform Indicator - SineWave
hilbert = ta.HT_SINE(dataframe)
dataframe['htsine'] = hilbert['sine']
dataframe['htleadsine'] = hilbert['leadsine']
# Pattern Recognition - Bullish candlestick patterns
# ------------------------------------
"""
# Hammer: values [0, 100]
dataframe['CDLHAMMER'] = ta.CDLHAMMER(dataframe)
# Inverted Hammer: values [0, 100]
dataframe['CDLINVERTEDHAMMER'] = ta.CDLINVERTEDHAMMER(dataframe)
# Dragonfly Doji: values [0, 100]
dataframe['CDLDRAGONFLYDOJI'] = ta.CDLDRAGONFLYDOJI(dataframe)
# Piercing Line: values [0, 100]
dataframe['CDLPIERCING'] = ta.CDLPIERCING(dataframe) # values [0, 100]
# Morningstar: values [0, 100]
dataframe['CDLMORNINGSTAR'] = ta.CDLMORNINGSTAR(dataframe) # values [0, 100]
# Three White Soldiers: values [0, 100]
dataframe['CDL3WHITESOLDIERS'] = ta.CDL3WHITESOLDIERS(dataframe) # values [0, 100]
"""
# Pattern Recognition - Bearish candlestick patterns
# ------------------------------------
"""
# Hanging Man: values [0, 100]
dataframe['CDLHANGINGMAN'] = ta.CDLHANGINGMAN(dataframe)
# Shooting Star: values [0, 100]
dataframe['CDLSHOOTINGSTAR'] = ta.CDLSHOOTINGSTAR(dataframe)
# Gravestone Doji: values [0, 100]
dataframe['CDLGRAVESTONEDOJI'] = ta.CDLGRAVESTONEDOJI(dataframe)
# Dark Cloud Cover: values [0, 100]
dataframe['CDLDARKCLOUDCOVER'] = ta.CDLDARKCLOUDCOVER(dataframe)
# Evening Doji Star: values [0, 100]
dataframe['CDLEVENINGDOJISTAR'] = ta.CDLEVENINGDOJISTAR(dataframe)
# Evening Star: values [0, 100]
dataframe['CDLEVENINGSTAR'] = ta.CDLEVENINGSTAR(dataframe)
"""
# Pattern Recognition - Bullish/Bearish candlestick patterns
# ------------------------------------
"""
# Three Line Strike: values [0, -100, 100]
dataframe['CDL3LINESTRIKE'] = ta.CDL3LINESTRIKE(dataframe)
# Spinning Top: values [0, -100, 100]
dataframe['CDLSPINNINGTOP'] = ta.CDLSPINNINGTOP(dataframe) # values [0, -100, 100]
# Engulfing: values [0, -100, 100]
dataframe['CDLENGULFING'] = ta.CDLENGULFING(dataframe) # values [0, -100, 100]
# Harami: values [0, -100, 100]
dataframe['CDLHARAMI'] = ta.CDLHARAMI(dataframe) # values [0, -100, 100]
# Three Outside Up/Down: values [0, -100, 100]
dataframe['CDL3OUTSIDE'] = ta.CDL3OUTSIDE(dataframe) # values [0, -100, 100]
# Three Inside Up/Down: values [0, -100, 100]
dataframe['CDL3INSIDE'] = ta.CDL3INSIDE(dataframe) # values [0, -100, 100]
"""
# Chart type
# ------------------------------------
# Heikinashi stategy
heikinashi = qtpylib.heikinashi(dataframe)
dataframe['ha_open'] = heikinashi['open']
dataframe['ha_close'] = heikinashi['close']
dataframe['ha_high'] = heikinashi['high']
dataframe['ha_low'] = heikinashi['low']
return dataframe
@@ -187,15 +96,16 @@ class Hyperopt(Backtesting):
"""
Save hyperopt trials to file
"""
logger.info('Saving Trials to \'%s\'', self.trials_file)
pickle.dump(self.trials, open(self.trials_file, 'wb'))
if self.trials:
logger.info('Saving %d evaluations to \'%s\'', len(self.trials), self.trials_file)
dump(self.trials, self.trials_file)
def read_trials(self) -> Trials:
def read_trials(self) -> List:
"""
Read hyperopt trials file
"""
logger.info('Reading Trials from \'%s\'', self.trials_file)
trials = pickle.load(open(self.trials_file, 'rb'))
trials = load(self.trials_file)
os.remove(self.trials_file)
return trials
@@ -203,9 +113,15 @@ class Hyperopt(Backtesting):
"""
Display Best hyperopt result
"""
vals = json.dumps(self.trials.best_trial['misc']['vals'], indent=4)
results = self.trials.best_trial['result']['result']
logger.info('Best result:\n%s\nwith values:\n%s', results, vals)
results = sorted(self.trials, key=itemgetter('loss'))
best_result = results[0]
logger.info(
'Best result:\n%s\nwith values:\n%s',
best_result['result'],
best_result['params']
)
if 'roi_t1' in best_result['params']:
logger.info('ROI table:\n%s', self.generate_roi_table(best_result['params']))
def log_results(self, results) -> None:
"""
@@ -231,7 +147,8 @@ class Hyperopt(Backtesting):
trade_loss = 1 - 0.25 * exp(-(trade_count - self.target_trades) ** 2 / 10 ** 5.8)
profit_loss = max(0, 1 - total_profit / self.expected_max_profit)
duration_loss = 0.4 * min(trade_duration / self.max_accepted_trade_duration, 1)
return trade_loss + profit_loss + duration_loss
result = trade_loss + profit_loss + duration_loss
return result
@staticmethod
def generate_roi_table(params: Dict) -> Dict[int, float]:
@@ -247,87 +164,44 @@ class Hyperopt(Backtesting):
return roi_table
@staticmethod
def roi_space() -> Dict[str, Any]:
def roi_space() -> List[Dimension]:
"""
Values to search for each ROI steps
"""
return {
'roi_t1': hp.quniform('roi_t1', 10, 120, 20),
'roi_t2': hp.quniform('roi_t2', 10, 60, 15),
'roi_t3': hp.quniform('roi_t3', 10, 40, 10),
'roi_p1': hp.quniform('roi_p1', 0.01, 0.04, 0.01),
'roi_p2': hp.quniform('roi_p2', 0.01, 0.07, 0.01),
'roi_p3': hp.quniform('roi_p3', 0.01, 0.20, 0.01),
}
return [
Integer(10, 120, name='roi_t1'),
Integer(10, 60, name='roi_t2'),
Integer(10, 40, name='roi_t3'),
Real(0.01, 0.04, name='roi_p1'),
Real(0.01, 0.07, name='roi_p2'),
Real(0.01, 0.20, name='roi_p3'),
]
@staticmethod
def stoploss_space() -> Dict[str, Any]:
def stoploss_space() -> List[Dimension]:
"""
Stoploss Value to search
Stoploss search space
"""
return {
'stoploss': hp.quniform('stoploss', -0.5, -0.02, 0.02),
}
return [
Real(-0.5, -0.02, name='stoploss'),
]
@staticmethod
def indicator_space() -> Dict[str, Any]:
def indicator_space() -> List[Dimension]:
"""
Define your Hyperopt space for searching strategy parameters
"""
return {
'macd_below_zero': hp.choice('macd_below_zero', [
{'enabled': False},
{'enabled': True}
]),
'mfi': hp.choice('mfi', [
{'enabled': False},
{'enabled': True, 'value': hp.quniform('mfi-value', 10, 25, 5)}
]),
'fastd': hp.choice('fastd', [
{'enabled': False},
{'enabled': True, 'value': hp.quniform('fastd-value', 15, 45, 5)}
]),
'adx': hp.choice('adx', [
{'enabled': False},
{'enabled': True, 'value': hp.quniform('adx-value', 20, 50, 5)}
]),
'rsi': hp.choice('rsi', [
{'enabled': False},
{'enabled': True, 'value': hp.quniform('rsi-value', 20, 40, 5)}
]),
'uptrend_long_ema': hp.choice('uptrend_long_ema', [
{'enabled': False},
{'enabled': True}
]),
'uptrend_short_ema': hp.choice('uptrend_short_ema', [
{'enabled': False},
{'enabled': True}
]),
'over_sar': hp.choice('over_sar', [
{'enabled': False},
{'enabled': True}
]),
'green_candle': hp.choice('green_candle', [
{'enabled': False},
{'enabled': True}
]),
'uptrend_sma': hp.choice('uptrend_sma', [
{'enabled': False},
{'enabled': True}
]),
'trigger': hp.choice('trigger', [
{'type': 'lower_bb'},
{'type': 'lower_bb_tema'},
{'type': 'faststoch10'},
{'type': 'ao_cross_zero'},
{'type': 'ema3_cross_ema10'},
{'type': 'macd_cross_signal'},
{'type': 'sar_reversal'},
{'type': 'ht_sine'},
{'type': 'heiken_reversal_bull'},
{'type': 'di_cross'},
]),
}
return [
Integer(10, 25, name='mfi-value'),
Integer(15, 45, name='fastd-value'),
Integer(20, 50, name='adx-value'),
Integer(20, 40, name='rsi-value'),
Categorical([True, False], name='mfi-enabled'),
Categorical([True, False], name='fastd-enabled'),
Categorical([True, False], name='adx-enabled'),
Categorical([True, False], name='rsi-enabled'),
Categorical(['bb_lower', 'macd_cross_signal', 'sar_reversal'], name='trigger')
]
def has_space(self, space: str) -> bool:
"""
@@ -337,17 +211,17 @@ class Hyperopt(Backtesting):
return True
return False
def hyperopt_space(self) -> Dict[str, Any]:
def hyperopt_space(self) -> List[Dimension]:
"""
Return the space to use during Hyperopt
"""
spaces: Dict = {}
spaces: List[Dimension] = []
if self.has_space('buy'):
spaces = {**spaces, **Hyperopt.indicator_space()}
spaces += Hyperopt.indicator_space()
if self.has_space('roi'):
spaces = {**spaces, **Hyperopt.roi_space()}
spaces += Hyperopt.roi_space()
if self.has_space('stoploss'):
spaces = {**spaces, **Hyperopt.stoploss_space()}
spaces += Hyperopt.stoploss_space()
return spaces
@staticmethod
@@ -361,63 +235,26 @@ class Hyperopt(Backtesting):
"""
conditions = []
# GUARDS AND TRENDS
if 'uptrend_long_ema' in params and params['uptrend_long_ema']['enabled']:
conditions.append(dataframe['ema50'] > dataframe['ema100'])
if 'macd_below_zero' in params and params['macd_below_zero']['enabled']:
conditions.append(dataframe['macd'] < 0)
if 'uptrend_short_ema' in params and params['uptrend_short_ema']['enabled']:
conditions.append(dataframe['ema5'] > dataframe['ema10'])
if 'mfi' in params and params['mfi']['enabled']:
conditions.append(dataframe['mfi'] < params['mfi']['value'])
if 'fastd' in params and params['fastd']['enabled']:
conditions.append(dataframe['fastd'] < params['fastd']['value'])
if 'adx' in params and params['adx']['enabled']:
conditions.append(dataframe['adx'] > params['adx']['value'])
if 'rsi' in params and params['rsi']['enabled']:
conditions.append(dataframe['rsi'] < params['rsi']['value'])
if 'over_sar' in params and params['over_sar']['enabled']:
conditions.append(dataframe['close'] > dataframe['sar'])
if 'green_candle' in params and params['green_candle']['enabled']:
conditions.append(dataframe['close'] > dataframe['open'])
if 'uptrend_sma' in params and params['uptrend_sma']['enabled']:
prevsma = dataframe['sma'].shift(1)
conditions.append(dataframe['sma'] > prevsma)
if 'mfi-enabled' in params and params['mfi-enabled']:
conditions.append(dataframe['mfi'] < params['mfi-value'])
if 'fastd-enabled' in params and params['fastd-enabled']:
conditions.append(dataframe['fastd'] < params['fastd-value'])
if 'adx-enabled' in params and params['adx-enabled']:
conditions.append(dataframe['adx'] > params['adx-value'])
if 'rsi-enabled' in params and params['rsi-enabled']:
conditions.append(dataframe['rsi'] < params['rsi-value'])
# TRIGGERS
triggers = {
'lower_bb': (
dataframe['close'] < dataframe['bb_lowerband']
),
'lower_bb_tema': (
dataframe['tema'] < dataframe['bb_lowerband']
),
'faststoch10': (qtpylib.crossed_above(
dataframe['fastd'], 10.0
)),
'ao_cross_zero': (qtpylib.crossed_above(
dataframe['ao'], 0.0
)),
'ema3_cross_ema10': (qtpylib.crossed_above(
dataframe['ema3'], dataframe['ema10']
)),
'macd_cross_signal': (qtpylib.crossed_above(
if params['trigger'] == 'bb_lower':
conditions.append(dataframe['close'] < dataframe['bb_lowerband'])
if params['trigger'] == 'macd_cross_signal':
conditions.append(qtpylib.crossed_above(
dataframe['macd'], dataframe['macdsignal']
)),
'sar_reversal': (qtpylib.crossed_above(
))
if params['trigger'] == 'sar_reversal':
conditions.append(qtpylib.crossed_above(
dataframe['close'], dataframe['sar']
)),
'ht_sine': (qtpylib.crossed_above(
dataframe['htleadsine'], dataframe['htsine']
)),
'heiken_reversal_bull': (
(qtpylib.crossed_above(dataframe['ha_close'], dataframe['ha_open'])) &
(dataframe['ha_low'] == dataframe['ha_open'])
),
'di_cross': (qtpylib.crossed_above(
dataframe['plus_di'], dataframe['minus_di']
)),
}
conditions.append(triggers.get(params['trigger']['type']))
))
dataframe.loc[
reduce(lambda x, y: x & y, conditions),
@@ -427,7 +264,9 @@ class Hyperopt(Backtesting):
return populate_buy_trend
def generate_optimizer(self, params: Dict) -> Dict:
def generate_optimizer(self, _params) -> Dict:
params = self.get_args(_params)
if self.has_space('roi'):
self.analyze.strategy.minimal_roi = self.generate_roi_table(params)
@@ -437,10 +276,11 @@ class Hyperopt(Backtesting):
if self.has_space('stoploss'):
self.analyze.strategy.stoploss = params['stoploss']
processed = load(TICKERDATA_PICKLE)
results = self.backtest(
{
'stake_amount': self.config['stake_amount'],
'processed': self.processed,
'processed': processed,
'realistic': self.config.get('realistic_simulation', False),
}
)
@@ -450,30 +290,18 @@ class Hyperopt(Backtesting):
trade_count = len(results.index)
trade_duration = results.trade_duration.mean()
if trade_count == 0 or trade_duration > self.max_accepted_trade_duration:
print('.', end='')
sys.stdout.flush()
if trade_count == 0:
return {
'status': STATUS_FAIL,
'loss': float('inf')
'loss': MAX_LOSS,
'params': params,
'result': result_explanation,
}
loss = self.calculate_loss(total_profit, trade_count, trade_duration)
self.current_tries += 1
self.log_results(
{
'loss': loss,
'current_tries': self.current_tries,
'total_tries': self.total_tries,
'result': result_explanation,
}
)
return {
'loss': loss,
'status': STATUS_OK,
'params': params,
'result': result_explanation,
}
@@ -491,6 +319,27 @@ class Hyperopt(Backtesting):
results.trade_duration.mean(),
)
def get_optimizer(self, cpu_count) -> Optimizer:
return Optimizer(
self.hyperopt_space(),
base_estimator="ET",
acq_optimizer="auto",
n_initial_points=30,
acq_optimizer_kwargs={'n_jobs': cpu_count}
)
def run_optimizer_parallel(self, parallel, asked) -> List:
return parallel(delayed(self.generate_optimizer)(v) for v in asked)
def load_previous_results(self):
""" read trials file if we have one """
if os.path.exists(self.trials_file) and os.path.getsize(self.trials_file) > 0:
self.trials = self.read_trials()
logger.info(
'Loaded %d previous evaluations from disk.',
len(self.trials)
)
def start(self) -> None:
timerange = Arguments.parse_timerange(None if self.config.get(
'timerange') is None else str(self.config.get('timerange')))
@@ -503,67 +352,35 @@ class Hyperopt(Backtesting):
if self.has_space('buy'):
self.analyze.populate_indicators = Hyperopt.populate_indicators # type: ignore
self.processed = self.tickerdata_to_dataframe(data)
dump(self.tickerdata_to_dataframe(data), TICKERDATA_PICKLE)
self.exchange = None # type: ignore
self.load_previous_results()
logger.info('Preparing Trials..')
signal.signal(signal.SIGINT, self.signal_handler)
# read trials file if we have one
if os.path.exists(self.trials_file) and os.path.getsize(self.trials_file) > 0:
self.trials = self.read_trials()
self.current_tries = len(self.trials.results)
self.total_tries += self.current_tries
logger.info(
'Continuing with trials. Current: %d, Total: %d',
self.current_tries,
self.total_tries
)
cpus = multiprocessing.cpu_count()
logger.info(f'Found {cpus} CPU cores. Let\'s make them scream!')
opt = self.get_optimizer(cpus)
EVALS = max(self.total_tries//cpus, 1)
try:
best_parameters = fmin(
fn=self.generate_optimizer,
space=self.hyperopt_space(),
algo=tpe.suggest,
max_evals=self.total_tries,
trials=self.trials
)
with Parallel(n_jobs=cpus) as parallel:
for i in range(EVALS):
asked = opt.ask(n_points=cpus)
f_val = self.run_optimizer_parallel(parallel, asked)
opt.tell(asked, [i['loss'] for i in f_val])
results = sorted(self.trials.results, key=itemgetter('loss'))
best_result = results[0]['result']
except ValueError:
best_parameters = {}
best_result = 'Sorry, Hyperopt was not able to find good parameters. Please ' \
'try with more epochs (param: -e).'
# Improve best parameter logging display
if best_parameters:
best_parameters = space_eval(
self.hyperopt_space(),
best_parameters
)
logger.info('Best parameters:\n%s', json.dumps(best_parameters, indent=4))
if 'roi_t1' in best_parameters:
logger.info('ROI table:\n%s', self.generate_roi_table(best_parameters))
logger.info('Best Result:\n%s', best_result)
# Store trials result to file to resume next time
self.save_trials()
def signal_handler(self, sig, frame) -> None:
"""
Hyperopt SIGINT handler
"""
logger.info(
'Hyperopt received %s',
signal.Signals(sig).name
)
self.trials += f_val
for j in range(cpus):
self.log_results({
'loss': f_val[j]['loss'],
'current_tries': i * cpus + j,
'total_tries': self.total_tries,
'result': f_val[j]['result'],
})
except KeyboardInterrupt:
print('User interrupted..')
self.save_trials()
self.log_trials_result()
sys.exit(0)
def start(args: Namespace) -> None: