layer_size = [(8, ), (10, ), (50, ), (10, 4), (8, 3)]
solvers = ['adam', 'lbfgs']
learning_rate = ['constant', 'invscaling', 'adaptive']
learning_rate_init = [0.001, 0.005, 0.01, 0.05]
act = ['relu', 'logistic', 'tanh']
# Get all possible configs
configs = []
for l in layer_size:
for s in solvers:
for lr in learning_rate:
for lri in learning_rate_init:
for a in act:
configs.append({
'params': {
'hidden_layer_sizes': l,
'solver': s,
'learning_rate': lr,
'learning_rate_init': lri,
'activation': a
},
'x_train': x_train,
'y_train': y_train,
'x_test': x_test,
'y_test': y_test
})
return configs
ModelFactory.register_model('mlp', MLPModel)
def predict(self, x, **kwargs):
pred = self.model.predict(x)
return pred
@with_xy
def get_grid_search_configs(self, **kwargs):
kernels = ['rbf']
cs = np.arange(0.5, 5.0, 0.5)
gammas = ['auto', 0.001, 0.005, 0.01, 0.05]
# Get all possible configs
configs = []
for k in kernels:
for c in cs:
for g in gammas:
configs.append({
'params': {
'kernel': k,
'C': c,
'gamma': g,
},
'x_train': kwargs.get('x_train'),
'y_train': kwargs.get('y_train'),
'x_test': kwargs.get('x_test'),
'y_test': kwargs.get('y_test')
})
return configs
ModelFactory.register_model('svc', SVCModel)
# Maximum number of iterations of the k-means algorithm for a single run
max_iter = [150, 300, 600]
# Relative tolerance with regards to inertia to declare convergence
tolerance = [0.0001, 0.0005, 0.0010]
# Get all possible configs
configs = []
for c in n_clusters:
for n in n_init:
for i in max_iter:
for t in tolerance:
configs.append({
'params': {
'n_clusters': c,
'init':
'k-means++', # k-means++, random or ndarray
'n_init': n,
'max_iter': i,
'tol': t,
'n_jobs': None,
},
'x_train': kwargs.get('x_train'),
'y_train': kwargs.get('y_train'),
'x_test': kwargs.get('x_test'),
'y_test': kwargs.get('y_test')
})
return configs
ModelFactory.register_model('kmeans', KMeansModel)
@with_xy
def get_grid_search_configs(self, **kwargs):
weights = ['uniform', 'distance']
n_neighbors = [2, 3, 5, 9]
leaf_size = [10]
P = [2]
# Get all possible configs
configs = []
for w in weights:
for n in n_neighbors:
for l in leaf_size:
for p in P:
configs.append({
'params': {
'n_neighbors': n,
'weights': w,
'leaf_size': l,
'p': p,
'n_jobs': None
},
'x_train': kwargs.get('x_train'),
'y_train': kwargs.get('y_train'),
'x_test': kwargs.get('x_test'),
'y_test': kwargs.get('y_test')
})
return configs
ModelFactory.register_model('knn', KNNModel)
self.params.get('alpha'), self.params.get('beta')))
return
@with_x
def get_grid_search_configs(self, **kwargs):
x_train = kwargs.get('x_train')
x_test = kwargs.get('x_test')
seasonal_periods = range(1, 14)
alpha = np.arange(0.5, 3, 0.5)
beta = np.arange(0.0, 3.0, 0.5)
# Get all possible configs
configs = []
for p in seasonal_periods:
for a in alpha:
for b in beta:
configs.append({
'params': {
'seasonal_periods': p,
'alpha': a,
'beta': b,
},
'x_train': x_train,
'x_test': x_test
})
return configs
ModelFactory.register_model('expsmooth', ExpSmoothModel)
return to_discrete_double(forecast, -0.01, 0.01)
except (ValueError, np.linalg.linalg.LinAlgError):
logger.error('{}: convergence error'.format(str(self.model.get_params())))
@with_x
def get_grid_search_configs(self, **kwargs):
x_train = kwargs.get('x_train')
x_test = kwargs.get('x_test')
p_values = range(0,6)
d_values = range(0,6)
q_values = range(0,6)
# If series is stationary, don't apply differentiation
adf = adfuller(x_train) # 0 is score, 1 is pvalue
if adf[1] < 0.05: # Null hp rejected, series is stationary and requires no differentiation
logger.info('Series is stationary, no need for differencing')
d_values = [0] # Set d = 0
# Get all possible configs
configs = []
for p in p_values:
for d in d_values:
for q in q_values:
configs.append({
'params':{'order': (p,d,q)},
'x_train' : x_train,
'x_test' : x_test
})
return configs
ModelFactory.register_model('arima', ARIMAModel)
if has_negative(x):
x = scale(x, scaler='minmax', feature_range=(0, 1))
pred = self.model.predict(x)
return pred
@with_xy
def get_grid_search_configs(self, **kwargs):
alpha = [
0.0000000001, 0.001, 0.005, 0.01, 0.05, 0.1, 0.3, 0.5, 0.8, 1.0,
1.5, 2.0
]
fit_prior = [True, False]
# Get all possible configs
configs = []
for k in alpha:
for f in fit_prior:
configs.append({
'params': {
'alpha': k,
'fit_prior': f
},
'x_train': kwargs.get('x_train'),
'y_train': kwargs.get('y_train'),
'x_test': kwargs.get('x_test'),
'y_test': kwargs.get('y_test')
})
return configs
ModelFactory.register_model('mnb', MNBModel)
@with_x
def get_grid_search_configs(self, **kwargs):
x_train = kwargs.get('x_train')
x_test = kwargs.get('x_test')
p_values = range(0, 6)
d_values = range(0, 6)
q_values = range(0, 6)
# If series is stationary, don't apply differentiation
adf = adfuller(x_train) # 0 is score, 1 is pvalue
if adf[1] < 0.05: # Null hp rejected, series is stationary and requires no differentiation
logger.info('Series is stationary, no need for differencing')
d_values = [0] # Set d = 0
# Get all possible configs
configs = []
for p in p_values:
for d in d_values:
for q in q_values:
configs.append({
'params': {
'order': (p, d, q)
},
'x_train': x_train,
'x_test': x_test
})
return configs
ModelFactory.register_model('sarima', SARIMAModel)
import logging
from lib.log import logger
from lib.symbol import Symbol, DatasetType
from lib.models import ModelFactory
from lib.job import Job
from lib.report import ReportCollection
import pandas as pd
logger.setup(filename='../job_test.log',
filemode='w',
root_level=logging.DEBUG,
log_level=logging.DEBUG,
logger='job_test')
ModelFactory.discover()
ohlcv = pd.read_csv("./data/result/ohlcv.csv",
sep=',',
encoding='utf-8',
index_col='Date',
parse_dates=True)
btc = pd.read_csv("./data/coinmetrics.io/btc.csv",
sep=',',
encoding='utf-8',
index_col='date',
parse_dates=True)
_sym = 'BTC'
s = Symbol(_sym,
ohlcv=ohlcv,
blockchain=btc,
column_map={