import nn as nn
from stock import Stock
import stock_data as sd
import getSP500 as sp
from top10 import top10
import time
import datetime
import os
if os.geteuid() != 0:
exit("You need to have root privileges to run this script.\nPlease try again, this time using 'sudo'. Exiting.")
run_this = True
while True:
now = datetime.datetime.now()
if now.hour >= 0 and now.hour < 16:
run_this = True
if now.hour >= 16 and run_this == True:
nn.run()
run_this = False
time.sleep(3550)
for c in Lambda for d in weight for e in batch_size
for f in epochs for g in activation]
parm_dict = {} # holds the hyperparameter combination for one run
results = [] # holds the hyperparemeters and results for each run
start_time = time.time()
loop = count = 1
for x in parms:
for i in range(loop):
parm_dict['l1_size'] = x[0]
parm_dict['learning_rate'] = x[1]
parm_dict['lambda'] = x[2]
parm_dict['weight'] = x[3]
parm_dict['batch_size'] = x[4]
parm_dict['epochs'] = x[5]
parm_dict['activation'] = x[6]
job_name = "job_" + job_id + "/" + "run_" + str(count)
lift = run(data_dict, parm_dict, job_name)
tup = (count, parm_dict, lift)
results.append(tup)
count += 1
# Write out a summary of the results
writeResults(results, job_id)
job_id = int(job_id)
job.setJob(job_id + 1)
print("Job {} complete after {:,.0f} minutes".format(
str(job_id), (time.time() - start_time) / 60))
# train.hist(column='dQrauto_KCE', bins =100)
# 2. Scaling and Transform original data
utils.pTitle("2. Scaling and Transform original data")
train_val = train.values
scaler = StandardScaler()
# scaler = StandardScaler(with_mean=False)
scaler.fit(train_val)
print(scaler.mean_)
train_scal = scaler.transform(train_val)
Xtrain = train_scal[:, 0:2]
Ytrain = train_scal[:, 2]
test_val = test.values
test_scal = scaler.transform(test_val)
Xtest = test_scal[:, 0:2]
Ytest = test_scal[:, 2]
#! Explore
# plt.hist(Ytrain, bins = 100)
neuronas = [15, 45, 65, 75, 115, 150]
for i in neuronas:
capas = [2, 3, 4]
for c in capas:
utils.pTitle2("Neuronas: " + str(neuronas))
out = nn.run(Xtrain, Ytrain, Xtest, Ytest, resdir, i, c)
#! save scaled real data
# #predictions.T[0],Ytest
out = (out * np.sqrt(scaler.var_[2])) + scaler.mean_[2]
out.to_csv(resdir + "/Predic.csv")
for f in epochs for g in activation]
parm_dict = {}
count = 1
start_time = time.time()
loop = 1
for i in range(loop):
for x in parms:
loop_time = time.time()
parm_dict['l1_size'] = x[0]
parm_dict['learning_rate'] = x[1]
parm_dict['lambda'] = x[2]
parm_dict['weight'] = x[3]
parm_dict['batch_size'] = x[4]
parm_dict['epochs'] = x[5]
parm_dict['activation'] = x[6]
results = run(data_dict, parm_dict, count)
save_results(count, parm_dict, results)
count += 1
job_id += 1
job.setJob(job_id)
summary.close()
print('Total time: {:,.0f} minutes'.format(
(time.time() - start_time) / 60))
sys.exit()