def main(train_stock,
val_stock,
window_size,
batch_size,
ep_count,
strategy="t-dqn",
model_name="model_debug",
pretrained=False,
debug=False,
cpu=False):
""" Trains the stock trading bot using Deep Q-Learning.
Please see https://arxiv.org/abs/1312.5602 for more details.
Args: [python train.py --help]
"""
if cpu:
switch_k_backend_device()
print("#TF Version: ", tf.__version__)
print("Using device: ")
print(device_lib.list_local_devices())
agent = Agent(window_size,
strategy=strategy,
pretrained=pretrained,
model_name=model_name)
train_data = get_stock_data(train_stock)
val_data = get_stock_data(val_stock)
initial_offset = val_data[1] - val_data[0]
for episode in range(1, ep_count + 1):
train_result = train_model(agent,
episode,
train_data,
ep_count=ep_count,
batch_size=batch_size,
window_size=window_size)
val_result, _, actionCollection = evaluate_model(
agent, val_data, window_size, debug)
show_train_result(train_result, val_result, initial_offset)
def quick_train(train_stock): generate(train_stock) training_stock = 'training_data/' + train_stock + '.csv' strategy = 'double-dqn' window_size = 10 batch_size = 32 ep_count = 10 model_name = 'model_double-dqn_GOOG_50' pretrained = False debug = False coloredlogs.install(level="DEBUG") switch_k_backend_device() print(training_stock) main(training_stock, window_size, batch_size, ep_count, strategy=strategy, model_name=model_name, pretrained=pretrained, debug=debug)
if __name__ == "__main__":
args = docopt(__doc__)
train_stock = args["<train-stock>"]
val_stock = args["<val-stock>"]
strategy = args["--strategy"]
window_size = int(args["--window-size"])
batch_size = int(args["--batch-size"])
ep_count = int(args["--episode-count"])
model_name = args["--model-name"]
pretrained = args["--pretrained"]
debug = args["--debug"]
coloredlogs.install(level="DEBUG")
switch_k_backend_device()
try:
main(train_stock,
val_stock,
window_size,
batch_size,
ep_count,
strategy=strategy,
model_name=model_name,
pretrained=pretrained,
debug=debug)
except KeyboardInterrupt:
print("Aborted!")
def fn(ticker,strategy,ep_count,year):
stock = '{}'.format(ticker)
model_name = '{}_{}_{}'.format(ticker,strategy,ep_count)
test_stock = 'data/Nifty50/Split/{}_{}.csv'.format(ticker,year)
window_size = 10
debug = True
tf.keras.backend.clear_session()
K.clear_session()
agent = Agent(window_size, pretrained=True, model_name=model_name)
# read csv into dataframe
df = pd.read_csv(test_stock)
# df = df.iloc[:55]
# filter out the desired features
df = df[['Date', 'Adj Close']]
# rename feature column names
df = df.rename(columns={'Adj Close': 'actual', 'Date': 'date'})
# convert dates from object to DateTime type
dates = df['date']
dates = pd.to_datetime(dates, infer_datetime_format=True)
df['date'] = dates
df.head()
coloredlogs.install(level='DEBUG')
switch_k_backend_device()
test_data = get_stock_data(test_stock)
initial_offset = test_data[1] - test_data[0]
test_result, history = evaluate_model(agent, test_data, window_size, debug)
show_eval_result(model_name, test_result, initial_offset)
def visualize(df, history, title="trading session"):
# add history to dataframe
position = [history[0][0]] + [x[0] for x in history]
actions = ['HOLD'] + [x[1] for x in history]
df['position'] = position
df['action'] = actions
# specify y-axis scale for stock prices
scale = alt.Scale(domain=(min(min(df['actual']), min(df['position'])) - 50, max(max(df['actual']), max(df['position'])) + 50), clamp=True)
# plot a line chart for stock positions
actual = alt.Chart(df).mark_line(
color='green',
opacity=0.5
).encode(
x='date:T',
y=alt.Y('position', axis=alt.Axis(format='$.2f', title='Price'), scale=scale)
).interactive(
bind_y=False
)
# plot the BUY and SELL actions as points
points = alt.Chart(df).transform_filter(
alt.datum.action != 'HOLD'
).mark_point(
filled=True
).encode(
x=alt.X('date:T', axis=alt.Axis(title='Date')),
y=alt.Y('position', axis=alt.Axis(format='$.2f', title='Price'), scale=scale),
color='action'
).interactive(bind_y=False)
# merge the two charts
chart = alt.layer(actual, points, title=title).properties(height=300, width=1000)
return chart
chart = visualize(df, history, title=test_stock)
cap = [0]
inv = 0
ret = 0
b = 0
for i in range(len(df)):
if df.iloc[i]['action']=='BUY':
cap.append(cap[-1]+df.iloc[i]['actual'])
inv+=df.iloc[i]['actual']
b+=1
if df.iloc[i]['action']=='SELL' and b>0:
cap.append(cap[-1]-df.iloc[i]['actual'])
ret += df.iloc[i]['actual']
b-=1
req_cap = max(cap)
prof = ret+(df['action'].value_counts().get('BUY',0)-df['action'].value_counts().get('SELL',0))*df.iloc[-1]['actual']-inv
return pd.DataFrame([[ticker,strategy,ep_count,year,inv,ret,req_cap, prof, (prof/req_cap)*100]],columns=['ticker','strategy','ep_count','year','investment','returns','required capital','profit','profit percentage'])
