import numpy as np
import matplotlib.pyplot as plt
import pandas as pd
from sklearn.preprocessing import MinMaxScaler
import tensorflow as tf
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense, Dropout, LSTM
from tensorflow.keras.models import load_model
import matplotlib.dates as mdates
import pandas_datareader.data as web
import datetime as dt
def predict_stock_trend(ticker, start, end, train_sample, train_epoch, forecast_days):
df = web.DataReader(ticker, 'yahoo', start, end)
data = df['Adj Close'].values
#print(data)
#scale input between 0, 1
scaler = MinMaxScaler(feature_range=(0, 1))
scaled_data = scaler.fit_transform(data.reshape(-1, 1))
#print(scaled_data)
#print(scaled_data.shape[0])
x = []
y = []
sample = train_sample
#y lags x by n samples
for i in range(sample, scaled_data.shape[0]):
x.append(scaled_data[i-sample:i, 0])
y.append(scaled_data[i, 0])
x, y = np.array(x), np.array(y)
x = np.reshape(x, (x.shape[0], x.shape[1], 1))
#print(x.shape)
#(1208, n, 1)
#print(x[0])
print(y.shape)
#print(y)
#train on GPU
pysical_devices = tf.config.experimental.list_physical_devices('GPU')
#print("Num GPUs Available: ", len(pysical_devices))
tf.config.experimental.set_memory_growth(pysical_devices[0], True)
model = Sequential()
model.add(LSTM(128, input_shape=(sample, 1), return_sequences=True))
model.add(Dropout(0.2))
model.add(LSTM(128, return_sequences=True))
model.add(Dropout(0.1))
model.add(LSTM(128, return_sequences=True))
model.add(Dropout(0.1))
model.add(LSTM(128))
model.add(Dropout(0.1))
model.add(Dense(1))
model.summary()
model.compile(
loss='mean_squared_error',
optimizer='adam',
)
model.fit(x,
y,
epochs=train_epoch,
batch_size=32)
model.save("stock_xsample.h5")
x_forecast = scaled_data.flatten()
#print(x_forecast.shape)
#print(x_forecast[-50:])
x_forecast = x_forecast[-sample:]
y_forecast = np.array([])
forecast_period = forecast_days
#print(x_forecast)
#model = load_model('stock_xsample.h5')
#assume sample = 50. use last 50 known stock price to predict the next one
#then use last 49 known stock price + previous predicted price to predict next one...
#loop for forecast peroid, all forecasted price are recorded in y_forecast
for i in range(0, forecast_period):
if i < sample:
x_forecast_new = np.append(x_forecast[i:], y_forecast)
else:
x_forecast_new = y_forecast[i-sample:]
x_forecast_reshaped = np.reshape(x_forecast_new, (1, sample, 1))
predicted_stock_price_single = model.predict(x_forecast_reshaped)
y_forecast = np.append(y_forecast, predicted_stock_price_single[0][0])
#print(y_forecast)
#print(predicted_stock_price.shape)
#print(predicted_stock_price)
y_forecast = np.reshape(y_forecast, (len(y_forecast), 1))
predicted_stock_price = scaler.inverse_transform(y_forecast)
predicted_stock_price = predicted_stock_price.flatten()
#print(predicted_stock_price)
ax = plt.figure(figsize=(7, 5), dpi=100).add_subplot(111)
#fig, ax = plt.subplots()
t = df['Adj Close'].index
t = pd.to_datetime(t)
#print(t)
#create xaxis for predicted stock price
t_predicted = pd.bdate_range(end.strftime('%Y-%m-%d'), periods=forecast_period, freq="D")
#print(t_predicted)
ax.plot(t_predicted, predicted_stock_price, label="Predicted stock price", color="red")
ax.plot(t, data, label="Real Stock Price", color="blue")
ax.legend()
ax.xaxis.set_major_locator(mdates.DayLocator(interval=300))
ax.xaxis.set_major_formatter(mdates.DateFormatter('%b %d %Y'))
ax.set_title(ticker + " Price Prediction")
plt.ylabel("USD")
ax.xaxis.grid(True, which="major")
ax.yaxis.grid(True, which="major")
plt.xticks(rotation=45)
plt.subplots_adjust(bottom=.2)
zoom_factory(ax)
plt.show()
# for mouse scroll zoom
def zoom_factory(ax,base_scale = 1.2):
def zoom_fun(event):
# get the current x and y limits
cur_xlim = ax.get_xlim()
cur_ylim = ax.get_ylim()
# set the range
cur_xrange = (cur_xlim[1] - cur_xlim[0])*.5
cur_yrange = (cur_ylim[1] - cur_ylim[0])*.5
xdata = event.xdata # get event x location
ydata = event.ydata # get event y location
if event.button == 'up':
# deal with zoom in
scale_factor = 1/base_scale
elif event.button == 'down':
# deal with zoom out
scale_factor = base_scale
else:
# deal with something that should never happen
scale_factor = 1
print (event.button)
# set new limits
ax.set_xlim([xdata - cur_xrange*scale_factor,
xdata + cur_xrange*scale_factor])
ax.set_ylim([ydata - cur_yrange*scale_factor,
ydata + cur_yrange*scale_factor])
ax.figure.canvas.draw_idle() # force re-draw the next time the GUI refreshes
fig = ax.get_figure() # get the figure of interest
# attach the call back
fig.canvas.mpl_connect('scroll_event',zoom_fun)
#return the function
return zoom_fun
#predict_stock_trend(ticker, start, end, train_sample, train_epoch, forecast_days)
predict_stock_trend("FDX", dt.datetime(2015, 1, 1), dt.datetime.today(), 50, 100, 90)
