Untitled

import os
import pandas as pd
import matplotlib.pyplot as plt
import matplotlib.patches as patches
import seaborn as sns
import warnings
import numpy as np
from numpy import array
from importlib import reload # to reload modules if we made changes to them without restarting kernel
from sklearn.naive_bayes import GaussianNB
from xgboost import XGBClassifier # for features importance
warnings.filterwarnings('ignore')
plt.rcParams['figure.dpi'] = 227 # native screen dpi for my computer
import statsmodels.api as sm
from statsmodels.tsa.arima_model import ARIMA
from statsmodels.tsa.statespace.sarimax import SARIMAX
from statsmodels.graphics.tsaplots import plot_pacf, plot_acf
from sklearn.metrics import mean_squared_error, confusion_matrix, f1_score, accuracy_score
from pandas.plotting import autocorrelation_plot
import tensorflow.keras as keras
from tensorflow.python.keras.optimizer_v2 import rmsprop
from functools import partial
from tensorflow.keras import optimizers
from tensorflow.keras.models import Sequential, Model
from tensorflow.keras.layers import Input, Flatten, TimeDistributed, LSTM, Dense, Bidirectional, Dropout, ConvLSTM2D, Conv1D, GlobalMaxPooling1D, MaxPooling1D, Convolution1D, BatchNormalization, LeakyReLU
from bayes_opt import BayesianOptimization
from tensorflow.keras.utils import plot_model
import functions
import plotting
np.random.seed(66)
files = os.listdir('data/stocks')
stocks = {}
for file in files:
    if file.split('.')[1] == 'csv':
        name = file.split('.')[0]
        stocks[name] = pd.read_csv('data/stocks/'+file, index_col='Date')
        stocks[name].index = pd.to_datetime(stocks[name].index)
def baseline_model(stock):
    baseline_predictions = np.random.randint(0, 2, len(stock))
    accuracy = accuracy_score(functions.binary(stock), baseline_predictions)
    return accuracy
baseline_accuracy = baseline_model(stocks['tsla'].Return)
print('Baseline model accuracy: {:.1f}%'.format(baseline_accuracy * 100))
base_preds = []
for i in range(1000):
    base_preds.append(baseline_model(stocks['tsla'].Return))
    
plt.figure(figsize=(16,6))
plt.style.use('seaborn-whitegrid')
plt.hist(base_preds, bins=50, facecolor='#4ac2fb')
plt.title('Baseline Model Accuracy', fontSize=15)
plt.axvline(np.array(base_preds).mean(), c='k', ls='--', lw=2)
plt.show()
print('Tesla historical data contains {} entries'.format(stocks['tsla'].shape[0]))
stocks['tsla'][['Return']].head()
plt.rcParams['figure.figsize'] = (16, 3)
plot_acf(stocks['tsla'].Return, lags=range(300))
plt.show()
orders = [(0,0,0),(1,0,0),(0,1,0),(0,0,1),(1,1,0)]
train = list(stocks['tsla']['Return'][1000:1900].values)
test = list(stocks['tsla']['Return'][1900:2300].values)
all_predictions = {}
for order in orders:
    
    try:
        history = train.copy()
        order_predictions = []
        
        for i in range(len(test)):
            
            model = ARIMA(history, order=order) # defining ARIMA model
            model_fit = model.fit(disp=0) # fitting model
            y_hat = model_fit.forecast() # predicting 'return'
            order_predictions.append(y_hat[0][0]) # first element ([0][0]) is a prediction
            history.append(test[i]) # simply adding following day 'return' value to the model    
            print('Prediction: {} of {}'.format(i+1,len(test)), end='\r')
        
        accuracy = accuracy_score( 
            functions.binary(test), 
            functions.binary(order_predictions) 
        )        
        print('                             ', end='\r')
        print('{} - {:.1f}% accuracy'.format(order, round(accuracy, 3)*100), end='\n')
        all_predictions[order] = order_predictions
    
    except:
        print(order, '<== Wrong Order', end='\n')
        pass
fig = plt.figure(figsize=(16,4))
plt.plot(test, label='Test', color='#4ac2fb')
plt.plot(all_predictions[(0,1,0)], label='Predictions', color='#ff4e97')
plt.legend(frameon=True, loc=1, ncol=1, fontsize=10, borderpad=.6)
plt.title('Arima Predictions', fontSize=15)
plt.xlabel('Days', fontSize=13)
plt.ylabel('Returns', fontSize=13)
plt.annotate('',
             xy=(15, 0.05), 
             xytext=(150, .2), 
             fontsize=10, 
             arrowprops={'width':0.4,'headwidth':7,'color':'#333333'}
            )
ax = fig.add_subplot(1, 1, 1)
rect = patches.Rectangle((0,-.05), 30, .1, ls='--', lw=2, facecolor='y', edgecolor='k', alpha=.5)
ax.add_patch(rect)
plt.axes([.25, 1, .2, .5])
plt.plot(test[:30], color='#4ac2fb')
plt.plot(all_predictions[(0,1,0)][:30], color='#ff4e97')
plt.tick_params(axis='both', labelbottom=False, labelleft=False)
plt.title('Lag')
plt.show()
plt.figure(figsize=(16,5))
plt.hist(stocks['tsla'][1900:2300].reset_index().Return, bins=20, label='True', facecolor='#4ac2fb')
plt.hist(all_predictions[(0,1,0)], bins=20, label='Predicted', facecolor='#ff4e97', alpha=.7)
plt.axvline(0, c='k', ls='--')
plt.title('ARIMA True vs Predicted Values Distribution', fontSize=15)
plt.legend(frameon=True, loc=1, ncol=1, fontsize=10, borderpad=.6)
plt.show()
tesla_headlines = pd.read_csv('data/tesla_headlines.csv', index_col='Date')
tesla = stocks['tsla'].join(tesla_headlines.groupby('Date').mean().Sentiment)
tesla.fillna(0, inplace=True)
plt.style.use('seaborn-whitegrid')
plt.figure(figsize=(16,6))
plt.plot(tesla.loc['2019-01-10':'2019-09-05'].Sentiment.shift(1), c='#3588cf', label='News Sentiment')
plt.plot(tesla.loc['2019-01-10':'2019-09-05'].Return, c='#ff4e97', label='Return')
plt.legend(frameon=True, fancybox=True, framealpha=.9, loc=1)
plt.title('Tesla News Sentiment and Daily Return', fontSize=15)
plt.show()
pd.DataFrame({
    'Sentiment': tesla.loc['2019-01-10':'2019-09-05'].Sentiment.shift(1), 
    'Return': tesla.loc['2019-01-10':'2019-09-05'].Return}).corr()
scaled_tsla = functions.scale(stocks['tsla'], scale=(0,1))
X = scaled_tsla[:-1]
y = stocks['tsla'].Return.shift(-1)[:-1]
xgb = XGBClassifier()
xgb.fit(X[1500:], y[1500:])
n_steps = 21
scaled_tsla = functions.scale(stocks['tsla'], scale=(0,1))
X_train, \
y_train, \
X_test, \
y_test = functions.split_sequences(
                        
    scaled_tsla.to_numpy()[:-1], 
    stocks['tsla'].Return.shift(-1).to_numpy()[:-1], 
    n_steps, 
    split=True, 
    ratio=0.8