arima.py

import requests
import pandas as pd
from bs4 import BeautifulSoup
import io   
import yfinance as yf
from datetime import datetime
import numpy as np
import statsmodels.api as sm

# import plotly.express as px
import matplotlib.pyplot as plt

# from statistics import covariance


import warnings
warnings.filterwarnings("ignore")

import statsmodels.api as sm
from sklearn.metrics import mean_squared_error
import pandas as pd

df = pd.read_csv('us-shareprices-daily.csv', sep=';')

def get_model_accuracy(data, ticker_symbol):
    
    stock_data = data[data['Ticker'] == ticker_symbol]


    # get MSE for testing data using 85/15 split for chosen stock symbol

    train_data, test_data = stock_data[0:int(len(stock_data)*0.85)], stock_data[int(len(stock_data)*0.85):]
    training_data = train_data['Close'].values
    test_data = test_data['Close'].values
    history = [x for x in training_data]
    model_predictions = []
    N_test_observations = len(test_data)
    for time_point in range(N_test_observations):
        model = sm.tsa.statespace.SARIMAX(history, order=(1,1,1))
        model_fit = model.fit(disp=0)
        output = model_fit.forecast()
        yhat = output[0]
        model_predictions.append(yhat)
        true_test_value = test_data[time_point]
        history.append(true_test_value)

    MSE_error = mean_squared_error(test_data, model_predictions)
    return 'Testing Mean Squared Error is {}'.format(MSE_error)


def main(tickers, earliest_date):
    df = pd.read_csv('data_and_sp500.csv')
    for ticker in tickers:
        x = np.array(df['Date'])
        y = np.array(df[ticker])
        ticker_df = pd.concat([df['Date'], df[ticker]], axis=1)

        model = sm.tsa.statespace.SARIMAX(ticker_df[ticker], order=(21,1,7))
        model_fit = model.fit(disp=-1)
        # print(model_fit.summary())
        forecast = model_fit.forecast(7, alpha=0.05)#.predict(start=1259, end=1289)

        data = pd.Series(y, x)

        # plt.plot(fit1.fittedvalues, marker="o", color="blue", label='smoothing')

        plt.plot(x, y, label='{} historical'.format(ticker))
        plt.plot(forecast, label='{} forecast'.format(ticker))
        plt.legend(loc="upper left")
        step = 10
        plt.xticks([x[i+step] for i in range(-step, len(x), step) if i+step < len(x)], rotation=90)

    plt.title('ARIMA forecast model vs. actual for {}'.format('ticker'))
    plt.xlim(df.shape[0]-100, df.shape[0]+21)
    plt.show()
    # plot_df = data.to_frame().reset_index().rename(columns={'index': 'date', 0: 'price'})
    # plot_df.columns = ['date', 'price']
    # print(plot_df)
    # fig = px.line(plot_df, x='date', y='price')
    # fig.show()


def stock_covariance(stocks):
    df = pd.read_csv('djia_2017-2022.csv')
    x = df[df['Name']==stocks[0]].dropna(how='any')['Close']
    y = df[df['Name']==stocks[0]].dropna(how='any')['Close']

    cov = covariance(x, y)
    return cov

def dji_covariance(stock):
    df = pd.read_csv('djia_2017-2022.csv')
    x = df[df['Name']=='^DJI'].dropna(how='any')['Close']
    y = df[df['Name']==stock].dropna(how='any')['Close']

    cov = covariance(x, y)
    return cov




if __name__ == "__main__":
    main(['AA', 'IBM', 'AAPL', 'AMD'], '2021-01-01')