Stock Price Prediction Using LSTM

Web scraping and stock price prediction model

Note:

The project was completed on 31/05/2022, and the data was collected on the same day.
The websites related to the web scraping section may have different structures in the future. Therefore, if the source code is run at a different time, some unexpected errors may occur.

Project Description

The project includes following sections:

1. Data Collection - Web Scraping

• Selected 100 stock symbols from HOSE (Ho Chi Minh Stock Exchange): https://quotes.vcbs.com.vn/a/exchange.html
• Scraped the closing prices of stock symbols for the years 2020, 2021, and 2022.

2. Data Pre-processing

• Normalised data using TimeSeriesScalerMeanVariance.
• Plotted the the stock price chart over the given years.

3. Train and Predict

• Trained the LSTM model using collected data.

• Predicted the first 4 months of 2022. The prediction was done in 2 ways:

  • Predict consecutive 4 months: Using the values from the end of 2021, predicted the stock price trend for the first 4 months of 2022. In other words, the predicted values were used to predict more values. For example, the predicted price on 01/01/2022 was used to predict the price on 02/01/2022.

  

  • Predict day by day: Using the values from the end of 2021 and the beginning of 2022, predicted the stock price day by day for the first 4 months of 2022. In other words, only the available values were used to predict new values.

  

Installation

Install Chrome browser (if not installed already)

Download and install Chrome from the homepage: https://www.google.com/chrome/

Download Chrome Driver for data collection

Method 1: Manually download hromedriver
Select and download the suitable ChromeDriver at: https://chromedriver.storage.googleapis.com/index.html?path=100.0.4896.60/

• Ensure Chromedriver is compatible with the operating system.
• Since this source code is run on Windows, chromedriver_win32.rar was downloaded.
• After successfully downloading the driver, unzip the file to get chromedriver.exe.
• Place chromedriver.exe in the same folder as this project.

With this method, create the webdriver using:

from selenium import webdriver
wd = webdriver.Chrome(executable_path='chromedriver.exe')

Note: If the following error occurs "SessionNotCreatedException: Message: session not created: This version of ChromeDriver only supports Chrome version 100" when running the code "webdriver.Chrome(executable_path='chromedriver.exe')", switch to method 2.

Method 2: Using WebDriverManager library
Install WebDriverManager using

pip install webdriver-manager

With this method, create the webdriver using:

from selenium import webdriver
from webdriver_manager.chrome import ChromeDriverManager
wd = webdriver.Chrome(ChromeDriverManager().install())

Library versions

Library	Version
numpy	1.19.5
pandas	1.1.5
matplotlib	3.3.4
selenium	3.141.0
sklearn	0.19.0
tslearn	0.5.2
keras	2.4.3

Files in repository

• stock_price.ipynb: The main Jupyter Notebook file of the project.
• allDailyStockPrice.csv: CSV file containing the closing prices of all stock symbols over the years.
• chromedriver.exe: Chromedriver for Windows (should be replaced if the code is run on another operating system)
• dailyStockPriceCSV folder:
  • Contains the CSV files with closing prices of every stock symbol.
  • stockHasLessData folder includes CSV files of other stock symbols that have less data than others.