PROJECT GUIDE - 1999michael/AI-Finance-Project GitHub Wiki

Project Guide

Below we discuss how the workflow of our software, and how to run it.


Data Processing

1.0 Collecting Data

1.1 Extracting and Cleaning Data

1.2 Creating Variables

This step consists of taking the cleaned data from the .txt file containing all raw data points, and converting it to a .json file containing variables to be inputted (upon request) to our neural network. We currently have 22 variables, consisting of 8 ratios from financial statements, and 15 data points from stock prices and stock price indicators. Variable explanations and their relevant indexes can be found on this page ____.

To add new variables, we create a new function that follows the guidelines of ____ and running the function _____ that will take the .txt file and the function as input, and update the .json file containing new variables.

1.3 Data Formatting

Not to be confused with step 3, this step consists of taking the requested variables (up to 22) and format the data into items to be passed into the neural network for training. This step consists of initializing the data_collect class of the users choice, and running .format_data(True, filename). Explanation of the initialization parameters can be found on this page ____.

This concludes the explanation on data processing


Neural Network Training

Using an LSTM model (found in ___ file) and train the model until the training and validation accuracy and losses have plateaued. To run the LSTM model, we call ___ function, and run model.train(), with parameters explained on this page ___.


Investment Simulator

3.0 Initializing the Simulator Class

Call the class investment_account with initialization parameter explained on this wiki page ___.

3.1 Running the Simulation

The simulation consists of investing once per week, in the top num_top_stocks stocks you initialized the investment_account class with. It will buy, then sell on that same day

Assumptions

  • You are guaranteed to get what you demand, i.e. the stock is always instantly bought/sold.
  • You can sell/buy your stocks at the closing price of that day

3.2 Damage Mitigation Techniques

Some stocks are bound to not increase despite our model predicting an increase a week in the future. This section discusses counter measures to prevent / minimize loss when a false positive has occurred.

3.2.1 Holding Stocks

⚠️ **GitHub.com Fallback** ⚠️