This page is a project description by Anatoliy Pushkarev for the thesis of the 2024-2025 years at the Innopolis University
The core idea is to develop a robust ai trading model, which generates 10% of profit every month, which is a basic professional trader benchmark.
I am planning to test three financial markets, which are:
What we are talking about here is a time-series forecasting task. There are many approaches how to set it up, but basically we need to put in the model some past prices and predict future prices (the data is TIME x PRICE).
Data may look like this:
https://www.kaggle.com/datasets/camnugent/sandp500
Above you can see daily stock prices for a company. Each row represents a price for a certain day (so it is 1 day timeframe). ‘Open’ column represents price at the beginning of the day, ‘close’ — end of the day, ‘high’ — highest value of the day, ‘low’ — lowest value of the day, ‘volume’ — trading value of the asset.
Also there should be another algorithm (ai or manual setup) which takes input and output data of the time-series model and then decides:
So the time-series model will predict price data, and the ‘decider’ algorithm will say which positions should we take.