AI Trader analyses historical data from the past months on multiple time frames depending on the AI Attitude (Cheetah, Zebra, Elephant). After rigorous tests, our team identified the optimal amount of data to provide the training set for each AI Attitude. This was imperative, as overfitting or underfitting of data can vastly affect the performance of machine learning.
The weight allocated to neural networks is divided as follows: 85% Price, 10% Volume, 5% Indicator data (MACD, RSI, STOCHASTICS). We continuously run models on the live market on a 24 hours basis, which allows the machine learning pattern to change and adapt to the fluctuating market conditions.