In making predictions on crypto prices using LSTM (Long Short Term Memory) model, we have tried various methodologies and it seems to have the following conclusions:
1. The nature of the model is time series analysis (such as Autoregressive Model ARt, where t = length of training set), but more complicated than AR model is that we are using CHLOV (close, high, low, open, volume) data instead of simply closing prices;
2. When we lengthen the time span of training set, the result of the model will still change but converge when getting over a certain point, because market prices too far back apparently have fewer relevant information;
3. On server-end deployment, we need to store much more historical data than traditional machine learning models;
In order to deliver alpha version faster, Banca team decide we are not going to include the LSTM deep learning model in our alpha version. We will take more time to test our assumptions in it and deliver it later versions.
1. BANCA team continues to work on front-end and back-end development of CoinShell alpha version.
2. We have completed testing of most simple trading signals, initial parameters have been calculated. Now the scoring system is being developed.
The BANCA platform serves the global cryptocurrency community. BANCA’s dynamic eco-chain uses AI and expert system that includes automatic management. The BANCA platform analyzes Big Data and delivers precise services tailored to the specific needs of our individual users.
Telegram channel: https://t.me/BancaPublic