Due to the Coronavirus situation, Chinese government has lengthened Chinese New Year Holidays and further advised companies to have their employees work remotely. Considering the seriousness of the virus outbreak, like many other companies in China at the moment, BANCA team has continued its operation with all team members working from home. We expect to return to our office within one week. This will cause a little delay in the delivery of CoinShell but we still expect the alpha version to be completed and delivered in the first half of 2020.
When the modeling of our signals moves from single regression to multiple regression, problems surface about how to avoid over-fitting. The answer is that there will always be some level of over-fitting, but it is more about how we can reduce its effect. BANCA team reduces model over-fittings mainly through three ways:
First, assumptions should be made before model building. We need to come up with reasonable assumptions before flooding the model with too many factors.
Second, we can try to increase data amount used in the training. The more data we use, the harder it becomes for various factors to “perfectly” explain historical price movement.
Third, we can reduce the number of independent factors. More specifically, we can kick out factors that have low unique explanatory power over the dependent variable.
Apparently, in deep learning models, we also use other methods such as early stopping and dropout to fight over-fitting problem.
1. Development continues for CoinShell Alpha Version.
2. Testing of machine learning models continues for CoinShell Alpha Version and future versions.
3. We started to redesign some of the page layouts for CoinAI and CoinUltra.
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.