Practical Lessons we Learned While Building Cryptocurrency Price Predictions Using Deep Learning
Building predictive models for crypto assets is brutally difficult.
During the last few months, we at IntoTheBlock have embarked in an effort to build predictive models for crypto assets using deep neural networks. We failed a lot, learned a lot and were able to achieve some very promising results. Some of the initial predictive models will be soon available in the IntoTheBlock platform but, in the meantime, I would like to share some of the things we have learned starting with some very practical lessons.
Price predictions are, in many forms, the ultimate expression of crypto analytics. Many of the metrics and indicators that we use regularly are intended to help us formulate a predictive thesis about the market. Price prediction algorithms are just a programmatic way to abstract those theses into machine learning model.
Price Predictions for Crypto Assets in Three Steps
The process for creating deep learning models for crypto assets can be summarized in three easy steps:
1) What to predict?: Formulating a prediction thesis.
