One of the most commonly acknowledged facts about Bitcoin – and the broader cryptocurrency market – is the unpredictability of price fluctuations. Many have tried adapting and applying some of the charting techniques used to predict stock market price movements for the digital assets market, with varying degrees of accuracy. However, what about machine learning?
A 2nd December 2019 blog post published by Abinhav Sagar, a data scientist from India’s Vellore Institute of Technology, the focus is on exactly that question. Prior to demonstrating his four step process, Sagar made note of how the computing science of machine learning has been a little neglected in this field. However, he does reason that it likely a result of the numerous factors that cryptocurrency prices are affected by.
“Although machine learning has been successful in predicting stock market prices through a host of different time series models, its application in predicting cryptocurrency prices has been quite restrictive. The reason behind this is obvious as prices of cryptocurrencies depend on a lot of factors like technological progress, internal competition, pressure on the markets to deliver, economic problems, security issues, political factor etc.” Sagar wrote.
Sagar’s proposed prediction method is a four step process that works by making use of the LSTM neutral network (Long Short-term Memory).
The first step of Sagar’s process requires the collection of data in real-time (which Sagar sourced from CryptoCompare). The second step to prepare the data before feeding it into the neutral network. The third step is to test the give data on the LSTM network. Lastly, the prediction is visualised.
Sagar points out that the method used to measure the average magnitude of error in the prediction is mean absolute error. Following which, would be the process of plotting and then visualising the results of the test.
Get real time update about this post categories directly on your device, subscribe now.
