Abinhav Sagar, a data scientist at Vellore Institute of Technology in India, has developed an algorithm that could help predict cryptocurrency prices in real time by use of machine learning, Cointelegraph reports on November 2nd.
Sagar attempted to open up this new area of artificial intelligence application, which by now been neglected due to the volatility on the underlying market. By utilizing a Long Short-Term Memory (LSTM) neural network, Sagar fed specially prepared data sets based on CryptoCompare's price and volume histories. The results from the model were then visualized alongside actual price movements and compared.
While machine learning and artificial intelligence already found their application in predicting prices and devising trading strategies for the more traditional assets, cryptocurrency markets remain elusive due to the high volatility, faked data and market manipulations. It is interesting to compare the results of Sagar's model with actual price movement. The easy to spot deviations could be the result of the model's imperfections, but also the reflection of the irregularities in the crypto market. It is those irregularities that "shake out" most of the day traders but also what makes trading on the market potentially lucrative. Once a fully working prediction model is established, those irregularities should decline or already be accounted for and it remains to be seen what potential impact such a model could have on the market.