Today, Modulus, a US-based developer of ultra-high-performance trading and surveillance technology that powers global equities, derivatives, and cryptocurrency exchanges, launched an introductory white paper explaining a revolutionary new predictive modeling algorithm for Bitcoin.
“Predictive modeling of financial time series data, in general, is complex due to volatility, but data regarding cryptocurrencies are even more complex to capture since traders don’t rely on a consistent source of technical indicators, as traditional financial traders would. After years of research, employing artificial neural networks, we now have data that shows the Modulus algorithm to be successful in predicting a binary change in, and future prices of, cryptocurrencies.”
– Richard Gardner, CEO of Modulus
“While some naive resources may recommend the use of simple, linear forecasting methods, financial data rarely meet the assumptions of such statistical models. Thus, linear modeling techniques, such as ARIMA, should be avoided,” explained Gardner. In 2014, research conducted by Oancea and Ciucu showed empirically how Long-Short Term Memory (LSTM) neural network models perform better than feedforward neural networks for cryptocurrency price forecasting.
The Modulus algorithm utilizes uniform distributions based on the number of neurons within a layer, rather than using default weight randomization processes for neural networks. “While model architecture is a key component in predicting cryptocurrency fluctuations, the selection of quality input data is just as, if not more, important. Modulus has decades of experience in financial trading, including with cryptocurrencies, that give us the upper hand in meticulously selecting data sources that are relevant to price fluctuations in the cryptocurrency market.”
Modulus is known throughout the financial technology segment as a leader in the development of ultra-high frequency trading systems and exchanges. Over the past twenty years, the company has built a client list which includes Goldman Sachs, Merrill Lynch, JP Morgan Chase, Bank of America, Barclays, NASA, Siemens, Shell, Yahoo!, Microsoft, Cornell University, and the University of Chicago. Among other data collected, Modulus gathers blockchain details fromblockchain.com, website traffic data from Amazon Web Services, and relevant keyword information from Google Trends.
