High Frequency Trading Strategy based on Deep Neural Networks
In this presentation, a high-frequency strategy using a Deep Multilayer Perceptron (DMLP), a type of Deep Neural Network (DNN), is presented. The input information to the DMLP consists of: (I). Current time (hour and minute); (II). The last n one-minute logarithmic pseudo-returns, where n is the sliding window size parameter and the one minute logarithmic pseudo-returns are computed as the logarithmic difference of two consecutive one minute average returns; (III). The last n one-minute standard deviations of the price; (IV). The last n trend indicator, computed as the slope of the linear model fitted using the transaction prices inside a particular minute. The DMLP output prediction is the next one-minute logarithmic pseudo-return that in turn allows to predict the next one-minute average price. This prediction is used to build a high-frequency trading strategy that buys (sells) at the beginning of a minute period if the current price is below (above) of the DMLP predicted next one-minute average price and closes the position either when predicted average price is reached or at the end of the minute
PhD Candidate, MBA + BSc Computer Engineering. 14 years’ experience across multiples industries/ sectors (Utilities, Telecommunications, Healthcare, IT, Manufacturing, Finance) in different positions including software development, project management, business analyst, as well as top management positions. I consider myself as highly reliable, very analytical and goal oriented. Nowadays, I am dedicated to research, design and develop Deep Learning Algorithmic Trading Strategies.