Machine learning algorithms in trading

Abstract

Algorithmic trading involves the use of electronic platforms for all the trading operations such as entering trading orders where different trading aspects of the order such as timing, price, quantity etc. are decided based on an algorithm. It also at times involves initiating the order without any human interference or intervention. Algorithmic trading is technologydriven mode of trading. The algorithms are essentially based on mathematical models. These advanced and highly sophisticated models are used for making trading decisions in the financial markets. The models decide the optimum time of the trade so as to minimize the impact of the order on the stock’s price. Attempts are directed towards a time that would cause the least impact. The large orders are broken into child orders of small sizes. The orders for these small share blocks are given at the optimal time decided by the algorithm and the models. The division into these smaller blocks, not necessarily of equal size, is also done by the algorithm. This mode of trading is most commonly used by large institutional investors such as banks, pension funds, hedge funds, etc. as they have large amount of volumes traded every day. This justifies the initial technology cost required to setup algorithmic trading platforms. Use of complex algorithms enables these institutions to extract the optimum prices for the trades without significantly increasing the transaction costs associated with such trades. A very special class of trading, known as “high frequency trading” is very popular in hedge funds. Here sophisticated algorithms are used to initiate orders based on available information in fractions of a second. The highly optimized complicated algorithms are devised mainly to make use of any market inefficiency or arbitrage opportunity to make profits. Algorithmic trading can be used in market making, arbitrage, speculation or any other investment strategy. The placing of orders may be completely automated or may be augmented with manual support.