Home / Students / Machine Learning Based Approach for High Frequency Trading

Machine Learning Based Approach for High Frequency Trading

With tremendous progress in technology, there have been different attempts to apply machine learning in various fields. Machine learning is a subfield of computer science that draws models and methods from statistics, algorithms, computational complexity, artificial intelligence, control theory, and a variety of other disciplines. Algorithmic approach is used to study the market, including holding periods, order types (e.g. passive versus aggressive), and strategies (momentum or reversion, directional or liquidity provision, etc.). This details every order placed, every execution, and every cancellation, and that thus permits the faithful reconstruction (at least for equities) of the full limit order book, both historically and in real time.

 

 

HIGH FREQUENCY TRADING

Earlier Stock market had a large screen showing change in rates and a hectic crowd of traders with babble of voices to buy and sell shares but this picture has changed with time, traders are replaced by computers and these computers are quick with trading programs. To make this possible HFT is split second faster than the other trading means and that is only possible as they are closed to stock market either by shorter optical fibre cable or through super fast broadband connection which gives them an edge over their competitors. For example- if price of a certain commodity starts increasing constantly, since High frequency trading programs can process this information before, they have the opportunity to first respond to the price change and can earn huge profits. This resulted in High frequency trading. High Frequency Trading constitutes of a large portion of stock market trading. HFT firms use computerized algorithms, based on mathematical formulas for proprietary trading, and they engage in electronic market making, cross-trading, venue price arbitrage, short-term statistical arbitrage, and various other opportunistic strategies. HFT firms operates across stock market, their activity has the potential to introduce a systematic component to the trading environment.

 

As firms have a goal of reducing expenditure, a well studied machine learning method is applied, known as reinforcement learning. Reinforcement Learning has its roots in the older field of control theory. It is a branch of machine learning for dynamic state-based policies from data. Reinforcement Learning is about how software agents take actions automatically, predict target values so as to maximize the reward.

Rather than seeking to reduce costs for executing a given trade, it would be more likely if models run themselves profitably and decide when to trade (that is, under what conditions in a given state space) and how to trade (that is, in which direction and with what orders), for alpha generation purposes.

With the market share of Automated Trading increasing, the regulators have been concerned with the reliability and efficiency in the markets. Regulations aimed at moving more derivatives trading into exchanges are seen as compounding the opportunities for electronic trading. High- speed decision making, arbitrage and speculation are enhanced by automated trading. But it is important to understand that automated trading is a way to trade efficiently, and not a trading strategy by itself. History has shown that those, who innovate and adapt in the face of change, those who are able to invest and integrate new technology are more likely to come out ahead in long run.

 

Cons of HFT

  • Normal traders cannot match to the speed of High Frequency traders and for this high frequency traders have to pay not only explicit stock market fees but also have to add up with Implicit HFT markup.
  • There are situations when two or more programs repeatedly react one another and subsequently trap into a loop, the problem is the speed with which it happens, only special programs will be able to solve this problem as humans are way slower to access it and intervene.

 

SEBI ON ALGORITHM TRADING
In India, there are various stock broking firms under investigations which have used dark fibre network to connect to the co-location facility of NSE. These lines are controlled by the users rather than service provider.

Algorithm trading can lead to wild upswings, so to curb that SEBI has established certain regulatory guidelines that are:

  • Price quoted in order should not violate defined price of bonds,
  • All orders should be routed by broker services available in India
  • Stock exchanges need to have mechanisms so as to identify and read out malfunctioning algorithms.

 

 

-KRATIKA KARAMCHANDANI (Analytics and Finance)

About Admin

Avatar

Check Also

Five reasons you need an MBA after engineering

Today, an MBA has become the peak of education. Without an MBA, you often run ...

Leave a Reply

Your email address will not be published. Required fields are marked *