STUDENT ENTRY

Student Entry Name - David Yossif

Submission Date - May 21. 2019

Topic Title - "AI Prediction"

​

​

AI and Machine Learning in the Stock Market

​

1. Introduction:

The stock market is one of the major driving forces behind the American economic market. Through the digitalization of the stock market over the past two decades there has been an increase in data and exchanges to massive amounts daily in the market. Naturally with these increases, big data along with data collection has become a major part of the modern stock market (Trippi). Right after big data has been introduced machine learning is inevitably introduced into that same field, and the stock market is no exception. One of the major applications is predicting how the stock market will perform, as a whole and per individual stock options. There are so many factors that go into a stock, such as quarterly financial results, company releases, mergers, and various announcements about organizations. Machine learning can potentially be a great aid in reducing these unpredictable factors. Additionally the further extension of machine learning, neural networks, can be used in forecasting and the improvement of risk assessment. Finally not only can investors utilize these power tools but so can the U.S. Securities and Exchange Commission in regulating the stock market and enforcing the proper rules.

​

2. The Problems that Machine Learning Aim to solve:

There are two main types of Analysis when working with the stock market: fundamental analysis and technical analysis. Fundamental analysis is all about looking at the current business position and financial standing of the company in its current state along with its history. Technical Analysis falls more into the analysis of physical mathematical charts and statics that have mathematical backing (Patel). Therefore it can be noted that some components of the fundamental analysis can have components that cannot be completely represented in tangible data that is ready to be inputted into a machine learning technique or model.

​

2a. Predicting the Future:

First and foremost the most important thing for an investor to is know how a stock will do in the future. So this is the obvious problem that machine learning can solve, know whether a stock price will rise. This can be done through a relatively simple binary classification. The binary part being whether there will be an increase or decrease. The complicated part about this problem is which points of data to train the model on. For instance we can use Natural Language Processing to use previous announcements and how they affected the previous stock prices. The main sub problems within this overall problem is not collecting data or lack thereof data, but which data to use (Patel). This is where machine learning will reach its limitations since it is up to the data scientists and engineers to decide which information is truly helping the prediction. If the too much information is passed we reach the problem of over fitting where the model is only accurate for the previous training data but not new data being passed in. There have many implementations of machine learning other than classification that have had limited success in the stock market. That is why research is still growing in this field due to the low success rates and the sheer difficulty of trying to in some sense predict the future. Another implementation is Averaging, where based on previous years we can average all those prices of a given stock and use the average as a prediction of the upcoming data. Linear Regression can also be applied in this situation where we assume the stock price points take on a form of a mathematical equation that is on a 2D plane (Patel). This curve will allow the model to predict the next data point of the stock.  

​

2b. Regulating:

In this problem the Acting Director of the regulation of the stock market defines AI in the financial sector as “Fintech.” However he then notes that the same technology that fuels the Fintech also applies to the new developing sector “Regtech” (Bauguess). This refers to the use of AI in the regulation of the stock market which a very important task that involves again massive amounts of data. This is due to the fact of millions of trades that can occur at any given moment within the stock market. This technology makes compliance and rule enforcement less labor intensive, more efficient, and much faster than the previous alternatives. A simple implementation of a simple rule is as follows: an investor can only make X amount of trades on a particular stock within a given amount of time.  This may seem like a simple enough rule to implement but when there are millions of trades occurring investors can try to break up their trading and hide within the masses. The introduction of neural networks is easily able to look at past cases where investors followed the rules and cases where infractions occurred (Kirkland). Then many machine with little bots are able to cover every trade that occurs within the given time period according to the rule. The use of the neural nets that progressive learns from the data and improves with each training case it sees is also able to parse through unquantifiable amounts of data that would previously result in breakage of rules going unpunished.

​

3. Fintech

Fintech refers to financial technology that has been developed in order to replace traditional financial methods that were used to deliver financial services. This services can include mobile banking, investing services where we see the major rise of AI and machine learning within this sector, and crypto currency.  Additionally there have been recent studies on assessing the potential of Fintech on the finance industry with specific focuses in AI and stability of services. Within these studies arguments have also been made that there will be an increase in costs to regulate and coordinate those economic regulations that can also be aided by AI (Philippon). The rise of Fintech is relatively new and will bring about many new changes to this field.