Machine learning and artificial intelligence are two of the trendy buzzwords that are getting thrown around in every domain and topic nowadays. But, to an extent, the powers of machine learning are real, and machine learning is being applied to solve different problems in ways that humans are not capable of. How does it work, and can it be used to predict things such as stock prices?
What Is Machine Learning?
Machine learning is a collective term for a series of different algorithms and techniques. When we talk about algorithms here, we refer to a sequence of instructions that can be programmed into a computer to make it perform a certain task.
The simplest algorithms found in computers may do things such as open a document when you click on it, turn up the sound when you press the sound button, or turn off your computer when it is overheating. Each of these tasks requires an algorithm for the computer to know what to do. If there is no algorithm to open a document, or if it has not been coded into the computer, then the computer cannot open the document.
Machine learning algorithms are specific algorithms that can be used to make a prediction or inference based upon some given data. One basic algorithm is the line of best fit, where a line is plotted through a series of points on a scatter graph, and the line should be as close as possible to all the points. This line can then be used to predict new values on the graph.
In machine learning, this would be known as regression. Existing data is collected, and then an equation is fitted on the data which minimises the error, or the distance of the equation from each point.
There are many other algorithms that are used in machine learning. This includes the decision tree, where a tree of questions is created and the answers branch off to further questions, eventually leading to an answer of prediction – feeding different data points into this tree would result in different predictions.
What Data Is Needed For A Machine Learning Algorithm?
The type of data needed for a machine learning algorithm depends on the algorithm being used. Different types of data can include numbers (integers or real numbers), words, pictures, audio, or Booleans (yes or no data).
For many machine learning algorithms, including regression, the data has to be converted into numbers. For types of data such as pictures and audio, they are often already stored as numbers in a computer. For types of data such as words, you can convert them into numbers by letting each unique word be represented by a different number.
In an example, the word “to” could become the number 4, the word “go” could become the number 11, and the word “bed” could become the number 6. The phrase “to go to bed” would then become “4 11 4 6” when converted to numbers. This method can be used to convert sentences or even whole articles or books into numbers.
How Does This Help Us Predict Stock Prices?
By collecting different types of data, data scientists are able to create machine learning models such as regression and decision tree models that attempt to predict the future prices of stocks. These predictions can then be used by investors and traders in their decision making.
The potential in these models is great, which is why teams of data scientists often work together to improve their predictions. The Jane Street Market Prediction competition on data science website Kaggle currently has over 3,000 data scientists working together and competing to predict future prices of real-world stocks. In this competition, data scientists are working with historical stock data to predict the values of future stocks. The winner of the competition is simply whoever can create the machine learning model with the best predictions.
Researchers have also been looking into what other data can be used to predict stock prices. Research at the K.J. Somaiya College of Engineering in Mumbai, India looked at using news articles to predict the future prices of Apple stock and was able to detect whether Apple stock prices would increase or decrease with up to 92% accuracy.
Similar research has been carried out by looking at social media. Researchers have been able to predict increases and decreases in stock prices by looking at whether conversations on social media platforms such as Twitter and Reddit have been positive or negative. Research in this field has been applied in large businesses for years. According to the Jane Street Market Prediction competition organisers, “Jane Street has spent decades developing their own trading models and machine learning solutions to identify profitable opportunities and quickly decide whether to execute trades. These models help Jane Street trade thousands of financial products each day across 200 trading venues around the world”.
It comes as no surprise that these investors want to use every tool at their disposal, considering the large amounts of money at stake in the stock market. The speed at which computers are also able to process data will allow investors to spot new and unseen areas of improvement in their trades. Investment in machine learning is not cheap though, and large corporations will need to invest heavily in the staff and technology required to make the optimal trades with split-second decisions. According to research by the CFA Institute in 2019, only 10% of portfolio managers were using artificial intelligence or machine learning techniques in their decision making. Most portfolio managers were still relying on tools such as Excel.
To summarise, machine learning techniques can be used to assist investors in their decision making. In fact, some investors are already using machine learning and artificial intelligence to help with their decisions, although research is still ongoing in many related areas. Overall, it looks as though machine learning will become an important tool for investors in the near future.
Comments