Linux operating system can be used to make predictions on when someone is likely to buy a security stock.
Researchers at the University of Waterloo and IBM have developed a machine learning algorithm to predict when a user will buy stocks, and what the chances of them doing so will be.
“In the real world, it’s very easy to buy the right thing, but you can’t predict what that will be,” said Daniel Lutz, an associate professor at the university and the principal investigator of the new study.
“But with a machine-learning model, you can.”
The results are published in the journal Scientific Reports, and they reveal a new kind of machine learning model that can identify people who are likely to spend money on a stock.
The computer, which can be programmed to predict which stock an investor is most likely to invest in, is called a machine.
The researchers developed the machine learning system by studying data from about 30 million stock purchases over the past two years, Lutz said.
The algorithm used this data to generate three different versions of its algorithm: one that uses a single point of view on people who invest in stocks, another that uses two points of view, and another that only uses one point of perspective.
“The first version that we created, which was used by the system to predict how much an investor would invest, was very good at predicting when people were likely to be willing to spend,” Lutz told CBC News.
“It predicted that they would be more likely to make that investment if they were a female or white male.”
Lutz said the second version of the algorithm had a very different set of biases that were likely the result of a more nuanced view of people who might invest in stock, such as gender, age, race, or income.
“This system also had a better prediction about who would be able to buy,” Luts said.
“That is very interesting, and it’s going to give us a better understanding of what the human mind is capable of.”
Luts said the data collected by the machine was similar to the information that would be collected about the people that people would buy from the bank that they used to create the prediction.
“We know that people tend to spend their money when they’re excited about something and when they feel they have the ability to do something that they might otherwise not be able do,” he said.
The new model was able to predict more than 100% of the time that a person would buy a financial product when they were just one step away from doing so.
“When you’re trying to predict the behavior of a particular person, the machine can be trained to predict about 100% accuracy,” Lutts said.IBM said the work was an important step towards creating a better machine learning program that could better predict the actions of people in the real-world.
“Using machine learning to predict human behavior is a huge challenge for any artificial intelligence system,” said Jim Wunderman, vice president of advanced research at IBM Watson Research, in a statement.
“But these results demonstrate the power of machine-vision technology for predictive models of human behavior.”
The researchers are currently working to improve their system so that it can also predict what kind of behavior a person might have when they are in a certain position, such a sitting position.
“These are all things that could have been improved by training our system to recognize when a human would sit down or take a drink,” said Lutz.
Lutz, who is also a research fellow at the IBM Watson Institute for AI and Cognitive Science, said his machine learning work will be closely followed by the University’s research group, which is currently investigating ways to better understand how humans use artificial intelligence and artificial learning.
“Our machine-knowledge system will have to evolve to be able better predict what kinds of things people are interested in, and the machine will have a better ability to understand what those are,” he added.