Skip to main content

AI for Fintech – Using AI to predict stock market trends

So why develop AI software to predict stock market trends?

The environment of the stock market has an aura of mystery, risk, and excitement. For many years traders and brokers have always been adapting and forming trading strategies to get an edge on the market place.

Nowadays, due to the revelations of technology over the past two decades; there has been a wide array of innovations in the space. These innovations range from revolutionary advancements in circuit breakers to getting the latest news and the option to trade on smart phones. This has provided much needed stability for the buying and selling infrastructure, but what about risk analysis?

This is where AI Fintech software comes in. It has always been difficult for investors to maximise profit and mitigate loss. This is what piqued my interest, so I decided to investigate the problem further to gain an understanding of the motivators behind stock fluctuations. This would provide a solid foundation to work from in relation to risk analysis.

Finding data for AI Software to analyse

The power of social media

This area of data is particularly important, it provides insight into the sentiments of people; and as many of us know stock market fluctuations are directly proportional to human behaviour. This data can be used to analyse reactions to business news and company announcements. Social media data is pivotal in establishing a correlation with directions and trends of the stock market.

The importance of business news

Many traders would be actively monitoring business news networks such as Bloomberg or Yahoo finance to anticipate how the market would react to announcements. Business news networks have one purpose that they need to be good at to survive; and that is capturing events as they happen. In the context of training Fintech AI software to analyse this data, it would be very easy to capture how business news announcements have a positive or negative impact on stock prices.

Creating a hybrid dataset

The next step was to pick a company and analyse historical events that influenced stock prices. Who was chosen? Microsoft…naturally ????. There is always something interesting going on at Microsoft and there is no shortage of information that could be garnered from social media and business news networks.

Mapping the data for stock market prediction

The next stage was to select KPI’s to work with to analyse and benchmark. In the context of financial indicators, information in relation to stock price on open, high, low, and close was garnered. This information was useful for mapping out volatility over a timeline of events.

Natural language processing was then utilised to collect information from financial news, an example of how this was used; was gathering data from Twitter via its hashtag function. Topic based analysis of the Twitter data and how prices were impacted based on the context of tweets gave clear indication of the influence of public sentiment had over the increase or decrease of stock prices.

How accurate was the AI software once built and trained?

Based on the historical data gathered, the AI model was a whopping 72% accurate. Although this model could contribute towards reducing risky investments, one would need to note that a big data set is required to get there. I had a lot of fun constructing this platform to determine the feasibility of predicting stocks. More can be found in my research dissertation I conducted with CIT.

Next steps…

Utilising AI to solve problems is always an interesting area. Over the past 12 – 14 months Spéire has embraced the era of Artificial Intelligence. The potential for solving problems is huge, we are committing our time to develop AI solutions to help businesses innovate and keep up with the technological revolution of Artificial Intelligence. Learn more here or e-mail