Q&A with PhD Student Harika Tuzcuoglu

Harika (Ozhalepli) Tuzcuoglu is a third-year PhD student. She received an MBA with a minor in finance from SUNY, Binghamton, and a Bachelor of Computer Engineering from Koç University in Turkey. The first student to fast-track from the MSc to the PhD, Harika was instrumental in the creation of an accelerated graduate research program at the Sprott School of Business.

What attracted you to Sprott’s PhD in Management program?

I love finance. I love numbers, and I love how money works. My previous degree—the MBA—was not research-based, though it does include a minor in finance. After completing my degree, I knew that I wanted to keep studying finance in a way that would draw on my MBA and engineering background. More specifically, I wanted to look into the role of technology in finance—FinTech. At that time, there weren’t many professors working in this area. I was able to narrow my search to just a few universities, and Sprott was one of them.

When exploring graduate programs, I read the research of Mohamed Al Guindy, who is now my supervisor, and was intrigued by it. I also found that whenever I reached out to program administrators at the various schools, I’d hear back from Sprott very quickly and positively. So that’s how I ended up here, but I started as a master’s student because I wasn’t sure if I could do a PhD.

During my first year, I loved all the courses I took. They were completely different from other courses I had taken. They taught me how to do research, quantitative analyses, and literature reviews, and how to read academic articles. The more time I spent at school, the more I liked being here. I decided I wanted to pursue a PhD, and all my professors supported my decision.

What kind of research did you do during your coursework?

Much of it focused on connecting Artificial Intelligence [AI] and Finance. For example, in one project, I analyzed 8 million tweets about Bitcoin to determine the impact of Twitter on Bitcoin price return and volatility. I began by cleaning the tweets to get them ready for natural language programming (NLP). Among other things, this process involved the removal of punctuation marks, stop words (words like “I am” and “the”), extra spaces, and numbers.

After performing NLP, I fed the machine the cleaned tweets and employed unsupervised topic modelling algorithms, which basically allow the machine to decide by itself how to cluster words. In one cluster, the tweets included emojis like the rocket and stars—symbols people were using to indicate that Bitcoin was skyrocketing. In another topic, the most frequent emojis included the angry red face and the clown, along with words like “short” and “alert.” The emojis in this topic were indicating that Bitcoin was doing badly. These results showed that emojis convey meaning because the algorithms were themselves able to meaningfully cluster the emojis along with the words.

For the Finance part of this project, I compared the Twitter data to cryptocurrency market data. I found that increased tweet volume predicted a rise in Bitcoin price volatility.

What is your current project?

It’s an investigation of how Bitcoin and cryptocurrency discussions have evolved over time. We have tens of millions of Bitcoin tweets, starting from the inception of Bitcoin in 2008, and going until today. I will employ unsupervised machine learning algorithms, such as topic modelling, to trace the evolution of Bitcoin topics on Twitter. I want to determine whether there are any trends around issues like Bitcoin investment, environmental concerns, regulatory challenges, and illegal activities.

I also want to analyze how cryptocurrency discussions relate to global events and impact Bitcoin price return and volatility. For example, let’s say my algorithm produced a topic for this month related to the security of Bitcoin blockchain, and I find that whenever this topic is being discussed, Bitcoin price volatility jumps. I would then tell investors, “Security is the trend topic right now, and I found an association between security and Bitcoin price volatility. You may want to be careful because your Bitcoin may be more volatile at this time.”

I’m very fortunate to be at Sprott and to have access to the Social Media and Finance Lab because running algorithms on datasets as large as mine is very computationally demanding. It actually requires a supercomputer, and Sprott is the only business school in Canada to have such computing power. Without access to this Lab, I wouldn’t be able to conduct this research.

Do you think you’ll find a connection between Twitter discussions and the markets, or between the topics and global events?

I may find a relationship or I may not. That’s the beauty of finance; it’s not really predictable. I just have to do my research and see.