Thesis project on text analysis to help predicting stock peaks & falls

Stocks are about public perception of the value of a company and thousands of websites are tracking those companies & the products. I hope to teach my algorithm to go through them and make decisions for me.

I am on my final year in the university and I am doing my thesis. I am thinking about writing vigorously about how my research is progressing and how I am trying to do it. So, far I have only done basics.

Let’s be clear. I don’t have much knowledge on stock market in any way. But I find it pretty interesting. Last year, I played Fantasy Premiere League. And it seemed fun trying to predict the outcome of the matches and score points. I always had a fasination about stock market and how it works. Even if my thesis is kind of a failure, I guess still I get a pretty good chance to know more about it.

Anyway, so initially my super brilliant idea was this — Everything (well, almost) about a publicly traded company is a mass interest. Blogs, newspapers, stock analysis websites extensively track these companies. So, I am thinking, what if a system could crawl through all these and judging by those informations decide what to buy and what not.

Now that’s very generic. Obviously the system has to work on many fronts. I have thought about some approaches.
- Building a graph of products. Each node would contain information about who owns the products, the competitor of the products.
- Rating each of the individual components i.e the products, owner, competitor.
- Crawling through the news, blogs, and publicly available information if any of the objects gets affected and that would impact their rating.
- Making the decision to buy and sell.

It feels pretty advanced, and truth be told I am kind of out of my depth here.

But let’s so how it goes from here now. Cheers.