Researchers from the University of California, Riverside and Yahoo Research Barcelona have devised a model that uses data about Twitter volumes to predict how financial markets will behave. Their model bested other baseline strategies by 1.4% to 11% and outperformed the Dow Jones Industrial Average during a four-month simulation.
"These findings have the potential to have a big impact on market investors," said Vagelis Hristidis, an associate professor at the Bourns College of Engineering. "With so much data available from social media, many investors are looking to sort it out and profit from it."
The research, focused on what Twitter volumes, retweets and who is doing the tweeting might say about individual stocks, differs from that of earlier work focused on making sense of the broader market based on positive and negative sentiments in tweets.
As with so many stock-picking techniques, the researchers here tossed out plenty of caveats about their system, which they said might work quite differently, for example, during a period of overall market growth rather than the down market that their research focused on.
University of Texas, Dallas scientists have developed software dubbed Frankenstein that's designed to be even more monstrous than the worst malware in the wild so that such threats can be understood better and defended against. Frankenstein can disguise itself as it swipes and messes with data, and could be used as a cover for a virus or other malware by stitching together pieces of such data to avoid antivirus detection methods.
"[Mary] Shelley's story [about Dr. Frankenstein and his monster] is an example of a horror that can result from science, and similarly, we intend our creation as a warning that we need better detections for these types of intrusions," said Kevin Hamlen, associate professor of computer science at UT Dallas who created the software, along with doctoral student Vishwath Mohan. "Criminals may already know how to create this kind of software, so we examined the science behind the danger this represents, in hopes of creating countermeasures."
Such countermeasures might include infiltrating terrorist computer networks, the researchers say. To date, they've used the NSF and Air Force Office of Scientific Research-funded technology on benign algorithms, not any production systems.