Chris’s Domain

I am slowly, but progressing through my final session in university steadily. I have just handed in my final Honours thesis, a culmination of all the hard work I’ve put in the past 6 months (really, it’s the last few weeks of experiments that really boost some of my results). Now that it has been handed in and can be publicised, I’m allowed to lift the shroud and reveal some broad concepts behind it.

The research centres around the idea of building a sentiment detection engine for internet stock message boards. There has been limited studies on this in the US, but this is the first time a project of such a scale has been attempted on an Australian source. Gathering a vast amount of data from HotCopper/Sharescene and the like, the aim is to predict the general market sentiment of a particular stock and apply this to financial analytics. This is useful in 2 areas, particularly in technical trading strategies where it’s a source of new information, perhaps using sentiment as a factor in regression studies to crowd-source information rather than relying on traditional media; it is also useful in surveillance where analysts (like those guys working at ASX and ASIC) can use the sentiment prediction to gauge market reaction and possibly explain spikes in prices so they can be further investigated. Online forums contain so much content (a lot of noise too, but if we can filter this, we can mine a lot of useful information) where there is ample research opportunities to leverage this information, both for market surveillance as well as to profit from. Human analysts can only process so much information in a given time period, but with high frequency market executions, there is now a greater need for advanced, intelligent analytic tools more than ever.