Sentiment analysis not quite there yet
Treasury has been working with the CSIRO, along with the Department of Human Services and other government agencies and companies, to pilot a social media monitoring tool called'Vizie'.
Alexander hopes to gain further insights into how Treasury can mine and use social media to create useful information on the cost of living, for example. However, this hasn't proved to be effective so far.
"We got some interesting results from that, but they were a bit limited. We looked if it correlated against other metrics that we have, but it didn't always. It looked like the sentiment was saying something was happening, but it just wasn't happening.
"Whether that means there's something in the measures, or what people say on social media has no relevance to reality. We couldn't draw any conclusions because we just don't know."
Big data and machine learning a possibility
Alexander is looking at how Treasury could use sensor data and see if it can provide useful information that relates to economic situations.
"It's about looking at massive unstructured data sets like being able to analysis out of utilities, or get a machine to run analysis over utilities, or traffic light information, crime statistics and come up with some correlation that tells you something interesting," he said.
"We are still pretty early days in that kind of space, but we are looking around [in the market] at the moment. But there's not a lot out there. There are lots of vendors that say 'I've got this fantastic big data tool', but really what they are trying to sell you is their business analytics tool.
"The fact that it can run over a gigantic data set is great, but it's not going to do that machine learning, correlation of disparate and massive data sets," he said.
Source: CIO Australia
Sign up for Computerworld eNewsletters.