Melbourne Science Hackfest

anomalyfinder

Anomaly detection and plotting them on the map

Sometimes, thing happen when we don't really expect them. It usually indicates that we don't have a full understanding of the process and can't fully predict it with the data we possess.

For example, we all know that bushfires occur in the dry areas when the temperature is extremely high for a prolonged period of time. But what if there is a fire that happened under completely different conditions? Should we assume it's an error in the record and there is no fire? Maybe meteostation in the area isn't functioning correctly and recording temperatures incorrectly? Or we're simply missing important factor that effects bushfires?  Or maybe the origin of the fire is not natural? Whatever that is, simply ignoring this point and discarding it as an error without further investigation may severely compromise our understanding of the nature of the process. 

People who collect the data are not the same people that analyse it. As the result anomalies on data are often left uninvestigated or even simply discarded as error readings.

We offer an easy to use 'no-coding-required' tool that will allow researchers who are not familiar with statistics to find anomalies, and maps them geographically encouraging further investigation into the reasons of certain anomaly.

Github