Hello NuPIC, Here is a really easy way to experiment with NuPIC...
If you have NuPIC installed, you can run this simple script to see anomalies in any scalar stream of data in River View. Here's an example. > git clone https://github.com/nupic-community/river-runner.git > cd river-runner > ./run.py --plot --river=mn-traffic-sensors --stream=T3561 --field=occupancy This will run 5000 points of data from this data stream through NuPIC: http://data.numenta.org/mn-traffic-sensors/T3561/data.html Be advised, you must have matplotlib installed for the `--plot` option to work. If you leave it out, the anomalies will be written to a file that you can plot however you wish. If you let this run out, you'll see a high anomaly likelihood on Tuesday, Oct 6th: https://gyazo.com/c40dadcd17c862bc028629c64d7020d6 Although I can't find a historical "traffic incident" data feed, if you look at this twitter stream https://twitter.com/MSP_Traffic you'll see that there were quite a few traffic incidents happening on that day. Extra points if anyone can actually find an traffic incident that happened in the afternoon rush hour of Oct 6 nearby this traffic sensor: http://data.numenta.org/mn-traffic-sensors/T3561/meta.html Regards, --------- Matt Taylor OS Community Flag-Bearer Numenta PS: You can use https://github.com/nupic-community/river-runner to push any scalar data streams in River View into a NuPIC anomaly model. Try out a few different streams and see what you come up with!
