Forum Replies Created
-
Posts
-
You also get “points” for correctly identifying good anemometers.
score = sum(YourSubmission==GroundTruth)
So if YS = [1 0 0 1] and GT = [0 0 0 1], score = 3
See here:
http://www.phmsociety.org/node/697“Please note that to diminish the chance of using gradient information, the scores are obscured a bit. However, the rank order is correct – i.e., 1st listed team has best score, 2nd team has 2nd best score, etc.”
We rank teams based on how they are actually doing, but we report scores based on a subset. I have tweaked the way the subset is chosen to diminish the disparity a bit.
Max score is 675.
You may read the fine manual here:
https://www.phmsociety.org/competition/phm/11/scoringRegarding the training data: they are normal (unfaulted) operating data for the respective types of anemometer data. Your task is to use these files as prototypes of normal behavior and to figure out how to identify data that differs from it. There are numerous approaches for that.
You may submit ONE new result file per day.
The leaderboard shows the rank order of the last file submitted.
There are six individual cutter records, c1…c6. Records c1, c4 and c6 are training data, and records c2, c3, and c5 are test data.
I mis-typed about c6. If you downloaded *c3.zip* before 15 June, the wear file is incorrect – please discard it.
The c6.zip wear file is fine.
