User login

Weekly Report -- 08/06/2012




Marked the latest batch of libtrace assignments for 513.

Continued manually checking for anomalies in my various test time series. Found a few scenarios where a certain type of pattern in the time series would cause the prediction algorithm to go a bit crazy for a while. After a bit of investigation, I found the problem was caused by the denoising algorithm which was creating the problematic pattern.

Throwing all my manual classifications into WEKA unfortunately did not reveal any obvious avenues for improvement, so it is back to the drawing board a bit on this one. None of the metrics I currently have can help resolve the false positives and false negatives, so I'll need to look for a new one.





ARIMA and outliers


I stumbled upon your blog and found it interesting. I am not sure how long you have been at this, but we started this process which you are trying to replicate in 1976. Take a look at the Capabilities ppt on our website and it might spark some ideas for you to change your code. I would be happy to take this off-line with test examples to show you how our software handles data and compare it to your output or continue using "comments" to show you "stuff". The ARIMA modeling process may be more complicated than you have thought. See my Autobox discussion group on and you will see my comments that provide further descriptive info on this nightmare which you are experiencing and we have solved. Now, it is true that once out of 1,000 times we fail.

Tom Reilly