| forum.alglib.net http://forum.alglib.net/ |
|
| Detecting abnormalities using Principal Componenets Analysis http://forum.alglib.net/viewtopic.php?f=2&t=52 |
Page 1 of 1 |
| Author: | moex [ Sat Aug 21, 2010 7:11 am ] |
| Post subject: | Detecting abnormalities using Principal Componenets Analysis |
Hello All, I have been studying PCA for a while now, and I finally figured out that when applying PCA to normalized, mean-adjusted data it only transforms the data to a coordinate system that describes the most variance in data. This is a very important feature of PCA since it can change our perspective to data. We can understand these new perspectives visually if the number of variables are 2 or 3 variables. But in real life this is not the case, and we usually have to handle hundreds if not thousands of variables. This way PCA would produced a perspective that is difficult for us to comprehend. Our goal anyway is to locate events in the data which are considered abnormal, and then map the event to one or more variables according to correlations and patterns in the data. This can be done through two charts: 1- SPE (Standard Prediction Error) 2- T^2 (T Square) I was wondering if alglib does provide any of the two previous methods or other methods that further extend PCA. Thanks for your time. |
|
| Author: | Sergey.Bochkanov [ Sat Aug 21, 2010 7:31 pm ] |
| Post subject: | Re: Detecting abnormalities using Principal Componenets Anal |
No, it doesn't. Actually, data analysis algorithms are not my main area of interest now. Currently I focus on more "basic" algorithms (optimization, root finding, linear algebra), new languages, low level optimizations and multithreading. |
|
| Page 1 of 1 | All times are UTC |
| Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group http://www.phpbb.com/ |
|