It is possible to apply SVM algorithm on small data set of 10 attributes?

Hi Shubhangi,

i don't see a problem of "just" using 10 attributes. The success, i.e. a potential increase of accuracy, will depend on the quality of attributes.

Greetings, Sebastian

.

as Sebastian says above, 10 **attributes** (that is variables or features) is not a problem for svm

now, re-reading your question, i am not sure what you mean exactly : have you a data set of N labelled instances described on 10 features ... or have you a small data set of 10 labelled instances described on F features ("small data set" indeed !)?

.

Yes, it is possible to use SVM for a real-time implementation of abnormal process detection in Windows using only 10 attributes- only if these 10 attributes are enough to make the machine differentiate between a normal and an abnormal process. As Sebastian said, "the accuracy of the SVM model will depend on how good the attributes are". Regarding the time factor, for real-time implementation - Not a problem for a small dataset.

Thanks Sebastian, FabriceClerot, Abhilash and Andreas for the valubale suggestions.. @ Fabrice Clerot in our problem,  the case is  data set of N labelled instances described on 10 features.. Whatever these 10 parameters data of standard windows process helpful to detect the abnormal activities..all of these parameters are important because not only single parameter get altered by attcker(or any suapicious activity) it will having some affect on another parametrs also ..so we consider relatively important parameter.. Normal windows starting process are in less number so whatever data about these process in terms of these parametrs are less thats why confusion is any supervised learning technique should be applicable in such cases..

Shubhangi,

It is possible, but you need to use a specific type of SVM that learns incrementally. However, with only 10 instances, I am afraid that you could get low prediction accuracy. Consequently, I recommend increasing your data and make much bigger. 

HTH.

Samer