|International Conference and Workshop on Communication, Computing and Virtualization|
|Foundation of Computer Science USA|
|ICWCCV2015 - Number 2|
|Authors: Maya Aikara, R.r. Sedamkar, Sheetal Rathi|
Maya Aikara, R.r. Sedamkar, Sheetal Rathi . An Improved Hybrid Approach of Mining Graphs using Dual Active Feature Sample Selection and LTS (Learn-To-Search). International Conference and Workshop on Communication, Computing and Virtualization. ICWCCV2015, 2 (September 2015), 0-0.
In software engineering, discriminative sub graphs are used to identify the bug signatures (context of bug). Most of discriminative sub graph mining algorithms estimate the discriminative sub graphs from a positive and negative labelled graph dataset. The labelling is done manually, which is time as well as cost consuming. A hybrid discriminative sub graph mining algorithm using dual active feature sample selection and LTS, which reduces the manual labelling by 60%. But, this hybrid approach does query graph computation without considering the features of the labelled input graph dataset. Even the precision limit is set to 4, which may not be optimal for all type of input dataset. This paper presents an improved hybrid approach, which does a query graph computation considering all graphs in the input dataset. An additional tool is used for input pre-processing method. The average precision limit is determined so as to achieve maximum recall for any type of input dataset. The experiments and results shows that the improved hybrid approach can achieve an average recall of 66. 67% when the precision limit is set to 3, whereas the earlier hybrid approach attained an average recall of 33. 33% when precision limit was set to 4.