Browse Prior Art Database

Neural net to improve turn around time of validation testing

IP.com Disclosure Number: IPCOM000132022D
Original Publication Date: 2005-Nov-29
Included in the Prior Art Database: 2005-Nov-29
Document File: 3 page(s) / 36K

Publishing Venue

IBM

Abstract

Using Neural Nets To Improve Turn Around Time Of Validation Testing - Neural nets can be used to automatically prioritize the order of testcases to be run for validation testing. Based on software changes and a history of previous runs a neural net can be used to determine which testcases are most likely to reveal a problem. These testcases are ordered to be run first so that fixes can be found earlier in the testing cycle and therefore reduce turnaround time. Since neural nets learn over time, as the software program matures, the neural net will do a better job predicting which test structures to run first.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 54% of the total text.

Page 1 of 3

Neural net to improve turn around time of validation testing

Neural nets can be utilized to help the turnaround time of validation testing. Any software that changes and has associated testcases that are run can be used for this process. The goal is to identify which testcases have the highest probability of giving an unexpected result. These testcases are run first so that problems with the software can be highlighted quicker and fixed quicker, therefore reducing the turnaround time to do the testing. The developer does not need to wait for the entire test suite to finish. Problems would most likely be highlighted very early in the runs. The current method is for the developer to manually prioritize the testcases based upon their knowledge of what changed and what the testcases check. Alternatively if no prioritization takes place the developer must use brute force by running all testcases through in no specific order. Problems may not be highlighted until the end of this process. As the software matures and more of a history builds up, the neural net will learn and more accurately predict results.

The first step is to obtain the changes to the software. This can be extracted in several ways. Highlighted below is one way of obtaining the information using a diff command on the files in a revision control system.

svn diff -r HEAD | grep @@ @@ -1,72 +0,0 @@ @@ -1,37 +0,0 @@ @@ -1,82 +0,0 @@ @@ -1,108 +0,0 @@ @@ -1,17 +0,0 @@ @@ -1,8 +0,0 @@

This information is then processed into a usable input format. In this case it reduced to file number and percentage location in the file

filenum %loc filenum %loc filenum %loc ... 4 18 4 42 8 15

The percent location is defined as: linenum/total lines * 100 %. Line percent should work better then line number which would change almost everytime.

Using fann (...