Browse Prior Art Database

Verification records classification and prioritization based on SVM algorithm Disclosure Number: IPCOM000236675D
Publication Date: 2014-May-08
Document File: 4 page(s) / 74K

Publishing Venue

The Prior Art Database


Significant test and development resources are needed to verify resolved internal defects and field bugs (APARs). A classification is proposed of the remaining verifications records based on Support Vector Machine (SVM) algorithm, in order to reduce the required efforts - especially to reduce time and resources spent on verifying correctly fixed items.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 60% of the total text.

Page 01 of 4

Verification records classification and prioritization based on SVM algorithm

    Four hundreds fields apars and defects ported to current stream from maintenance stream generates 400 verification records assigned to test team. Single verification cost is around 1 person day. To resolve verification backlog around 400 person days (around 2 years for single test engineer) are needed. Let's assume that a test team consists of 5 test engineers, the whole team will be excluded from major test activities within development stream for 4 months. The most important thing is that less than 10% of ported items cause regression (in our case). 90 % of spent time produces no results/input to product and current release. That time/effort could be put and invested in different areas of test e.g. automation and extension of regression suite .

    Let's imagine how much the manual work can decrease if all verification records are categorized to 2 classes: impossible regression, possible regression and within each class the confidence factor is assigned to each verification record . It will allow the test team to focus in first place on items from set 2 (real regression) and re-use confidence factor for work prioritization.

Support Vector Machines algorithm:

    Phase 1: Learning on historical data (already verified - training set) using selected features - generation of separating hyperplane

    Phase 2: Classification of new apars to 2 classes: (OK, REGRESSION) and distance calculation (risk =...