| |
 |
| Paper: |
Supervised Classification of Software Problem Reports Characterized by Error Log Profiles |
| Volume: |
522, Astronomical Data Analysis Software and Systems XXVII |
| Page: |
603 |
| Authors: |
Miranda, N.; Gil, J. P. |
| Abstract: |
ALMA uses a ticket system based on JIRA to track problem reports (PR) that appears during regular operations. Some of those problems are assigned to a team of software engineers on charge of analyzing event logs around the time each problem was reported and relating this information to either new or ongoing investigation tickets (ICT). This can be modeled as a supervised classification problem that associates a new PR to an existing ICT ticket. The classifier is based on a list of features extracted from the error log profile by using text mining techniques tuned by the engineers acting as domain experts. In this work a real database of PR / ICT is used. An event log coloring technique was implemented to make clusters of similar events as a pre-processing stage to improve feature extraction used later to build a matrix. A probabilistic method is then used in order to generate scored suggestions, so the user who performs the investigation can have some key indicators of what previous IT could be associated with the problem in hand, and thus help in the troubleshooting process. |
|
|
|
|
|
