Sensitivity Analysis of the Negative Selection Algorithm Applied to Anomalies Identification in Builds

The search for replacing manual processes with automated processes brings with it an increase in complexity related to its controls and monitoring. The use of builds, that is, automated software delivery processes, is a good example. Its primary objective is the construction, packaging, testing, and delivery of system versions. The execution of tests in the context of software delivery automation materializes the existing controls in the execution of manual processes and can basically result in success or failure. The failure state occurs when one or many of the steps that make up the automated process do not obtain the expected result. The software industry invests a lot of time in investigating build failures, as they can fail for reasons not directly related to the tests performed. Such failures are called anomalies. This article presents a way to automatically identify anomalies using a natural computing algorithm inspired by artificial immune systems, called the Negative Selection Algorithm (ASN), in order to obtain the correct classification of failures in builds. The focus of the article is on the sensitivity analysis of the ASN in relation to the neighborhood radius of the detectors and the number of detectors generated.