Clasificación y distribución de errores de automatización empleando técnicas de Machine Learning en la empresa Digital Ware.

Abstract

In this paper, the development of a system for classifying and distributing errors through Machine Learning techniques is exposed, which allows solving one of the most common problems in the automation area of the Digital Ware company, such as Difficulty in the revision and readjustment of automated tests due to the high handling of tests and information increasing the time required making the process slow. For this, it is proposed that this system make use of the selection and distribution of errors, taking as a reference the results obtained in massive automation launches through Azure pipeline connected to the business project, with the algorithms and techniques provided in the .NET libraries such as ML.NET and console applications such as NET CORE; where the algorithm has previously adequate and structured data entry, making use of data processing formats such as .txt, separating columns with tabulations for easy reading by libraries, facilitating its extraction, and classifying it through the binary model with more stable technique such as logistic regression, among others. Finally, the functional classification system is obtained and evaluated by the metrics strictly given for binary classifications, resulting in a high percentage of precision and confidentiality.