ANÁLISE QUALITATIVA DOS TEXTOS DE CORREÇÕES DE ERRO DE UM ERP E DISTRIBUIÇÃO AUTOMÁTICA DAS TAREFAS

Abstract

An integrated management system or ERP (Enterprise Resource Plain) is a business software package which comprises the management of several areas of a company, centralized in a single database. The development of such versatile software requires multidisciplinary teams respectively responsible for each ERP module. This paper will address the flow of tasks in the process of maintaining an integrated management system for large scale, with the objective of making possible and minimizing errors in the classification process of these tasks, proposing that the interpretation of the texts be done by an algorithm in detriment to the interpretation of the employees inherent to the ERP maintenance process. The "Support Vector Classifier" algorithms were analyzed; "Multinomial Naive Bayes" and the implementation a Perceptron neuronal network to interpret tasks and distribute them among teams. We detail the data preparation process and the definition of the most effective algorithm for the solution.