Intention Mining Data preprocessing based on Multi-Agents System

Abstract

Abstract Today vast and diverse event records of applications exist for almost every scientific domain, making their integration and intelligent exploitation challenging. Indeed, complex data require expressive data representation models. This work deals with the intention mining that is a very active and promising research area. Performance, flexibility, and adaptation are the biggest challenges for intention mining. Those issues have been illustrated from research on method engineering and guidance. Intention mining is the ability to predict a user’s goals. Knowing the user’s intention can support the decision-making of the network administrators. In addition, the user-friendliness is another big challenge to the intention mining in this paper. For this, based on the communication and coordination of intelligent Agents, a new multi-Agents based System approach is gathered in order to discover an intentional process model and provide specific recommendations. In addition, the input of intention mining is the trace of activities, which is an unstructured file because the activities are distributed from different sources. However, the dataset that will be used in intention mining must be well structured and filtered, so often some efforts are required to filter the relevant data. The main input of all algorithm used to discover intentional process model is the log file (traces activities), which is unstructured dataset and not ready to be feed as-is to machine learning algorithm. Therefore, this paper aims to describe the data preprocessing steps, which transform the unstructured log file to a structured one.