A two-tier hybrid parameterization framework for effective data classification

• Published in: Frontiers in Artificial Intelligence and Applications
• Main Author: Mohamad M.; Selamat A.
• Format: Conference paper
• Language: English
• Published: IOS Press BV 2018
• Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85063374979&doi=10.3233%2f978-1-61499-900-3-321&partnerID=40&md5=b99cb61584af8c77494a6276dba6e3e4

The classification process is a decision-making task. In order to obtain a good decision, the classification process needs to be conducted by following a standard framework or approach. The selection of a good parameterization method plays an important role in executing an effective parameterization process. The parameterization process will generate an optimized parameter reduction set that helps the classifier in generating a significant result. However, the size and characteristics of the dataset might also influence the generation of results in the classification process. The process of parameterization becomes more complex when the dataset is big and consists of uncertain and inconsistent data. Therefore, these problems need to be considered during the decision-making process. Many solutions have been provided recently by researchers, but most of the research works did not consider the highlighted issues as problems that should be solved together through the use of a single framework. In this paper, an alternative framework was proposed for decision-makers in conducting the decision-making process. The framework was demonstrated by the use of a two-tier hybrid parameterization phase that involving two main processes for identifying the most optimized parameter set for uncertain and inconsistent big datasets. The results showed that the proposed framework can be implemented to significantly increase the classification performance by returning an accuracy rate of more than 70% for all datasets. © 2018 The authors and IOS Press. All rights reserved.