Improvements the HANN-L2F for classification by using k-means

• Main Authors: Jirawat Teyakome, Narissara Eiamkanitchat
• Format: Conference Proceeding
• Published: 2018
• Subjects: Computer Science; Engineering
• Online Access: https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=84966545425&origin=inward; http://cmuir.cmu.ac.th/jspui/handle/6653943832/54377
• Institution: Chiang Mai University

© 2015 IEEE. This paper presents the improved algorithm for the Hybrid Approach of Neural network and Level-2 Fuzzy set (HANN-L2F). The main structure is including 2 parts. The first part is Neuro-Fuzzy system, including the MLP Neural network with the combination of the level-2 Fuzzy system. The second part is using k-nearest neighbor to classify the output from Neuro-fuzzy. The HANN-L2F is an algorithm with high classification performance. However, the process to determine the number of membership functions in HANN-L2F is take time and sometimes results the high number of clusters, this make the high complexity of the overall process. This paper, including the data adjustment in HANN-L2F by using the determination process and implement k-means algorithm to find the appropriate number of clusters. Several types of standard datasets from UCI repository machine learning are used to verify the performance of the propose algorithm. The additional data adjustment process can improve the classification performance of HANN-L2F, furthermore, can reduce the classification time as displayed in the experimental results.