Comparison Activation Function of Deep Neural Network for Class Recognition Classification of the Elderly  
Author Dong Su Lee

 
Co-Author(s) Youn Su Kim; In Hong Chang

 
Abstract This paper identifies the degree of subjective class recognition of the elderly. We examine the degree of class recognition classification of the elderly by various variables that the general characteristics of elderly people, subjective expectation, life satisfaction, household income, living cost, etc. There is a significance to suggest a new dimension to the quality of life of the elderly by examining the influence factors of the elderly’s life not only on the individual dimension but also on the subjective class recognition of the individual in the community. This paper presents the types of activation functions during the deep learning analysis and compares the classification rates by type. In this paper, 4,485 elderly people aged 55 years and over using the 6th data of the aging research panel of the Korea Employment Information Service in 2018 are subjected to using the deep neural network during the deep learning analysis. The results show that the PReLU among functions that the ReLU function, the Leaky ReLU function, the PReLU function, and ELU function have the highest class recognition classification of the elderly.

 
Keywords subjective class recognition, deep neural network, activation function, ReLU function
   
    Article #:  DSBFI19-60
 
Proceedings of ISSAT International Conference on Data Science in Business, Finance and Industry
July 3-5, 2019 - Da Nang, Vietnam