A Comparative Study of Missing Data Imputation Methods for Activity Recognition Task

Authors

  • Afia Sajeeda Institute of Information Technology, University of Dhaka, Dhaka-1000
  • B M Mainul Hossain Institute of Information Technology, University of Dhaka, Dhaka-1000
  • Sumon Ahmed Institute of Information Technology, University of Dhaka, Dhaka-1000

DOI:

https://doi.org/10.3329/dujase.v7i1.62880

Keywords:

Generative Adversarial Networks, GANs, Missing Data, Imputation, Deep Learning.

Abstract

While data is integral to address real-world problems using machine learning techniques, the availability of complete data is challenging as data goes missing during collection or even afterward. Missing values make data unsuitable for use and demand imputation techniques to resolve the problem. Here, we compare the performance of four existing missing data imputation techniques KNN, MICE, GAIN, and HexaGAN by applying them on KU-HAR dataset and two datasets of the collection, Nurse Care Activity Recognition Dataset. Our investigation suggests that HexaGAN learns the original data distribution better and demands further considerations, as the RMSE between the imputed data and the real data is lower. To investigate the role of activation function, we replace the underlying ReLU activation functions of the neural networks in HexaGAN architecture with the Swish activation function. Experimental results show that the modified version of HexaGAN possesses the potential to outperform the original one when applied to the same activity recognition datasets.

DUJASE Vol. 7(1) 1-8, 2022 (January)

Abstract
88
PDF
72

Downloads

Published

2023-02-01

How to Cite

Sajeeda, A., Hossain, B. M. M. ., & Ahmed, S. . (2023). A Comparative Study of Missing Data Imputation Methods for Activity Recognition Task. Dhaka University Journal of Applied Science and Engineering, 7(1), 1–8. https://doi.org/10.3329/dujase.v7i1.62880

Issue

Section

Articles