Title: Dual strategy based missing completely at random type missing data imputation on the internet of medical things

Authors: P. Iris Punitha; J.G.R. Sathiaseelan

Addresses: Department of Computer Applications, Bishop Heber College (Autonomous), Tiruchirappalli, Tamil Nadu, 620 017, India; Affiliated to: Bharathidasan University, India ' Department of Computer Applications, Bishop Heber College (Autonomous), Tiruchirappalli, Tamil Nadu, 620 017, India; Affiliated to: Bharathidasan University, India

Abstract: One problem that reduces performance in data analysis is missing data. Improper imputation of missing data may lead to an incorrect prediction. In the internet of medical things (IoMT) era, when a lot of data is created every second and data utilisation is a major issue for healthcare providers, missing values must be managed well. The literature proposes many missing data imputation methods. However, excessive missing value instances diminish the number of complete examples in the collection. Imputing missing data with few complete instances will not improve results. The number of complete instances could be raised by considering the imputed item as a complete object and utilising it alongside the existing complete instances for future imputations. So this work introduces a new dual strategy-based missing data imputation (DS-MDI) approach for IoMT missing completely at random (MCAR) data. The proposed DS-MDI technique uses cube-root-of-cubic-mean and enhanced Levenshtein distance-based clustering (ELDC) with cluster-directed closest neighbour selection (CSNN). This approach imputes more items using imputed objects. The Kaggle Machine Learning Repository's cStick IoMT dataset was processed using the suggested technique. The DS-MDI algorithm outperforms current missing data imputation algorithms in accuracy, precision, recall, and F-measure.

Keywords: missing data imputation; internet of medical things; IoMT; mean imputation and clustering; cluster-directed closest neighbour selection; CSNN; enhanced Levenshtein distance-based clustering; ELDC.

DOI: 10.1504/IJIEI.2023.136100

International Journal of Intelligent Engineering Informatics, 2023 Vol.11 No.4, pp.317 - 336

Received: 12 Apr 2023
Accepted: 24 Aug 2023
Published online: 16 Jan 2024 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article