Department of Business Development and Technology

Weighted objective distance for the classification of elderly people with hypertension

Research output: Contribution to journal/Conference contribution in journal/Contribution to newspaperJournal articleResearchpeer-review

  • Supansa Chaising, Mae Fah Luang University
  • ,
  • Punnarumol Temdee, Mae Fah Luang University
  • ,
  • Ramjee Prasad

Hypertension is a significant risk factor for cardiovascular disease (CVD) development and is commonly observed among elderly individuals. To prevent CVD complications, identifying groups according to their abilities to control hypertension is highly important. In previous studies, objective distance was proposed as a measurement for identifying groups of elderly people with hypertension. However, this measurement still has a problem requiring the further elimination of insignificant factors affecting classification accuracy. Therefore, a weighted objective distance (WOD) is proposed as a new measurement, which can address significant individual factors to obtain higher classification accuracy. WOD is based on the concept that hypertension controllability in each individual can be represented by different significant individual factors, which can be used for group classification. Therefore, WOD represents a modification of the original objective distance measurement, using weighting factors. Weighting factors obtained from information gain can prioritize factors according to their associated weights. WOD is assumed to be able to provide greater classification accuracy than the original objective distance. To test this assumption, the secondary data for 1198 elderly people with hypertension were collected. The WOD method provided classification results with 97.33% accuracy, which is higher than the original objective distance method, which provided 70.03% accuracy, consistent with the proposed assumption. To validate the WOD method, the classification accuracy was compared against those provided by decision tree and neural network classifiers. The results showed that the WOD method was more accurate than the decision tree and neural network classifiers, which provided 71.12% and 71.62% accuracy, respectively.

Original languageEnglish
Article number106441
JournalKnowledge-Based Systems
Publication statusPublished - Dec 2020

    Research areas

  • Cardiovascular disease, Elderly people, Hypertension, Information gain, Objective distance, Weighting factors

See relations at Aarhus University Citationformats

ID: 199065569