Random Forest Method Implementation for Analyzing the Up and Go Test Results from a Fitness Training Routine Data Performed by a Taiwanese Elderly Group

• Main Authors: Diana Eloísa Roa Flores, 迪亞娜
• Other Authors: Tien-Lung Sun
• Format: Others
• Language: en_US
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/7t39vz

碩士 === 元智大學 === 工業工程與管理學系 === 105 === Taiwan is becoming in a "hyper-aged" society that carries with needs of social health care services. In order to tackle these needs, in 2008 Taiwan government launched the ten-year long-term care program. On the base of senior’s health promotion activity, a community center in Yilan County implemented a fitness training routine for an elderly group which was composed of 23 patients. The present research aimed to analyze the incidence of this fitness training routine over their functional fitness test results, specifically for the Up and Go test, by building a model able to classify the Up and Go test results. This in favor of enables healthcare professionals to modify the patients training routine and improve the efficiency the routine execution. In order to figure it out, a Random Forest Algorithm was implemented. In this process, the data was prepared and different approaches of the algorithm inputs were combined yielding eight models. Later, the models were evaluated on their Accuracy to classify and by using Receiver Operator Characteristic (ROC) graphs which allowed to obtain the ROC Area Under the Curve Score. After the evaluation, it was determined the best classification model which showed a Classification Accuracy equivalent to 0.86 and a ROC AUC Score equal to 0.83. Subsequently, the features importance rank was obtained which identified seventy-four features as the most incident on the Up and Go test results classifications.