Development of Clustering Algorithms for Older Faller in Malaysia

Falls are serious problem which lead to negative consequences on the quality of life especially for older people. Most falls are caused by the interaction of multiple risk factors. However, manual analysis in fall data are time consuming and high processing cost. Therefore, this study purpose to develop a clustering-based fall risk algorithm which can provide assistances for clinician. The proposed algorithm consists of several stages included data pre-processing, feature selection, feature extraction, clustering and characteristic interpretation. This study employed Malaysian Elders Longitudinal Research (MELoR) dataset. A total of 1279 subjects and 9 variables are selected for clustering. The combination of t-Distributed Stochastic Neighbour Embedding (t-SNE) for feature extraction, and K-means clustering algorithm are chosen to cluster the subjects into Low (13%), Intermediate A (19%), Intermediate B (21%) and High (31%) fall risk group. In comparison, older people with higher fall risk have slower gait, imbalance, weaker muscle strength, with cardiovascular disorder, poor performance in cognitive test, and advancing age. To conclude, the proposed fall risk clustering algorithm is capable to group the subjects that have similar features. It presents a potential as assessment tool in management of falls.