Understanding complex interactions between neighborhood environment and personal perception in affecting walking behavior of older adults: A random forest approach combined with human-machine adversarial framework

Abstract

The population aging is a growing problem worldwide. Walking is one of the most important ways of self-management of health for older adults, determined by many factors, such as neighborhood environment (NE) and socio-economic attributes. Although the previous studies have typically predicted elderly walking behavior through NE, they are limited by the methodological system and data collection, resulting in low prediction accuracy. To this end, this study incorporates residents' subjective perceptions of the environment and objective neighborhood environmental attributes into the evaluation system, uses human-machine adversarial framework and machine learning methods to predict elderly walking behavior, and assesses the nonlinear effects of each factor. The results show that (1) combining subjective and objective factors, the prediction accuracy of elderly walking behavior has been effectively improved based on human-machine adversarial framework and machine learning methods. (2) The nonlinear and threshold effects of environmental and perceptual factors on the walking time of the elderly were revealed. (3) The neighborhood attributes were incorporated into the walking behavior prediction, and were found to be of comparable importance to the influence of the NE on the behavior of the elderly. These results provide more reliable qualitative and quantitative auxiliary suggestions for planners.