The declining birthrate and the increasing aging population can exacerbate various societal issues such as social isolation, which can have a serious impact on the mental and physical health of older adults. Increased frequency of going out can reduce the possibility of future social isolation and facilitate recovery from social isolation. In this paper, we propose a novel method for generating travel recommendations for older adults to increase their frequency of going out. The proposed method builds a travel-recommendation model based on social media posts by older adults. The modelling process exploits the semi-supervised Latent Dirichlet Allocation (ssLDA) and object detection techniques to extract the interests of older adults by analyzing latent topics in textual and visual messages. Travel recommendations can be generated by matching the latent topics and the online information about travel destinations. Our feasibility study demonstrates a higher recall in predicting relevant topics for older adults compared to a baseline method that relies on the conventional Latent Dirichlet Allocation (LDA) model.