Early diagnosis of social isolation in older adults can prevent physical and cognitive impairment. This diagnosis usually consists on personal and periodic application of psychological assessment instruments. Unfortunately this is often a tedious process and therefore this is a situation that opens up opportunities in creating new ways of diagnosis. In this context, ambient intelligence and social networking sites are suitable technologies for automatic monitoring of significant changes in social interactions of older adults. However, current instruments that measure social isolation are based on subjective aspects since they only evaluate emotional social support and they do not consider objective aspects. This is the reason why a prediction model based on objective isolation variables is needed in order to measure the social interactions through computing mechanisms. This paper presents the development of a prediction model from the social interaction activities that can be registered through smartphone's sensory capabilities, the personal communications using an online social network and also radio-frequency identification mechanisms. The proposed model will benefit institutions interested in developing technological solutions to detect early stages of social isolation and improve the quality of life of older adults.