Big data analytics in smart living environments for elderly monitoring

Today, data collected in smart-living environments are constantly increasing in the dimensions of volume, velocity and variety, which characterize any big data application. In such a way, it makes sense to investigate big data analytics for elderly monitoring at home. The aim of this study is to conduct a preliminary investigation of state-of-the-art algorithms for abnormal activity detection and change prediction, suitable to deal with big data. The algorithmic approaches, under evaluation and comparison, belong to the three main categories of supervised, semi-supervised and unsupervised techniques. At this purpose, specific synthetic data are generated, including activities of daily living, home locations in which such activities take place, as well as physiological parameters. All techniques are evaluated in terms of abnormality-detection accuracy and lead-time of prediction, using the generated datasets with various kinds of perturbation. The achieved results, even though preliminary, are very encouraging, showing that unsupervised deep-learning techniques outperform traditional (machine learning) ones, with detection accuracy greater than 96% and prediction lead-time of about 15 days in advance.