An Integrated Approach for Healthcare Planning over Multi-dimensional Data Using Long-Term Prediction

The mining of temporal aspects over multi-dimensional data is increasingly critical for healthcare planning tasks. A healthcare planning task is, in essence, a classification problem over health-related attributes across temporal horizons. The increasingly integration of healthcare data through multi-dimensional structures triggers new opportunities for an adequate long-term planning of resources within and among clinical, pharmaceutical, laboratorial, insurance and e-health providers. However, the flexible nature and random occurrence of health records claim for the ability to deal with both structural attribute-multiplicity and arbitrarily-high temporal sparsity. For this purpose, two solutions using different structural mappings are proposed: an adapted multi-label classifier over denormalized tabular data and an adapted multiple timepoint classifier over multivariate sparse time sequences. This work motivates the problem of long-term prediction in healthcare, and places key requirements and principles for its accurate and efficient solution.