Revealing Cause-Effect Relations in Comorbidities Analysis Using Process Mining and Tensor Network Decomposition

The existence of certain comorbidities, the co-occurrence of different diseases in the same individual, is well-known in the medical community. However, finding temporal cause-effect relations between those diseases constitutes a big challenge. Clinical records can be used to extract the required information, but their analysis is elusive due to the vast and heterogeneous amount of data. We propose a new methodology for time-preserving tensor networks decomposition to be applied in the analysis of big data problems where the temporal dimension of the key factual fields must not be modified. This methodology will also allow the creation of a new process mining modeling which can capture the cause-effect relations as low-order tensors associated to the transitions of the mined processes, and whose structure takes the form of a multilayer complex network. All these theoretical and methodological advances will allow their application to real biomedical data to analyze comorbidities.