We develop statistical algorithms to infer possiblecardiac pathologies, based on data collected from 24h Holterrecording over a sample of 2829 labelled patients; labels highlightwhether a patient is sufering from cardiac pathologies. In the frstpart of the work we analyze statistically the heart-beat seriesassociated to each patient and we work them out to get a coarse-grained description of heart variability in terms of 49 markerswell established in the reference community. These markers arethen used as inputs for a multi-layer feed-forward neural networkthat we train in order to make it able to classify patients.However, before training the network, preliminary operations arein order to check the efective number of markers (via principalcomponent analysis) and to achieve data augmentation (becauseof the broadness of the input data). With such groundwork,we fnally train the network and show that it can classify withhigh accuracy (at most 85patients that are healthy from thosedisplaying atrial fbrillation or congestive heart failure. In thesecond part of the work, we still start from raw data and we geta classifcation of pathologies in terms of their related networks:patients are associated to nodes and links are drawn accordingto a similarity measure between the related heart-beat series.We study the emergent properties of these networks lookingfor features (e.g., degree, clustering, clique proliferation) ableto robustly discriminate between networks built over healthypatients or over patients sufering from cardiac pathologies. Wefnd overall very good agreement among the two paved routes.