How to best predict short bowel syndrome outcome with machine learning approaches?

In recent years, there has been an extensive use of machine learning techniques in the medical field for diagnostic or therapeutic prediction purposes. In the field of short bowel syndrome, numerous statistical studies have been proposed, but to date no machine learning techniques have been exploited to predict the outcomes of the surgery commonly performed in paediatric patients suffering from this pathology. One reason for this lack can be identified in the fact that this is a rare condition and therefore it is difficult to have a large dataset. This paper investigates the possibility of processing demographic data of paediatric short bowel syndrome patients by spot-checking machine learning algorithms on a dataset of 86 patients. The experimental setup was developed to ensure the best performance of each algorithm and to take into account the moderate unbalance of the output classes. The Decision Tree algorithms proved to be the best solution in terms of accuracy, precision, recall and F1-score (obtaining values of 0.85 for each metric considered), capable of better understanding the data model.