Machine learning per la pubblica amministrazione

[Machine translation] The article deals with the impact of big data on the knowledge assets of public administrations, starting from an experiment on the database of national public contracts that personally involved the University of Turin and the National Anti-Corruption Authority (ANAC). The article illustrates the various phases that a computer scientist or data scientist follows to arrive at the use of data for cognitive purposes: the initial statistical approach aimed at identifying the descriptive characteristics of the cases under study, is followed by the descriptive approach aimed at identifying the regularities and correlations of the available database; after these two phases, the research path continues through the predictive approach through machine learning techniques. The article concludes by promoting the prescriptive approach as functional to the identification of decisions that should be taken on the basis of the available data and that could suggest future good practices.