The Machine Learning Algorithm Selection Model: test with multiple datasets

The technological revolution known as Industry 4.0 is permeating and changing the way companies of all sizes manage their processes. The revolution is influencing companies process at all levels, including production, service, and management ones. Not surprisingly, the strong digitalisation currently occurring in the industrial scenario is contributing to the generation of unprecedented quantities of data that companies can exploit for several purposes and scopes. New data analysis approaches, able to exploit the computational power of modern PCs and workstations are being studied by researchers and practitioners to identify patterns and generate knowledge from data. Yet, despite being able to collect increasing quantities of data, many companies still lack the capabilities and competencies to use analytic approaches such as Machine Learning (ML), elaborate data into information and, thus, generate value. A model, namely the Machine Learning Algorithm Selection Model (MLASM), has been proposed to guide the unexperienced users in selecting a set of ML algorithms suitable for their analysis according to the scope of the analysis and the characteristics of the dataset. This paper describes the process used to test the MLASM with several datasets to verify its usefulness and the correctness of its suggestions. In accordance with the results, improvements and updates have been proposed for the MLASM.