Studying innovation with patents and machine learning algorithms: A laboratory for engineering students

Teaching innovation management to engineers is becoming increasingly relevant. However, it can be difficult to involve engineers in a discipline in which technical competences do not represent the core whereas professional and soft skills play a critical role. For this reason, adopting the proper teaching approach is key to capture the students' attention and interest. In our study, we propose a laboratory for teaching innovation based upon two elements that are very closed to the engineering mindset: patents and machine learning algorithms. The laboratory proposes the application of machine learning approaches to patents data, for studying the innovation activity of companies. Three machine learning algorithms, Least Squares, Deep Neural Networks and Decision Trees are exploited. Their application is proposed to capture the relationships between relevant patents output variables (such as, for example, the number of forward citations, as proxy of the company's innovation capability) and the related input features (such as, for example, the number and type of patent technological classes). By practically using this approach, students can be introduced to some relevant topics in innovation management, such as investments, protection, market identification, cumulation of knowledge.