Handling Veracity of SVM Predictions

Any concern about the veracity of predictions made by artificial intelligence (AI) systems might deter decision makers from using them to support their decisions. To dismiss such concerns in AI systems based on support vector machines (SVMs), in this paper we explore the use of L-grades for dealing with the veracity of SVM predictions and propose a novel method for obtaining those grades. An illustrative example shows how an L-grade can be used for denoting to what extent an SVM prediction can be trusted, as well as how to obtain L-grades inside a binary classification process.