Explaining Computer Predictions with Augmented Appraisal Degrees

Abstract

An augmented appraisal degree (AAD) has been conceived as a mathematical representation of the connotative meaning in an experience-based evaluation, which depends on a particular experience or knowledge. Aiming to improve the interpretability of computer predictions, we explore the use of AADs to represent evaluations that are performed by a machine to predict the class of a particular object. Hence, we propose a novel method whereby predictions made using a support vector machine classification process are augmented through AADs. An illustrative example, in which the classes of handwritten digits are predicted, shows how the augmentation of such predictions can favor their interpretability.

Contextualizing Support Vector Machine Predictions

“…For instance, Figure 3 depicts the overall influence of x in a 2-dimensional feature space, where  = {f 1 , f 2 }. In this case, if Figure 2 A contextual view of XSVMC [14].…”