On regression methods based on linguistic descriptions

The prediction precision of mathematical models and their interpretability go usually against each other. The increase of the quality of one of the features decreases the other. In this article we introduce a new mathematical model based on Perception-based Logical Deduction (see [18], [19]) which is an implicative fuzzy inference mechanism based on linguistics semantics, and which enables the users to create models described with expressions of natural language. Our mathematical model increases the accuracy of inference mechanism used in regression analysis while it maintains the underlying linguistic semantics, which are crucial for human-computer interaction. In other words, we have managed to increase the prediction precision based on Perception-based Logical Deduction and not to decrease the interpretability of the system.