Search for: [Abstrakt = "After building the classification model, at the stage of the class description we try to extract knowledge from the model. We search for the description of classification rules, the natural language. The paper presents the simple algorithm for building the ranking of predictor variables based on their descriptive power \(for every class separately\) and uses boxplots to enable interpretation and give some insight. The procedure is universal and can be applied to classic or data mining methods. SVMs, Random Forest, Neural Network and k\-Nearest Neighbours were used for illustration with R software."]

Number of results: 1