Indukcja reguł dla danych niekompletnych i niezbalansowanych: modele klasyfikatorów i próba ich zastosowania do predykcji ryzyka operacyjnego w torak
Classification rules extraction for missing and imbalance data: models of classifiers and initial results in the rules-based thoracic surgery risk pre

Summary/Abstract: The classification problem of multi-faceted imperfect data, e.g. with missing values and at the same time with class imbalance, is considered. Aspects of the classification effectiveness and interpretability of the results through classification rules extraction for the ”black-box” like classifiers are discussed. An approach based on a boosted SVM classifier and an oracle-based decision rules extraction procedure is proposed and applied to a sample hospital data base of Wrocław Thoracic Surgery Centre. The research was performed using Imbalanced Learning Module of the KEEL Data Mining software package and WEKA Ma-chine Learning environment.

• Details
• Contents