Razlaga klasifikatorjev na podlagi podkonceptov

Mušič, Nejc (author), Robnik Šikonja, Marko (mentor)

V diplomski nalogi preizkusimo pristop razlage klasifikatorjev na podlagi podkonceptov na umetni in realni podatkovni množici. Pri razlagi so pomembni algoritmi gručenja, zato jih testiramo na dveh umetnih in realni podatkovni množici.

• Language: Slovenian
• Keywords: razložljiva umetna inteligenca, MDEC, DBSCAN, HDBSCAN, K-MEANS, SHAP, odločitvena pravila, medoid, prototip
• Work type: Bachelor thesis/paper
• Typology: 2.11 - Undergraduate Thesis
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2023
• PID: 20.500.12556/RUL-152726
• COBISS.SI-ID: 167284995
• Publication date in RUL: 04.12.2023

Secondary language

• Language: English
• Title: Explanation of Classifiers based on Subconcepts

Abstract

In this work, we evaluate an approach to explaining classifiers based on subconcepts using both artificial and real-world datasets. Clustering algorithms play a crucial role in this explanation process, and we assess their performance on two artificial and one real-world dataset.

• Keywords: XAI, MDEC, DBSCAN, HDBSCAN, K-MEANS, SHAP, decision rules, medoid, prototype