1. ZHAW digitalcollection
  2. School of Engineering
  3. Abschlussarbeiten
  4. Bachelorarbeiten ZHAW School of Engineering
Please use this identifier to cite or link to this item: https://doi.org/10.21256/zhaw-29936
Full metadata record
DC FieldValueLanguage
dc.contributor.advisorReif, Monika Ulrike-
dc.contributor.advisorHochberg, Alan-
dc.contributor.authorHoma, Celina-
dc.date.accessioned2024-02-15T14:18:22Z-
dc.date.available2024-02-15T14:18:22Z-
dc.date.issued2023-
dc.identifier.urihttps://digitalcollection.zhaw.ch/handle/11475/29936-
dc.description.abstractArtificial intelligence (AI) has become a valuable tool in skin cancer classification. However, its widespread adoption is limited by concerns about trustworthiness and its black-box nature. One promising approach to increasing trust is to improve the explainability of AI systems. This thesis explores the potential of using algorithms from the AIX360 library to achieve this goal. The algorithms - ProtoDash, DIP-VAE, LIME, SHAP and CEM - were evaluated for their runtime, simplicity, explainability/interpretability and stability in the context of skin cancer classification. The results suggest that each algorithm offers unique insights, but none can be considered universally superior. In particular, LIME and SHAP showed a promising balance between interpretability and stability, making them strong candidates. Further work is needed to identify the optimal combination of algorithms to increase confidence and to adapt these algorithms to handle complex data sets. This work is a step towards developing AI models that are not only effective, but also transparent and accountable, furthering the quest for trustworthy AI in healthcare.de_CH
dc.format.extent78de_CH
dc.language.isoende_CH
dc.publisherZHAW Zürcher Hochschule für Angewandte Wissenschaftende_CH
dc.relation.ispartofseriesBachelorarbeiten ZHAW School of Engineeringde_CH
dc.rightshttp://creativecommons.org/licenses/by/4.0/de_CH
dc.subjectMachine learningde_CH
dc.subjectArtificial intelligencede_CH
dc.subjectExplainable AIde_CH
dc.subjectAIX360de_CH
dc.subjectSkin cancer detectionde_CH
dc.subjectImage classificationde_CH
dc.subject.ddc006: Spezielle Computerverfahrende_CH
dc.titleEvaluating XAI algorithms in skin cancer classification : a path towards trustworthy AI systemsde_CH
dc.typeThesis: Bachelorde_CH
dcterms.typeTextde_CH
zhaw.departementSchool of Engineeringde_CH
zhaw.publisher.placeWinterthurde_CH
dc.identifier.doi10.21256/zhaw-29936-
zhaw.originated.zhawYesde_CH
Appears in collections:Bachelorarbeiten ZHAW School of Engineering

Files in This Item:
File Description SizeFormat 
2023_Homa_Celina_BA_SoE.pdf5.26 MBAdobe PDFThumbnail
View/Open
Show simple item record
Homa, C. (2023). Evaluating XAI algorithms in skin cancer classification : a path towards trustworthy AI systems [Bachelor’s thesis, ZHAW Zürcher Hochschule für Angewandte Wissenschaften]. https://doi.org/10.21256/zhaw-29936
Homa, C. (2023) Evaluating XAI algorithms in skin cancer classification : a path towards trustworthy AI systems. Bachelor’s thesis. ZHAW Zürcher Hochschule für Angewandte Wissenschaften. Available at: https://doi.org/10.21256/zhaw-29936.
C. Homa, “Evaluating XAI algorithms in skin cancer classification : a path towards trustworthy AI systems,” Bachelor’s thesis, ZHAW Zürcher Hochschule für Angewandte Wissenschaften, Winterthur, 2023. doi: 10.21256/zhaw-29936.
HOMA, Celina, 2023. Evaluating XAI algorithms in skin cancer classification : a path towards trustworthy AI systems. Bachelor’s thesis. Winterthur: ZHAW Zürcher Hochschule für Angewandte Wissenschaften
Homa, Celina. 2023. “Evaluating XAI Algorithms in Skin Cancer Classification : A Path towards Trustworthy AI Systems.” Bachelor’s thesis, Winterthur: ZHAW Zürcher Hochschule für Angewandte Wissenschaften. https://doi.org/10.21256/zhaw-29936.
Homa, Celina. Evaluating XAI Algorithms in Skin Cancer Classification : A Path towards Trustworthy AI Systems. ZHAW Zürcher Hochschule für Angewandte Wissenschaften, 2023, https://doi.org/10.21256/zhaw-29936.


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.