XOPSIS: an explainable AI method based on the order of preference by similarity to ideal solution
This thesis is submitted in partial fulfillment of the requirements for the degree of Master of Science in Computer Science and Engineering, 2023.
| Autor Principal: | |
|---|---|
| Outros autores: | |
| Formato: | Thesis |
| Idioma: | English |
| Publicado: |
Brac University
2024
|
| Subjects: | |
| Acceso en liña: | http://hdl.handle.net/10361/22640 |
| id |
10361-22640 |
|---|---|
| record_format |
dspace |
| spelling |
10361-226402024-04-22T21:02:07Z XOPSIS: an explainable AI method based on the order of preference by similarity to ideal solution Rahman, Anika Alam, Md. Golam Rabiul Department of Computer Science and Engineering, Brac University Explainable AI XOPSIS LIME SHAP Maternal health Interpretability Transparency Artificial intelligence This thesis is submitted in partial fulfillment of the requirements for the degree of Master of Science in Computer Science and Engineering, 2023. Cataloged from PDF version of thesis. Includes bibliographical references (pages 131-134). Explainable AI (XAI) techniques are essential for comprehending machine learn- ing model predictions in a variety of fields, including healthcare. In this study, we highlight XOPSIS, a newly developed XAI method built on the TOPSIS score and intended to offer thorough justifications for gradient boosting models. We com- pare the performance of XOPSIS with two established XAI techniques, LIME and SHAP, using a comprehensive dataset encompassing different domains, including maternal health records, the benchmark Iris dataset, and the benchmark breast cancer dataset. Our findings demonstrate significant similarities between XOPSIS and LIME in generating explanations and consistently identifying the most influen- tial features contributing to the model’s predictions. In addition, by utilizing SHAP values, we acquire a comprehensive comprehension of the model’s behavior and the unique contributions of each feature to the predictions. The significance of our pro- posed approach lies in its ability to enhance interpretability in machine learning models, enabling stakeholders to make informed decisions across various domains. While we showcase the effectiveness of XOPSIS in maternal health risk prediction, the benchmark Iris dataset, and breast cancer diagnosis, its applicability extends to diverse domains such as finance, cybersecurity, and customer behavior analysis. The flexibility and generalizability of XOPSIS make it a valuable tool for understanding the underlying factors driving model predictions. In addition to the maternal health records, the benchmark Iris dataset, and the breast cancer dataset, we also apply XOPSIS to the Car Acceptability dataset, further expanding its applicability across diverse domains. By including this dataset, we demonstrate the versatility of XOP- SIS in providing comprehensive explanations for machine learning models in various contexts. Furthermore, future studies should focus on the practical implementation of XOPSIS in different domains, evaluating its effectiveness in real-world scenarios, and assessing its impact on decision-making processes. Furthermore, exploring the integration of XOPSIS in various industries and applications can provide valuable insights into the interpretability and transparency of machine learning models. By advancing XAI techniques like XOPSIS, we can foster trust, accountability, and widespread adoption of AI technologies in diverse fields, ultimately benefiting both industry practitioners and end-users. The continued development and refinement of XOPSIS and similar XAI methods will contribute to the responsible and ethical use of AI, promoting transparency and understanding in complex machine learning models. Anika Rahman M.Sc. in Computer Science 2024-04-22T04:36:59Z 2024-04-22T04:36:59Z 2023 2023-07 Thesis ID 21266028 http://hdl.handle.net/10361/22640 en Brac University theses are protected by copyright. They may be viewed from this source for any purpose, but reproduction or distribution in any format is prohibited without written permission. 150 pages application/pdf Brac University |
| institution |
Brac University |
| collection |
Institutional Repository |
| language |
English |
| topic |
Explainable AI XOPSIS LIME SHAP Maternal health Interpretability Transparency Artificial intelligence |
| spellingShingle |
Explainable AI XOPSIS LIME SHAP Maternal health Interpretability Transparency Artificial intelligence Rahman, Anika XOPSIS: an explainable AI method based on the order of preference by similarity to ideal solution |
| description |
This thesis is submitted in partial fulfillment of the requirements for the degree of Master of Science in Computer Science and Engineering, 2023. |
| author2 |
Alam, Md. Golam Rabiul |
| author_facet |
Alam, Md. Golam Rabiul Rahman, Anika |
| format |
Thesis |
| author |
Rahman, Anika |
| author_sort |
Rahman, Anika |
| title |
XOPSIS: an explainable AI method based on the order of preference by similarity to ideal solution |
| title_short |
XOPSIS: an explainable AI method based on the order of preference by similarity to ideal solution |
| title_full |
XOPSIS: an explainable AI method based on the order of preference by similarity to ideal solution |
| title_fullStr |
XOPSIS: an explainable AI method based on the order of preference by similarity to ideal solution |
| title_full_unstemmed |
XOPSIS: an explainable AI method based on the order of preference by similarity to ideal solution |
| title_sort |
xopsis: an explainable ai method based on the order of preference by similarity to ideal solution |
| publisher |
Brac University |
| publishDate |
2024 |
| url |
http://hdl.handle.net/10361/22640 |
| work_keys_str_mv |
AT rahmananika xopsisanexplainableaimethodbasedontheorderofpreferencebysimilaritytoidealsolution |
| _version_ |
1814308021937897472 |