Optimizing Thyroid Disease Classification: A Fuzzy OWA Distance-Based CxK-NN Approach
International Conference on Intelligent and Fuzzy Systems, INFUS 2024, Çanakkale, Türkiye, 16 - 18 Temmuz 2024, cilt.1089 LNNS, ss.322-328, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Cilt numarası: 1089 LNNS
- Doi Numarası: 10.1007/978-3-031-67195-1_38
- Basıldığı Şehir: Çanakkale
- Basıldığı Ülke: Türkiye
- Sayfa Sayıları: ss.322-328
- Anahtar Kelimeler: Classification, Fuzzy LR Data, Fuzzy OWA Distance-Based CxK-NN Algorithm, Bootstrapping, Thyroid Disfunction Diagnosis
- Kırklareli Üniversitesi Adresli: Evet
Özet
Fuzzy OWA Distance-Based CxK-NN (F-OWA-CKNN) Algorithm is an advanced version of the OWA Distance-Based CxK-NN Algorithm. It’s a fuzzy classification technique that relies on Fuzzy LR Data. This approach specifically looks at K neighbors from each class to make its classifications. It utilizes the OWA Distance metric to calculate the distance between a fuzzy point being classified and its K-nearest class. A fuzzy metric is chosen to measure dissimilarity, and this metric is constructed from both spread distances and center distances, incorporating weights for enhanced accuracy. Thyroid dysfunction has led to the exploration of alternative methods that go beyond traditional statistical approaches. One such promising avenue is the integration of fuzzy logic with advanced classification techniques. In this study, F-OWA-CKNN method is used to optimize thyroid disease classification. By combining the Fuzzy Ordered Weighted Averaging (OWA) concept with a distance-based CxK-NN approach, we aim to enhance the accuracy and reliability of thyroid dysfunction diagnosis. F-OWACKNN method is applied to real-life thyroid data comprising 80 samples collected in 2014. To enhance the robustness of our results, we employed the bootstrapping technique, increasing the sample size from 80 to 1000. The increased sample size facilitates a more robust evaluation of the proposed approach. Setting the value of k to 3, we observed accuracy rates of 79% for the single linkage approach, 70% for the average linkage approach, and 60% for the complete linkage approach.