Current Proceedings on Technology

Yazarlar: Ramazan Tekin, Yılmaz Kaya, Necmettin Sezgin

Konular:-

Anahtar Kelimeler:Electroencephalogram,Epilepsy,Discrete Wavelet Transform,Support Vector Machines,Feture Selection

Özet: The main purpose of extracting particular features from huge raw data sets is to reduce the data size as well as generate a data specific to the problem. This preprocess is required in most of real applications. By using the features most relevant to the problem, classification, training and clustering methods can be realized quickly with a higher accuracy. However, increasing the number of features increases the number of generated data sets exponentially such as that N features generates 2N sub-clusters. To prevent this complexity the feature extraction algorithms comprise implicit heuristic strategies.In this study, concerning the classification problem of biomedical data, spectral analyses of biomedical signals were performed by using Discrete Wavelet Transform (DWT). The most presentable statistical features of these spectral components were chosen by Relief, Focus and Las Vegas filtering methods. Then, these features were applied to Support Vector Machines for classification. The performances of these filters were compared in terms of the obtained results which revealed reasonable information.

---