Realization of intelligent techniques for classification of biomedical signals in the Labview

In this thesis study, new classification algorithms have been developed by using intelligent calculation methods to find out whether the patient is healthy or not from the dataset obtained using biomedical signals. These algorithms have been developed using the LabVIEW program, a language of visual programming that has become increasingly popular in the literature. The thesis work consists of two parts in general. In the first part, the performance of LabVIEW is compared with MATLAB. As the dimension reduction algorithms in the literature, Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA) and Kernel Principal Component Analysis (KPCA) were used. Support Vector Machine (SVM) is used as classification algorithms. LabVIEW was chosen because the performance of the classifications is about the same for both programming languages. Because LabVIEW has both fast and easy programmable structure and the recommended models for use in thesis work have not been done in LabVIEW before, the LabVIEW programming language was used in this thesis work. In the second part, two different classification algorithms have been proposed in order to reduce the selection difficulty of the kernel function which enables the different feature data to be separated linearly with SVM. The kernel functions are used to classify data samples that can not be distinguished linearly. By transferring the kernel functions to a non-linear mapping, the data is transformed into a dimension of features that can be separated into higher dimensional and linear dimensions. In the use of different kernel functions in Machine Learning (ML) algorithms, the success of the system for different signals also changes. This causes different kernel functions to be used for each signal. However, it is desirable to use the independent kernel function for signal analysis and to classify it with good success rate. For this purpose, the serial PCA-LDA-SVM (Principal Component Analysis-Linear Discriminant Analysis-Support Vector Machine), parallel PCA-LDA-SVM, serial KPCA-LDA-SVM (Kernel Principal Component Analysis-Linear Discriminant Analysis-Support Vector Machine) and parallel KPCA-LDA-SVM methods are proposed. Performance analyzes were performed for breast cancer data and EEG signals using the recommended methods. In addition, we used both holdout and cross validation methods to test the accuracy of the developed system and to show that the models did not make a memorizing. Closer results have shown that the developed system does not make a memorizing. With this method, when a new data which has never been used in the training and testing phase is entered into the system, the system can choose whether the data belongs to a patient or a healthy person. With this thesis, an easy and effective method for the classification and classification of healthy-patients in any biomedical work (Electroencephalography-EEG, breast cancer) is presented in a different programming language. In addition, in this thesis study, except for classification, heart rate was determined from ECG data and blood pressure was determined from blood pressure data. The results obtained from ECG data were interpreted according to American Heart Association standards and the results obtained for blood pressure data according to the standards of the World Health Organization International Hypertension Committee. Within the scope of this thesis, the algorithms performed in the LabVIEW environment and the case-finding for new data provided an innovative contribution to the literature.