Deniz Ustun

Diagnosis of Several Diseases by Using Combined Kernels with Support Vector Machine

Machine learning techniques have gained increasing demand in biomedical research due to capability of extracting complex relationships and correlations among members of the large data sets. Thus, over the past few decades, scientists have been concerned about computer information technology to provide computational learning methods for solving the complex medical problems. Support Vector Machine is an efficient classifier that is widely applied to biomedical and other disciplines. In recent years, new opportunities have been developed on improving Support Vector Machines’ classification efficiency by combining with any other statistical and computational methods. This study proposes a new method of Support Vector Machines for influential classification using combined kernel functions. The classification performance of the developed method, which is a type of non-linear classifier, was compared to the standart Support Vector Machine method by applying on seven different datasets of medical diseases. The results show that the new method provides a significant improvement in terms of the probability excess.

A novel and simple expression to accurately calculate the resonant frequency of annular-ring microstrip antennas

This paper proposes a novel and simple expression for effective radius of annular-ring microstrip antennas (ARMAs) obtained using a recently emerged optimization algorithm of artificial bee colony (ABC) in calculating the resonant frequency at dominant mode (TM 11 ). A total of 80 ARMAs having different parameters related to antenna dimensions and dielectric constants was simulated in terms of the resonant frequency with the help of an electromagnetic simulation software called IE3D™ based on method of moment. The effective radius expression was constructed and the unknown coefficients belonging to the expression were then optimally determined with the use of ABC algorithm. The proposed expression was verified through comparisons with the methods of resonant frequency calculation reported elsewhere. Also, it was further validated on an ARMA fabricated in this study. The superiority of the presented approach over the other methods proposed in the literature is that it does not need any sophisticated computations while achieving the most accurate results in the resonant frequency calculation of ARMAs.

Grey-based fuzzy algorithm for the optimization of the ball burnishing process

Abstract In the present study, Grey based fuzzy algorithm was used for the optimization of complex multiple performance characteristics of the ball burnishing process. Experiments have been planned according to Taguchis L16 orthogonal design matrix. Burnishing force, number of passes, feed rate and burnishing speed were selected as input parameters, whereas surface roughness and microhardness were selected as output responses. Using Grey relation analysis (GRA), Grey relational coefficient (GRC) and Grey relation grade (GRG) were obtained. Then, Grey-based fuzzy algorithm was applied to obtain Grey fuzzy reasoning grade (GFRG). Analysis of variance (ANOVA) was carried out to find the significance and contribution of parameters on multiple performance characteristics. Finally, a confirmation test was applied at the optimum level of GFRG to validate the results. The results also show the feasibility of the Grey-based fuzzy algorithm for continuous improvement in product quality in complex manufacturing processes.

ANFIS model for determining resonant frequency of rectangular ring compact microstrip antennas

In this work, a model constructed with the adaptive neuro - fuzzy inference system (ANFIS) for estimating the resonant frequency of rectangular ring compact microstrip antennas (RRCMAs) in UHF band is proposed. A total of 108 RRCMAs having different parameters related to the antenna dimensions and dielectric substrate were simulated with the help of electromagnetic packaged software IE3D TM based on method of moment (MoM) to generate the data pool for training and test processes of the ANFIS model. While 96 RRCMAs were employed for training, the remainders were used for test the ANFIS model. The resonant frequencies were computed with the average percentage errors (APE) as 0.014% and 0.666% for training and test, respectively. The accuracy of proposed model was successfully demonstrated by comparing with the results of a method over the simulated data previously published in the literature. Further to inspect the validity of the ANFIS model, a RRCMA operating at 2.44 GHz was designed and fabricated for this work, and the accurate results concerning the resonant frequency were achieved.

Design of a dual-wideband monopole antenna by artificial bee colony algorithm for UMTS, WLAN, and WiMAX applications

In this study, a dual-wideband monopole antenna has been designed and developed for the universal mobile telecommunications system (UMTS), wireless local area network (WLAN), and worldwide interoperability for microwave access (WiMAX) applications. A novel approach integrating artificial bee colony (ABC) with the HyperLynx® 3D electromagnetic platform based on the method of moments has been employed to calculate the design parameters of the monopole antenna performance for the respective target frequencies and return loss. The proposed dual-wideband antenna operates in the dual-frequency ranges of 1.69–3.99 and 4.75–6.22 GHz applicable for the UMTS, WLAN, and WiMAX applications and it is fabricated on the flame resistant-4 substrate plate of 42 × 51 × 1.6 mm3. The performance of the presented monopole antenna is analyzed in terms of gain, radiation pattern, and s-parameter. The input reflection coefficient (S11) parameter and radiation pattern of the antenna are verified through the measurements. The measured values of the antenna parameters are found to match well within tolerable limits with the simulation results. The results illustrate that the presented dual-wideband monopole antenna obtained by using the ABC algorithm exhibits better performance in point of operating bands and s-parameter as compared with the multi-band antennas previously published in the literature.

A study on the performance of the hybrid optimization method based on artificial bee colony and differential evolution algorithms

Nowadays, the attraction of the optimization techniques based on artificial intelligence has increased the due to obtaining the its successful results on the difficult optimization problems in various areas. The artificial bee colony (ABC) algorithm based on swarm intelligence and the differential evolution (DE) algorithm improved by inspiring the natural biological evolution mechanism is the most popular of the artificial intelligence optimization methods. The exploration ability of these optimization techniques is generally very good. However, there is a disadvantage of these algorithms for the difficult optimization problem. The exploitation capability of the algorithms is usually not sufficient. In this study, the investigation results of the performance for an improved hybrid optimization method (HOM) to overcome their disadvantages occurred in optimization process of the difficult problems by combining ABC with DE algorithms are presented. While the improved algorithm keeps the exploration ability by retaining their standard updating strategy in the employed bees step, it enhances the exploitation skill by using powerful mutation and crossover strategies of the DE algorithm into onlooker bees step. The HOM having very high convergence speed and performance is a novel and robust optimization method based on meta-heuristic approach. In order to appraise the performance of the HOM in the testing step, a sequence of the classical benchmark function has been utilized and the performance results of the proposed method are compared with the performance of the standard DE, ABC algorithms. The reached numerical results in this study illustrated that the search performance of the presented HOM is better than the standard ABC, DE algorithms.

Usage of artificial neural network for estimating of the electrospun nanofiber diameter

At the present time, nanomaterials are used in the medicine and biology applications such as drug and gene delivery, bio-detection of pathogens, MRI contrast enhancement, tumor destruction via heating, and protein detection. Tissue engineering which is another of these applications is being increasingly popular. Because extracellular matrix (ECM) consists nano-sized structure; the usage of nanomaterials in tissue engineering enables to produce of tissue scaffolds that are more closely resemble the ECM form. Thus, the success rate increases in tissue engineering as it is provided a more favorable environment for the growth of cells. Electrospinning is a popular method among nanomaterial production ones. The diameter of the fiber produced by electrospinning technique depends on the various parameters like process, solution, and environmental parameters. In this study, an ANN model based on multilayer perceptron (MLP) is presented for predicting the average fiber diameter (AFD) of electrospun gelatin/bioactive glass (Gt/BG) scaffold. The experimental results previously published in the literature, which include one solution parameter (BG content) as well as two process parameters (tip to collector distance and solution flow rate) related to producing of electrospun Gt/BG nanofiber, have been used. The values of average percentage error between the predicted average fiber diameters and experimental ones are achieved as 3.27 %. The results obtained from the proposed model have also been confirmed by comparing with results of AFD expression reported elsewhere. It is illustrated that the AFD of electrospun Gt/BG can be accurately predicted by the model proposed here without requiring any complicated or sophisticated knowledge of the mathematical and physical background.

Tensile shear strength and elongation of FSW parts predicted by Taguchi-based fuzzy logic

Abstract This paper represents the fuzzy logic model for modeling and prediction of tensile shear strength and percent elongation of parts produced by the friction stir welding (FSW) process. A Taguchi L16 orthogonal array is used to plan and select the parameters and their levels. Weld travel speed, pin diameter and tool rotation are used as input variables. Therefore, a three-input and two-output fuzzy model is used to correlate these variables to the responses of tensile shear strength and percent elongation using the fuzzy rules generated based on experimental results. Close agreement is obtained between the fuzzy predicted and experimental results with the correlation coefficients of 0.931 and 0.895 for tensile shear strength and elongation, respectively.

Modeling and optimization of CNC milling of AISI 1050 steel by a regression based differential evolution algorithm (DEA)

Abstract The present study is aimed at finding an optimization strategy for the CNC pocket milling process based on regression analysis including differential evolution algorithm (DEA). Milling parameters such as cutting speed, feed rate and depth of cut have been designed using rotatable central composite design (CCD). The AISI 1050 medium carbon steel has been machined by a high speed steel (HSS) flat end cutter tool with 8 mm diameter using the zig-zag cutting path strategy under air flow condition. The influence of milling parameters has been examined. The model for the surface roughness, as a function of milling parameters, has been obtained using the response surface methodology (RSM). Also, the power and adequacy of the quadratic mathematical model have been proved by analysis of variance (ANOVA) method. Finally, the process design parameters have been optimized based on surface roughness using bio-inspired optimization algorithm, called differential evolution algorithm (DEA). The enhanced method proposed in this study can be readily applied to different metal cutting processes with greater and faster reliability.

Design of Butterworth and Chebyshev low-pass filter with heuristic algorithms

Heuristic algorithms which are artificial bee colony (ABC), differential evolution algorithm (DEA) and particle swarm optimization (PSO) algorithms are powerful evolutionary computation techniques for electronic circuit design. In this work, usage of ABC, FGA and PSO algorithms in the design of tenth orders Butterworth, and 1-dB Chebyshev low-pass filters are investigated and the obtained comparative results are presented. PSO algorithm obtained the effective and efficient design parameter results for Butterworth and 1-dB Chebyshev low-pass filters.