Genetic Optimized Support Vector Machine Research Articles

Stability analysis of GMAW based on multi-scale entropy and genetic optimized support vector machine

The gas metal arc welding is a complex chaotic dynamic process. To study the relationship between arc electrical signal and welding stability, the multi-scale entropy method was introduced to analyze the current signals under different welding process parameters. Under the short-circuiting droplet transition mode, the larger shielding gas flow rate led to, the more stable welding and the smaller amplitude of multi-scale entropy curves. When the welding current parameter increased gradually, the droplet transition mode changed, and the amplitude of multi-scale entropy curves increased. As the welding voltage rose, the droplet transfer frequency decreased and the multi-scale entropy increased. Furthermore, the four-class prediction of welding forming quality was studied by combining with the genetic algorithm-based support vector machine (GA-SVM). The multi-scale entropy distribution was closely related to the type and stability of short-circuiting transfer in the welding process.

Efficient Prediction Method of Defect of Monitor Configuration Software

In order to solve the problem of low efficiency in software operation, we need to research the defect prediction of monitoring configuration software. The current method has the low efficiency in the defect prediction of software. Therefore, this paper proposed the software defect prediction method based on genetic optimization support vector machines. This method carried out feature selection for the measure of complexity of software, and built software defect prediction model of genetic optimized support vector machine, and completed the research on the efficient prediction method of software defects. Experimental results show that the proposed method improves the quality of software effectively.

Optical detection of wound infection in vivo by near infrared diffuse reflectance spectroscopy

ABSTRACTDetection of bacterial infection is important for deciding optimal treatment and management for wound. In the paper, near infrared diffuse reflectance spectroscopy is proposed for detection wound infection in vivo with optical properties. A porcine model is used as experimental subject in the detection of wound infection. The spectrally resolved data of the wounds are analyzed to extract the optical properties including reduced scattering coefficient and absorption coefficient. Based on optical properties, an ensemble model of chain-like agent genetic optimized Support Vector Machine is applied to discriminate wound infection. The experimental results demonstrate the feasibility and effectiveness of the method for rapid and noninvasive detection of wound infection using optical properties.