Shiru Zhou

Research of web classification mining based on classify support vector machine

With the development and widely used of Internet and information technology, the Web has become one of the most important means to obtain information for people. According to the web document classification and the theory of artificial neural network, a web classification mining method based on classify support vector machine (SVM) is presented in this paper. The SVM network structure that used for web text information classification is established, and we use the genetic algorithm (GA) to optimize SVM parameters, thereby enhancing the convergence rate and the classification accuracy. The structure of web classification mining system based on classify support vector machine is given. With the ability of strong pattern classification and self-learning and well generalization of SVM, the classification mining method can truly classify the web text information. The actual classification results show that this method is feasible and effective.

Energy-Saving Control System of Beam-Pumping Unit Based on Wavelet Neural Network

The energy saving process for beam-pumping unit is a complicated and nonlinear system, and it is very difficult to found the process model to describe it. The wavelet neural network has the ability of strong nonlinear function approach, adaptive learning, fast convergence and global optimization. In this paper, an energy-saving control system of beam-pumping unit based on wavelet neural network is presented. We adopt a method of reduce the number of the wavelet basic function by analysis the sparse property of sample data, and use the learning algorithm based on gradient descent to train network. The parameters of energy-saving control process of beam-pumping unit are measured using multi sensors. Then the control system can control the working state of beam-pumping unit real-time. The system is used in the oil recovery plant. The experimental results prove that this system is feasible and effective.

Wireless sensor network for community intrusion detection system based on classify support vector machine

A community intrusion detection system based on classify support vector machine (SVM) is presented in this paper. This system is composed of ARM (Advanced RISC Machines) data acquisition nodes, wireless mesh network and control centre. The data acquisition node uses sensors to collect information and processes them by image detection algorithm, and then transmits information to control centre with wireless mesh network. When there is abnormal phenomenon, the system starts the camera and the classify SVM is used to recognize the face image. The SVM network structure that used for face recognition is established, and we use the genetic algorithm (GA) to optimize SVM parameters, thereby enhancing the convergence rate and the recognition accuracy. With the ability of strong pattern classification and self-learning and well generalization of SVM, the recognition method can truly classify the face. This system resolves the defect and improves the intelligence and alleviates workers working stress.

Community intrusion detection system based on wavelet neural network

A community intrusion detection system based on wavelet neural network (WNN) is presented in this paper. This system is composed of ARM (Advanced RISC Machines) data acquisition nodes, wireless mesh network and control centre. The data acquisition node uses sensors to collect information and processes them by image detection algorithm, and then transmits information to control centre with wireless mesh network. When there is abnormal phenomenon, the system starts the camera and the WNN is used to recognize the face image. We adopt a algorithm of reduce the number of the wavelet basic function by analysis the sparseness property of sample data which can optimize the wavelet network, and give the network learning algorithm. With the ability of strong pattern classification and function approach and fast convergence of WNN, the recognition method can truly classify the face. This system resolves the defect and improves the intelligence and alleviates workers working stress.

The Research of Building Logistics Cost Forecast Based on Regression Support Vector Machine

Building logistics cost forecasting is a complicated nonlinear problem, due to the factors that influence building logistics cost are anfratuous, so it was difficult to describe it by traditional methods. The support vector machine (SVM) has the ability of strong nonlinear function approach and strong generalization and it also has the feature of global optimization, in this paper, a modeling and forecasting method of building logistics cost based on SVM is presented. The SVM network structure for forecasting building logistics cost is established. Moreover, we propose a self-adaptive parameter adjust iterative algorithm to confirm SVM parameters, thereby enhancing the convergence rate and the forecasting accuracy. We discussed and analyzed the effect factor of building logistics cost. With the ability of strong self-learning and well generalization of SVM, the modeling and forecasting method can truly forecast the building logistics cost by learning the index information. The actual forecasting results show that this method is feasible and effective.

Community Intrusion Detection System Based on Radial Basic Probabilistic Neural Network

A community intrusion detection system based on radial basic probabilistic neural network (RBPNN) is presented in this paper. This system is composed of ARM (Advanced RISC Machines) data acquisition nodes, wireless mesh network and control centre. The sensor is used to collect information in the data acquisition node and processes them by image detection algorithm, and then transmits information to control centre with wireless mesh network. When there is abnormal phenomenon, the system starts the camera and the radial basic probabilistic neural network algorithm is used to recognize the face image. We construct the structure of RBPNN that used for recognition face image, and adopt the K-Nearest Neighbor algorithm and least square method to train the network. With the ability of strong pattern classification and function approach and fast convergence of RBPNN, the recognition method can truly classify the face. This system resolves the defect and improves the intelligence and alleviates workers working stress.

Modeling for mobile communication fading channel based on wavelet neural network

Aimed at the complicated and nonlinear relationship between input and output property of wireless channel and the advantages of wavelet neural network (WNN), a method for wireless channel modeling and simulation based on WNN is presented in this paper. Moreover, we adopt an algorithm of reduce the number of the wavelet basic function by analysis the sparse property of sample data which can optimize the wavelet network in a large extent, and the learning algorithm based on the gradient descent was used to train network. We discussed the fading channel model and analyzed the impact factor of little-scale fading channel modeling. With the ability of strong nonlinear function approach and fast convergence rate of WNN, the modeling method can implement the modeling and simulation of fading channel rapidly and effectively by learning the propagation characteristic information of wireless channel. The simulation result shows the feasibility and validity of modeling method.

Research of Wireless Fading Channel Modeling Based on Radial Basis Function Network

Radial basis function network (RBFN) is one of the neural networks used widely. Aimed at the complicated and nonlinear relationship between input and output property of wireless channel and the advantages of RBFN, a method for wireless channel modeling and simulation based on RBFN is presented in this paper. We construct the structure of RBFN that used for wireless fading channel modeling, and adopt the K-Nearest Neighbor algorithm and least square method to train the network. We discussed the fading channel model and analyzed the impact factor of little-scale fading channel modeling. With the ability of strong function approach and fast convergence of RBFN, the modeling method can implement the modeling and simulation of fading channel rapidly and effectively by learning the propagation characteristic information of wireless channel. The simulation result shows the feasibility and validity of modeling method.

Modeling for Mobile Communication Fading Channel Based on Regression Support Vector Machine

Aimed at the complicated and nonlinear relationship between input and output property of wireless channel and the advantages of support vector machine (SVM), a method for wireless channel modeling and simulation based on regression SVM is presented in this paper. The SVM network structure for communication fading channel modeling is established. Moreover, we propose a self-adaptive parameter adjust iterative algorithm to confirm SVM parameters, thereby enhancing the convergence rate and the forecasting accuracy. We discussed the fading channel model and analyzed the impact factor of little-scale fading channel modeling. With the ability of strong nonlinear function approach and fast convergence rate and well generalization of SVM, the modeling method can implement the modeling and simulation of fading channel rapidly and effectively by learning the propagation characteristic information of wireless channel. The simulation result shows the feasibility and validity of modeling method.

The Oil Viscous Force Measurement System Based on Embedded System

An oil viscous force measurement system using the pulling and separating method based on embedded system is designed in this paper, ARM9 core S3C2410 chip was used as the kernel control unit of this measurement system. The high-precision strain gauge weight measurement is designed to weigh the liquid, and the viscous force can be computed according to the change of the liquid weight in pulling and separating process. The paper mainly focused on the design of the measurement and the temperature compensation methods, and the recurrent neural networks is used to compute the temperature compensation of weight measurement. The experimental results show that this system has good measurement accuracy, and this system can be used to measure automatically, in order to reduce the manual errors interference.