Performance Of Relevance Vector Machine Research Articles

Sludge Conditioning Soft Measuement Based on GA-RVM

To solve the highly non-linear problem of sludge conditioning, the Relevance Vector Machine (RVM), was used to softly measure the sludge conditioning process.The parameters of the kernel function seriously affect the comprehensive performance of RVM. Using genetic algorithm (GA)to obtain the optimal kernel function parameters is an effective solution. The simulation results indicated that the filtering performance was basically consistent with the actual value of dosage and the model prediction result, and the target value could be well predicted. It showed that the model had a certain value for industrial application.

A novel shearer cutting pattern recognition model with chaotic gravitational search optimization

The accurate recognition of the shearer cutting pattern is the focus in fully mechanized coal mining. Hence, a new cutting pattern recognition model based on the combination of Relevance Vector Machine (RVM) and Chaotic Gravitational Search Algorithm (CGSA) is proposed. Initially, the motor operation data, including voltage, current and motor speed, are collected as the detection signal and the RVM classifier based on Bayesian framework is chosen for pattern recognition. In order to optimize the parameters in RVM, which has a great influence on the performance of RVM, the optimization algorithm Gravitational Search Algorithm (GSA) is introduced. Finally, the basic GSA is modified into CGSA with the chaotic mapping for increasing the search diversity of the algorithm. The experimental study demonstrates the advantageous performance of the proposed model even without any feature extraction operations.

Sequence-based Prediction of Protein-Protein Interactions Using Gray Wolf Optimizer-Based Relevance Vector Machine.

Protein-protein interactions (PPIs) are essential to a number of biological processes. The PPIs generated by biological experiment are both time-consuming and expensive. Therefore, many computational methods have been proposed to identify PPIs. However, most of these methods are limited as they are difficult to compute and rely on a large number of homologous proteins. Accordingly, it is urgent to develop effective computational methods to detect PPIs using only protein sequence information. The kernel parameter of relevance vector machine (RVM) is set by experience, which may not obtain the optimal solution, affecting the prediction performance of RVM. In this work, we presented a novel computational approach called GWORVM-BIG, which used Bi-gram (BIG) to represent protein sequences on a position-specific scoring matrix (PSSM) and GWORVM classifier to perform classification for predicting PPIs. More specifically, the proposed GWORVM model can obtain the optimum solution of kernel parameters using gray wolf optimizer approach, which has the advantages of less control parameters, strong global optimization ability, and ease of implementation compared with other optimization algorithms. The experimental results on yeast and human data sets demonstrated the good accuracy and efficiency of the proposed GWORVM-BIG method. The results showed that the proposed GWORVM classifier can significantly improve the prediction performance compared with the RVM model using other optimizer algorithms including grid search (GS), genetic algorithm (GA), and particle swarm optimization (PSO). In addition, the proposed method is also compared with other existing algorithms, and the experimental results further indicated that the proposed GWORVM-BIG model yields excellent prediction performance. For facilitating extensive studies for future proteomics research, the GWORVMBIG server is freely available for academic use at http://219.219.62.123:8888/GWORVMBIG.

Computer Network Security and Precaution Evaluation based on Incremental Relevance Vector Machine Algorithm and ACO

Abstract It is well-known that the correct evaluation for computer network attack is very significant to avoid computer network attack. In order to improve the generalization performance of relevance vector machine, an incremental relevance vector machine algorithm(IRVM) based on ant colony optimization (ACO) is firstly presented and applied to computer network attack evaluation in the paper. In the experiments, we collect 85 samples of KDDCUP99 datasets as our experimental data, the detection accuracy of the hybrid algorithm of ACO and relevance vector machine is 97.65%, and the detection accuracy of the hybrid algorithm of relevance vector machine is 94.12%.It is shown that the detection accuracy of the hybrid algorithm of ACO and relevance vector machine for computer network attack is higher than that of relevance vector machine.