In this paper, we analyzed the shortcoming of Feature Weighted SVM and Sample Weighted SVM, then a new SVM approach is proposed based on the comprehensive feature and sample weighted. This method estimates the relative importance (weight) of each feature by discernibility matrix. It utilizes the weights for computing the inner product in kernel functions. In this way the computing of kernel function can avoid being dominated by trivial relevant or irrelevant features. Then we estimate the weight of each training samples by the feature weight and similarity between samples, in order to reduce the influence of non-critical samples and noise data on the SVM learning and improve the noise immunity. Experimental results show that comparing with simply considering the importance of feature or sample, the proposed method can more effectively improve the classification accuracy of SVM.
When a Data Stream Management System (DSMS) becomes overloaded and fails to satisfy all kinds of requirements, such as tuple latency and result precision because the arrival rates are bursty. Especially, real-time queries have to be completed within certain deadlines for results to be full of value. Semantic load shedding is an effective approach to alleviate workload. A semantic load shedding algorithm based on priority table is presented which considers about execution costs and values of tuples at the same time when deciding which tuples are dropped in this paper. Experiment results show that this algorithm has better performance and flexibility to handle workload fluctuations gracefully.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.