As sensitivity and specificity of current microRNA identification methods are not ideal or imbalanced because of emphasizing new features but ignoring weak classification ability and redundancy of features. An ensemble algorithm based on feature clustering and random subspace method was proposed, named CLUSTER-RS. After eliminating some features with weak classification ability using information ratio, the algorithm utilized information entropy to measure feature relevance and grouped the features into clusters. Then it selected the same number of features randomly from each cluster to compose a feature set, which was used to train base classifiers for constituting the final identification model. By tuning parameter and selecting base classifiers to optimize the algorithm, experimental comparison of CLUSTER-RS and five classic microRNA identification methods (Triplet-SVM,miPred,MiPred,microPred,HuntMi) was conducted using latest microRNA dataset. CLUSTER-RS was only inferior to microPred in sensitivity and performed best in specificity, and also had advantage in accuracy and Matthew correlation coefficient. Experiments show that, CLUSTER-RS algorithm achieves good performance and is superior to the rivals in the aspect of balance between sensitivity and specificity.

To reduce the negative impact of the inherent time lag property of trust evaluation on the result of evaluation, new time lag weakening policies for enhancing trust evaluation reliability were proposed by integrating the trust trends of target entity into trust evaluation process with second derivative as the measurement tool. In order to check the effectiveness of this strategy, with the idea of reverse engineering, the authors used it to extend the famous trust revaluation model proposed by Srivatsa. The results show the parts of trust evaluation result with the extended model is closer to the real behavior of the target entity than that of the original model, and the former is more capable of inhibiting the fluctuating behavior of malicious entity. All of these indicate that the time lag weakening policies can be used to enhance the reliability of trust evaluation.

To resolve the problems of high dimensionality, low accuracy and overlapping in image clustering, an effective link-clustering based image multiple-cluster partition method was proposed in this paper. This method utilized image distance to measure similarity and identified overlapping clusters by using link-clustering. As a result, an image may be partitioned into multiple clusters, and this multiple-cluster partition makes each cluster more specific compared with others. To validate this method, experiments were carried out on the datasets returned by search engine when searching for some key words. The result shows that the proposed method can find explicit clusters with partial overlapping.