Search

Rare events detection is regarded as an imbalanced classification problem, which attempts to detect the events with high impact but low probability. Rare events detection has many applications such as network intrusion detection and credit fraud detection. In this paper we propose a novel online algorithm for rare events detection. Different from traditional accuracy-oriented approaches, our approach employs a number of hypothesis tests to perform the cost/benefit analysis. Our approach can handle online data with unbounded data volume by setting up a proper moving-window size and a forgetting factor. A comprehensive theoretical proof of our algorithm is given. We also conduct the experiments that achieve significant improvements compared with the most relevant algorithms based on publicly available real-world datasets. [ABSTRACT FROM AUTHOR]

The Meshfree Local Petrov-Garlerkin Method (MLPG) was introduced to study the behavior of extra-long and large-diameter piles. The nonlinear behavior of the soil and pile shaft materials was firstly analyzed, based on these, the solving steps of MLPG was put forward involving the elastic-plastic model of soil and pile shaft reinforcement. Then a testing example was analyzed in contrast to the Finite Element Method (FEM), and the results indicated that the MLPG was of higher accuracy. Lastly, the MLPG was applied to analyze the data of static test on a full scale extra-long pile to study the behavior of pile. [ABSTRACT FROM AUTHOR]

Averaged One-Dependence Estimators [1], simply AODE, is a recently proposed algorithm which weakens the attribute independence assumption of naïve Bayes by averaging all the probability estimates of a collection of one-dependence estimators and demonstrates significantly high classification accuracy. In this paper, we study the selective AODE problem and proposed a Cross-Entropy based method to search the optimal subset over the whole one-dependence estimators. We experimentally test our algorithm in term of classification accuracy, using the 36 UCI data sets recommended by Weka, and compare it to C4.5[5], naïve Bayes, CL-TAN[6], HNB[7], AODE and LAODE[3]. The experiment results show that our method significantly outperforms all the other algorithms used to compare, and remarkably reduces the number of one-dependence estimators used compared to AODE. [ABSTRACT FROM AUTHOR]

Sensor webs consisting of nodes with limited battery power and wireless communication are deployed to collect useful information from a variety of environments. A new challenge in the wireless sensor networks (WSN) is the compromised nodes problem. Compromised nodes may exhibit arbitrary behavior and may collude with other compromised nodes. In this paper, we propose a novel security strategy with assistant cluster heads, SSACH, which focuses on limiting the attack from compromised nodes. It adds the Assistant Cluster Heads (ACH) so as to monitor Cluster Head (CH) and take precautions against the inside attack. [ABSTRACT FROM AUTHOR]