Yanheng Liu

Feature evaluation and selection with cooperative game theory

Recent years, various information theoretic based measurements have been proposed to remove redundant features from high-dimensional data set as many as possible. However, most traditional Information-theoretic based selectors will ignore some features which have strong discriminatory power as a group but are weak as individuals. To cope with this problem, this paper introduces a cooperative game theory based framework to evaluate the power of each feature. The power can be served as a metric of the importance of each feature according to the intricate and intrinsic interrelation among features. Then a general filter feature selection scheme is presented based on the introduced framework to handle the feature selection problem. To verify the effectiveness of our method, experimental comparisons with several other existing feature selection methods on fifteen UCI data sets are carried out using four typical classifiers. The results show that the proposed algorithm achieves better results than other methods in most cases.

Building a trusted route in a mobile ad hoc network considering communication reliability and path length

In a mobile ad hoc network (MANET), a source node must rely on other nodes to forward its packets on multi-hop routes to the destination. Unlike most previous studies that sought only the shortest path, our study proposes a novel trusted route that considers communication reliability and path length for a reliable and feasible packet delivery in a MANET. In most MANET routing schemes, security is an added layer above the routing layer. We introduce the concept of attribute similarity in finding potentially friendly nodes among strangers; so security is inherently integrated into the routing protocol where nodes evaluate trust levels of others based on a set of attributes. Unlike the fixed probability of dropping packets adopted in other routing mechanisms, our new forwarding rule is designed based on the attribute similarity and provides a recommended method in calculating the degree of similarity between attributes. The simulations show that the proposed routing scheme behaves better than the Dynamic Source Routing (DSR) protocol in warding off blackhole and changing behavior attacks and that it is unaffected by slander attacks. We also investigate the effects of transmission range, velocity, and number of nodes on routing performances.

Using cooperative game theory to optimize the feature selection problem

Feature selection is an important preprocessing step in machine learning and pattern recognition. Recent years, various information theoretic based measurements have been proposed to remove redundant and irrelevant features from high-dimensional data set as many as possible. One of the main disadvantages of existing filter feature selection methods is that they often ignore some features which have strong discriminatory power as a group but are weak as individuals. In this work, we propose a new framework for feature evaluation and weighting to optimize the performance of feature selection. The framework first introduces a cooperative game theoretic method based on Shapley value to evaluate the weight of each feature according to its influence to the intricate and intrinsic interrelation among features, and then provides the weighted features to feature selection algorithm. We also present a flexible feature selection scheme to employ any information criterion to our framework. To verify the effectiveness of our method, experimental comparisons on a set of UCI data sets are carried out using two typical classifiers. The results show that the proposed method achieves promising improvement on feature selection and classification accuracy.

A scalable random forest algorithm based on MapReduce

Random Forest is a popular data classification algorithm for machine learning. This paper proposes SMRF algorithm--an improved scalable Random Forest algorithm based on Map Reduce model. This new algorithm makes data classification in computer cluster or cloud computing environment for massive datasets. SMRF processes and optimizes the subsets of the data across multiple participating computing nodes by distributing. The experimental results show that the SMRF algorithm has the equally accuracy degradation but higher performance while comparing with traditional Random Forest algorithm. SMRF algorithm is more suitable to classify massive data sets in distributing computing environment than traditional Random Forest algorithm.

A virtual square grid-based coverage algorithm of redundant node for wireless sensor network

A virtual square grid-based coverage algorithm (VSGCA) is proposed in this paper. Each sensor node divides its sensing range into virtual square grids, if all the grids are covered by neighbors, the target node is redundant node. Compared with some previous algorithms, VSGCA can guarantee the coverage and connectivity in the whole network and has a computational complexity of O(nxM), which is less than most of other algorithms. Simulation results show that VSGCA can achieve better performance with fewer active nodes and lower computational complexity.

Selection of interdependent genes via dynamic relevance analysis for cancer diagnosis

Microarray analysis is widely accepted for human cancer diagnosis and classification. However the high dimensionality of microarray data poses a great challenge to classification. Gene selection plays a key role in identifying salient genes from thousands of genes in microarray data that can directly contribute to the symptom of disease. Although various excellent selection methods are currently available, one common problem of these methods is that genes which have strong discriminatory power as a group but are weak as individuals will be discarded. In this paper, a new gene selection method is proposed for cancer diagnosis and classification by retaining useful intrinsic groups of interdependent genes. The primary characteristic of this method is that the relevance between each gene and target will be dynamically updated when a new gene is selected. The effectiveness of our method is validated by experiments on six publicly available microarray data sets. Experimental results show that the classification performance and enrichment score achieved by our proposed method is better than those of other selection methods.

A trust propagation scheme in VANETs

In vehicular ad hoc networks (VANETs), a source node must rely on other nodes to forward its packets on multi-hop routes to the destination. Trust propagation is the principle by which new trust relationships can be derived from pre-existing trust relationship. Modeling for trust propagation is a fundamental block in an ad hoc environment. The main contribution of this work is a novel scheme to enhancing trust propagation in VANETs. We achieve trust routing by introducing the concept of attribute similarity to find some potential friendly nodes among strangers. Based on similarity degree we put forward a new forwarding behavior. Also we give the recommendatory methods to calculate the similarity degree of attribute. We present numerical results which demonstrate the effectiveness of the proposed trust propagation scheme. Our work appears to be the first to incorporate attribute similarity into trust routing in VANETs.

Node selection optimization for collaborative beamforming in wireless sensor networks

The communication distance and the energy of the nodes are limited in large-scale wireless sensor networks (WSNs). Collaborative beamforming is an effective way to solve such problem. However, the distribution of node location is not uniform, which leads to poor performance of the mainlobe and causes the high sidelobe level (SLL). This paper presents a novel collaborative communication method based on node selection optimization algorithm (NSOA). The method to calculate the optimal number of array nodes and to select the optimal array nodes for setting up a virtual antenna array are shown in NSOA. NSOA has the ability to select the CB nodes with optimal excitation amplitude and excitation phase by firefly algorithm to obtain the optimal radiation beampattern. In addition, energy consumption and communication delay of the nodes can be reduced. Simulation results show that the maximum SLL of the radiation beampattern obtained by NSOA is lower comparing with those obtained by the CCB and CSNA, meanwhile, the convergence rate of NSOA is faster than that of CCB. Compared with the traditional?clustering routing?algorithm,?NSOA has advantages in terms of communication delay, energy consumption, and prolonging network lifetime.

Internet worm early detection and response mechanism

In recent years, fast spreading worm has become one of the major threats to the security of the Internet and has an increasingly fierce tendency. In view of the insufficiency that based on Kalman filter worm detection algorithm is sensitive to interval, this article presents a new data collection plan and an improved worm early detection method which has some deferent intervals according to the epidemic worm propagation model, then proposes a worm response mechanism for slowing the wide and fast worm propagation effectively. Simulation results show that our methods are able to detect worms accurately and early.

Predictable Energy Aware Routing based on Dynamic Game Theory in Wireless Sensor Networks

A routing algorithm named Sub-Game Energy Aware Routing (SGEAR) modeled by Dynamic Game Theory is proposed in this paper to make better routing choices. SGEAR takes the residual energy of the nodes and the energy consumption of the path into consideration and achieves Nash Equilibrium using Backward Induction. Compared with Energy Aware Routing, SGEAR can provide stable routing choices for relaying nodes and the energy of the network can still burn evenly. Moreover, this algorithm is more suitable for being combined with sleeping scheduling scheme and thus prolongs the lifetime of Wireless Sensor Networks. Simulation results show that, combined with sleeping scheduling scheme, SGEAR has an increase of 20% in energy saving compared with Energy Aware Routing.