模式识别与人工智能
Home      About Journal      Editorial Board      Instructions      Ethics Statement      Contact Us                   中文
Pattern Recognition and Artificial Intelligence
22 Judgement and Disposal of Academic Misconduct Article
22 Copyright Transfer Agreement
22 Proof of Confidentiality
22 Requirements for Electronic Version
More....
22 Chinese Association of Automation
22 National ResearchCenter for Intelligent Computing System
22 Institute of Intelligent Machines,Chinese Academy of Sciences
More....
 
 
2015 Vol.28 Issue.6, Published 2015-06-30
Papers and Reports    Researches and Applications   
   
Papers and Reports
481 Algorithm for Feature Extraction with Effective Medical Meaning for the Prenatal Diagnosis of Placenta Accreta
PAN Xiao-Xiao , YE Dong-Yi , YAN Jian-Ying , ZHANG Dong , YANG Dan-Lin
Due to inconspicuous clinical characteristics of placenta accreta, there is no prenatal diagnosis methods with high sensitivity and specificity in clinical medicine. In this paper, feature selection methodis introduced into the prenatal diagnosis of placenta accreta. From the view of feature correlation, a multi-objective feature optimization problem is formulated to extract features with effective medical meaning for the prenatal diagnosis of placenta accreta, and then an improved non-dominated sorting genetic algorithm II (NSGA-II) is described to solve this problem. The computational result based on real clinical data for placenta accreta shows that the proposed method can extract placenta accreta features with effective medical meaning from complex clinical data of placenta accreta. The analysis based on receiver operating characteristic (ROC) curve shows that medical meaning of the extracted features has high diagnostic values, and it can be an effective decision tool for obstetricians to study the pathogenesis of placenta accreta and to make a timely prenatal diagnosis. The study reveals that some biochemistry characteristics in real diagnosis are very important and it can provide a reliable criterion for the prenatal diagnosis of placenta accreta.
2015 Vol. 28 (6): 481-489 [Abstract] ( 623 ) [HTML 1KB] [ PDF 481KB] ( 631 )
490 The HGSD Method for Consumption Sentiment Classification
LIN Ming-Ming , QIU Yun-Fei , SHAO Liang-Shan
Aiming at the sentiment classification for Chinese consumption comments, a sentiment classification method combining dictionary semantic concept and context semanteme is proposed. Firstly, a method of extracting benchmark words set of different domains is put forword. Then, the sentiment words are extracted through the similarity of HowNet based on the unigram model. Finally, HowNet and Google similarity distance (HGSD) combining the HowNet similarity and the Google similarity distance is presented to classify the sentences, which reflects the original meaning of the word and the meaning in the context. Experiments of consumption comments on books, computers and hotels show the higher F-measure of the proposed method, and meanwhile the contrast experiment shows the effectiveness of the proposed algorithm.
2015 Vol. 28 (6): 490-498 [Abstract] ( 535 ) [HTML 1KB] [ PDF 517KB] ( 631 )
499 Learning and Ensemble of One-Order Bayesian Derivative Classifiers with Continuous Attributes
WANG Shuang-Cheng,GAO Rui,LENG Cui-Ping

Considering the efficiency and reliablity, Gaussian copula function, Gaussian kernel function with smoothing parameter, the criteria of classification accuracy and the greedy selection of attribute parent node are combined to carry out the learning, optimization and ensemble of one-order Bayesian derivative classifiers with continuous attributes. Experiment and analysis are executed by using datasets with continuous attributes in UCI dataset. The results show that the classifiers after optimization and ensemble have good classification accuracy.
2015 Vol. 28 (6): 499-506 [Abstract] ( 616 ) [HTML 1KB] [ PDF 583KB] ( 711 )
507 An Attribute Reduction Method Based on Rough Sets for Dominating Sets of Graph
TAN An-Hui , LI Jin-Jin , CHEN Jin-Kun , LIN Guo-Ping
The relationship between attribute reduction problem in rough sets and dominating set problem in graph is discussed. By constructing an information system, the attribute reduction problem in rough sets is associated with the dominating set problem in graph, so as to transformed the dominating set problem into the attribute reduction problem. Firstly, it is proved that the minimal dominating set of a graph is exactly the attribute reduction of the constructed information system. Then, a minimum dominating set algorithm based on information entropy is proposed. Finally, A practical example illustrates the feasibility and efficiency of the proposed algorithm.
2015 Vol. 28 (6): 507-512 [Abstract] ( 516 ) [HTML 1KB] [ PDF 326KB] ( 633 )
513 Fine-Grained Emotional Elements Extraction and Affection Analysis Based on Cascaded Model
SUN Xiao, TANG Chen-Yi
For the fine-grained emotional elements extraction problem in product reviews,a cascaded model combining conditional random fields (CRFs) and support vector machine (SVM) is put forward. Aiming at the recognition of sentiment objects and emotional words, the review of syntactic and semantic informations are introduced into CRFs model to further improve the robustness of feature templates in CRFs. In SVM model, the features of deep semantic information of sentiment objects and emotional words and basic emotional orientation of emotional words are introduced to improve the traditional bag-of-words model. The sentiment of <sentiment object, emotional word> word pair is classified to acquire key information from product reviews, namely triples of (sentiment object, sentiment word, sentiment trend). Experimental results show that the proposed CRFs and SVM cascaded model efficiently improves the precision of emotional elements extraction and emotion classification.
2015 Vol. 28 (6): 513-520 [Abstract] ( 432 ) [HTML 1KB] [ PDF 499KB] ( 814 )
521 Team Evolutionary Algorithm Based on PSO
CHEN Wei, XIANG Tie-Ming, XU Jie
Particle swarm optimization (PSO) is widely studied and applied due to its simple principle and easy implementation. Aiming at improving the convergence speed and the search precision, an algorithm based on PSO, team evolutionary algorithm (TeamEA), is presented.The optimization process of this algorithm is divided into two stages. At the first stage, to keep the diversity the players are divided into junior teams to optimize and the senior team is formed. At the second stage, to improve the convergence speed, only the senior team is optimized. In the process of the whole optimization, by evaluating the achievements of the players, the adjustment of step-length and the maximum space are controlled, and the coaching staff is formed to guide the progress direction of the players. Results on high-dimensional multimodal test functions validate the superiority and effectiveness of the proposed algorithm.
2015 Vol. 28 (6): 521-527 [Abstract] ( 452 ) [HTML 1KB] [ PDF 557KB] ( 488 )
Researches and Applications
528 Face Recognition Based on Two-Dimensional Neighborhood Preserving Discriminant Embedding
ZHANG Da-Wei, ZHU Shan-An
Two-dimensional neighborhood preserving discriminant embedding (2DNPDE) is proposed in this paper. 2DNPDE is supervised feature extraction algorithm based on 2D image matrices. For representing the within-class neighborhood structure and the between-class distance relationship of samples, the within-class affinity matrix and the between-class similarity matrix are constructed respectively. The projection space obtained by 2DNPDE not only makes the low dimensional embedding of data points from different classes far from each other, but also preserves the neighborhood structure of samples from the same class and the distance relationship of samples from the different classes. The experimental results on the ORL and AR face databases show that the proposed algorithm has better recognition performance.
2015 Vol. 28 (6): 528-534 [Abstract] ( 490 ) [HTML 1KB] [ PDF 446KB] ( 520 )
535 Discriminative Least Squares Ordinal Regression
YU Hai-Ben, TIAN Qing, CHEN Song-Can
Ordinal regression is a special machine learning paradigm and its objective is to classify patterns by using a between-class natural order property between the labels. Although many algorithms are proposed, the classical least squares regression (LSR) is not applied to the ordinal regression scenario. In this paper, a discriminative least squares ordinal regression (DLSOR) is proposed by using the cumulative labels and the margin-enlarging technique.Without constraints imposed on the regression function, DLSOR can embed ordinal information and expand between-class margin only through the label transformation. Thus, a high classification accuracy and low mean absolute errors can be guaranteed with the premise that the model complexity of DLSOR is consistent with that of LSR. The experimental results demonstrate the superiority of the proposed method in improving the ordinal regression performance.
2015 Vol. 28 (6): 535-541 [Abstract] ( 448 ) [HTML 1KB] [ PDF 427KB] ( 557 )
542 Human Trajectory Analysis Method Based on Hidden Markov Model in Home Intelligent Space
ZHAO Yang , TIAN Guo-Hui , YIN Jian-Qin , FAN Jian-Xia
Recognition and pre-reasoning of human activity is essential in home intelligent space. In this paper, an approach based on hidden markov model (HMM) is proposed for human trajectory analysis. Firstly, the plan is discretized into tile blocks, and HMM models for human trajectories are set up off-line. Then, aiming at the online analysis, a sliding-window-like approach is put forward to achieve real-time trajectory segmentation and activate model matching process intelligently. Finally, a predicition of human trajectory is made by the intelligent space according to matching results. Experimental results show that the proposed approach achieves good performance in real-time trajectory analysis. Furthermore, the proposed approach can help home intelligent space make wiser decision.
2015 Vol. 28 (6): 542-549 [Abstract] ( 628 ) [HTML 1KB] [ PDF 1094KB] ( 1127 )
550 Top-K Relative Community Query Method for Social Network
LI Zhi-Chao, CHEN Hua-Hui, QIAN Jiang-Bo, DONG Yi-Hong
To find top-K relative communities associated with the query point is of significance in practical research . In this paper, the concept of clique and relative community is defined, and a method to rapidly detect the top-K relative communities is explored. A down detection expansion algorithm is proposed. All the clique structures are detected from query point. By extending each clique structure outward to construct a community, the top-K relative communities of the query point is quickly acquired through loop iteration. Meanwhile, to reduce the searching space and computing time, the down detection expansion algorithm is improved. Through comprehensive experimental comparison, the validity of the original algorithm and the efficiency of improved algorithm is verified.
2015 Vol. 28 (6): 550-557 [Abstract] ( 434 ) [HTML 1KB] [ PDF 469KB] ( 549 )
558 Pedestrain Detection Method Based on Partition Ensemble
LUO Hui-Lan , PENG Kai , KONG Fan-Sheng
To improve the accuracy of pedestrain detection, an ensemble approach for pedestrian detection in still images is proposed. Firstly, a partition ensemble method is used to evenly split the entire training window to get small regions, and features of small regions are extracted. Then, the AdaBoost classifiers are trained on different regions to get part classifiers. A global classifier is formed by weighted summing of these part classifiers. More global classifiers are obtained by using different partitioning methods to repeat the process. To improve detection results and achieve better performance, two global classifiers are built by using histograms of oriented gradient, and multi-level version of HOG descriptor features respectively for each partitioning method. The classifier ensemble is used to detect new images and the weighted voting method is used to decide the final results. Experimental results show that the proposed method achieves better performance than the whole window detector on INRIA dataset.
2015 Vol. 28 (6): 558-567 [Abstract] ( 614 ) [HTML 1KB] [ PDF 2161KB] ( 652 )
568 Adaptive Boundary Approximation Prototype Selection Algorithm
LI Juan , WANG Yu-Ping
The traditional prototype selection algorithms are susceptible to pattern reading sequence, abnormal patterns etc. Aiming at these problems, an improved prototype selection algorithm based on adaptive boundary approximation is proposed by a detailed analysis of the prototype learning rule. The prototype absorption strategy of condensed nearest neighbor algorithm (CNN) is improved and the closer homogeneous boundary prototype parallel to its current nearest one is retained. Meanwhile, the prototype updating strategy is built for achieving dynamic periodic updating to the prototype set. The proposed algorithm can overcome the above mentioned issues and effectively reduce the scale of prototype set. Experiments are made on the artificial dataset and UCI benchmark dataset, and the results show that the final prototype set obtained by the proposed algorithm reflects the distribution of the original dataset much better. It improves the average reduction ratio performance, has better classification accuracy and runs faster than other algorithms.
2015 Vol. 28 (6): 568-576 [Abstract] ( 464 ) [HTML 1KB] [ PDF 1151KB] ( 938 )
模式识别与人工智能
 

Supervised by
China Association for Science and Technology
Sponsored by
Chinese Association of Automation
NationalResearchCenter for Intelligent Computing System
Institute of Intelligent Machines, Chinese Academy of Sciences
Published by
Science Press
 
Copyright © 2010 Editorial Office of Pattern Recognition and Artificial Intelligence
Address: No.350 Shushanhu Road, Hefei, Anhui Province, P.R. China Tel: 0551-65591176 Fax:0551-65591176 Email: bjb@iim.ac.cn
Supported by Beijing Magtech 　Email:support@magtech.com.cn