Computer and Modernization

Improved Collaborative Filtering Algorithm

LU Chun-xia, WANG Yi-zhi

2016, 0(3): 1-4+10. doi:10.3969/j.issn.1006-2475.2016.03.001

Asbtract ( 306 )

References | Related Articles | Metrics

Collaborative filtering is the most widely used recommendation technology in the personalized recommendation system. However, the rapid increase of the amount of users and data make the score matrix of user's preference information become more and more sparse, and the collaborative filtering algorithm encounters a bottleneck. The calculation of similarity is the most important step in collaborative filtering algorithm. In order to improve the accuracy of the similarity of sparse matrix, this paper improves the traditional collaborative filtering algorithm. We cluster the item set first, and then use the Slope One algorithm to fill matrix after clustering, finally introduce the degree of preference of each cluster for user as the weight. The experimental results show that the improved collaborative filtering algorithm can effectively alleviate the sparse problem of scoring matrix, so as to improve the quality of the recommendation system.

Prediction of Missile Degradation Condition Based on Improved Unequal Interval Grey Model

MA Hong-xia1, CONG Lin-hu2

2016, 0(3): 5-10. doi:10.3969/j.issn.1006-2475.2016.03.002

Asbtract ( 267 )

References | Related Articles | Metrics

Aiming at the test data having the feature of small sample and nonlinearity and low prediction accuracy using traditional UGM(1,1) in missile degradation condition prediction, the structure mode of background value in traditional UGM(1,1) is optimized and a new calculation formula of background value is designed through analyzing the modeling process of traditional UGM(1,1). Further more, a method of missile degradation condition prediction based on improved UGM(1,1) is proposed. Missile in storage condition as an example is used to realize key parameters predicting. Finally, the results validate the rationality and effectiveness of improved UGM(1,1) prediction model.

A Classification Method of Thyroid Disease Based on Rotation Forest

PAN Qiao, XU Teng, CHEN De-hua, XU Guang-wei

2016, 0(3): 11-15. doi:10.3969/j.issn.1006-2475.2016.03.003

Asbtract ( 248 )

References | Related Articles | Metrics

Thyroid disease is common in the field of endocrine, accurate identification of different types of thyroid disease is the primary problem of clinical treatment. By using the results of clinical experiments, this paper presents a new method for thyroid disease classification. The method uses principal component analysis to reduce data dimension, and then implements classification task based on rotation forest algorithm. Rotation forest algorithm can make the difference between the base classifiers more obvious, and then improve the accuracy of the classifier, and it can reduce the processing time at the same time. Experimental results show that the classification accuracy of this method can reach to 96.28% on the dataset from UCI machine learning repository. In order to verify the effectiveness of the method furthermore, this paper also chooses the real clinical medical data set, it is more complex than the UCI standard dataset in data quantity and data dimension. Compared with the other method, the classification accuracy of this method reaches to 96.37%.

An Improved Rakel Approach Based on Label Pairwise

ZHOU En-bo, YE Rong-hua, ZHANG Wei-wei, ZHOU Zi-han

2016, 0(3): 16-18+23. doi:10.3969/j.issn.1006-2475.2016.03.004

Asbtract ( 315 )

References | Related Articles | Metrics

Rakel (Random k-labelsets) randomly selects a number of label subsets from the original set of labels and uses the LP (Label Powerset) method to train the corresponding multi-label classifiers. But the models maybe have a poor performance because of randomization nature. Thus in this paper we firstly capture some pairwise relationships based on label co-occurrence between the labels to training LP classifier by PwRakel (Pairwise Random k-labelsets) algorithm. The method extends the training set by exploiting label correlations to improve classification performance effectively. The experimental results indicate that the proposed method improves multi-label classification accuracy compared with the Rakel algorithm and to other state-of-the-art algorithms.

A Graph Clustering Algorithm Based on Structural Similarity

JIN Chao, ZHANG Long-bo, WANG Hai-peng, AN Jian-rui, HUAI Hao, WANG Xiao-dan

2016, 0(3): 19-23. doi:10.3969/j.issn.1006-2475.2016.03.005

Asbtract ( 268 )

References | Related Articles | Metrics

Network clustering (or graph partitioning) is an important task for the discovery of underlying structures in networks. In this paper, a fast algorithm named GNSCAN based on structure similarity is proposed, and the related definition and algorithm implementation process are given. In order to test the performance of the algorithm, a group of real datasets are used to test the algorithm. The theoretical analysis and experimental results show that the proposed GNSCAN algorithm is improved in efficiency. On the basis of the above GNSCAN algorithm, IGNSCAN is proposed. And the computation complexity of the algorithm IGNSCAN is more efficiency.

Hail Information Extraction Based on Sina Weibo

WANG Ping, WANG He-ying

2016, 0(3): 24-29+34. doi:10.3969/j.issn.1006-2475.2016.03.006

Asbtract ( 216 )

References | Related Articles | Metrics

To obtain accurate hail information more easily and quickly, a three-level identification is designed, which is the first identification of microblog containing “hail” through Web crawler technology, the second identification of hail events based on classifier and the third identification of hail element information based on rules. In order to improve identification performance of hail events, an assessment function for extracting features is added, and a multi-assessment function to determine the feature vectors is proposed. Then a scheme based on combination of three classifiers is given. The test results show that hail events extraction rate is 89.5% by the presented method, mistaken identification rate is less than 13.4%; hail element information extraction rate is more than 96.0%, mistaken identification rate is less than 8.6%.

Large Population Parallel Genetic Algorithm Based on OpenCL

XU Pei-yan, SHI Hui-bin

2016, 0(3): 30-34. doi:10.3969/j.issn.1006-2475.2016.03.007

Asbtract ( 283 )

References | Related Articles | Metrics

In order to improve the accuracy rate of RNA secondary structure prediction and accelerate the genetic algorithm, this thesis proposed the implementation of a large population parallel genetic algorithm based on OpenCL. Through researching the potential parallelism of genetic algorithm, this thesis uses Acer TMP246M-MG-5086 as experimental platform, firstly realizes the genetic algorithm on CPU, then realizes the large population parallel genetic algorithm on GPU. Test results show that the accuracy rate of parallel genetic algorithm prediction has been increased about 49.88%, and the average speedup of using GPU is 9.76x.

Learning State Analysis Method of Students Based on Outlier Detection

LU Liu-sheng, YU Ming-hui

2016, 0(3): 35-40. doi:10.3969/j.issn.1006-2475.2016.03.008

Asbtract ( 231 )

References | Related Articles | Metrics

The student supervisors are facing a great challenge in Chinese universities that they have a lot of work to do and serve too many students directly, so that they can hardly give a personalized learning guide for every student. We propose a method of learning state analysis of students based on outlier detection to solve this problem and allocate the limited educational resources to the neediest students. This method finds the suspicious outlying students through mining the students’ scores based on the algorithm of density-based local outliers, and analyzes the learning state of these students. The case study shows that this method can efficiently find some students with exceptional learning state which may assist the college student supervisors in managing students more efficiently.

Bounded Model Checking of PTACTL Based on SMT

MAO Liang-wen1, XU Liang2,3

2016, 0(3): 41-45. doi:10.3969/j.issn.1006-2475.2016.03.009

Asbtract ( 231 )

References | Related Articles | Metrics

Probabilistic timed automata are an extension of timed automata with discrete transition probabilities, and can be used to model timed randomized protocols or fault-tolerant systems. We present a bounded model checking method on SMT for verifying probabilistic timed automata against PTACTL properties. This method adds the transition times and transition probabilities into the ACTL properties and changes the encodings of models and properties in order to verify them, which is come from the SMT-based bounded model checking. We also give two examples to show that this method is effectiveness and efficiency.

APRFODA: Adaptive Priority Ranking of Test Cases Based on Functional Occupancy and Demand Analysis

XUE Yi-fan1, MAO Yu-guang1,2

2016, 0(3): 46-51. doi:10.3969/j.issn.1006-2475.2016.03.010

Asbtract ( 275 )

References | Related Articles | Metrics

In the process of using the cover of the test case prioritization, the feature is usually expressed by the code coverage change information, and it is easy to ignore other factors. So an adaptive priority ranking of test cases based on functional occupancy and demand analysis is proposed. Firstly, it takes the reference to the call function in the process of test case prioritization, and uses the impact of the change of source code, then takes the regression testing analysis of the domain and the regression range of the test case set. Secondly, it considers the requirement of test case prioritization, and the requirement of the evaluation index is determined. Then, the weights are adaptively integrated with the call function source code change. Finally, by comparing the simulation results, the proposed method can improve the correct rate of defect detection and reduce the test cost.

Forest Fire Recognition Based on Deep Convolutional Neural Network Under Complex Background

FU Tian-ju, ZHENG Chang-e, TIAN Ye, QIU Qi-min, LIN Si-jun

2016, 0(3): 52-57. doi:10.3969/j.issn.1006-2475.2016.03.011

Asbtract ( 486 )

References | Related Articles | Metrics

According to the characteristics of forest fire, a forest fire image recognition method based on deep learning is proposed and designed. The structure of convolutional neural network (CNN) is given by experiment, which is used in forest fire recognition under the complex background, and it has been trained and tested. A parameters replacement method is presented for low recognition rate existing in small samples forest fire recognition. The results show that the method is of a high accuracy reaching to 98%, it can extract features automatically, the input image doesn’t need to pre-processing, and it overcomes many inherent shortcomings of traditional algorithm. Its application in the field of forest fire recognition achieves good results.

A Rapid Fire Recognition Method

ZHANG Yong-mei1,2,3, DU Guo-ping1,4, XING Kuo1

2016, 0(3): 58-63. doi:10.3969/j.issn.1006-2475.2016.03.012

Asbtract ( 237 )

References | Related Articles | Metrics

Intelligent fire recognition system is an important part of constructing wisdom city and preventing fire, which can well guarantee the safety of people's lives and property. Concerning the contradiction of accuracy and real-time in image fire recognition, a rapid fire image recognition method is proposed. Firstly, the segmentation method combining watershed segmentation and automatic seeded region growing is adopted to make suspected flame region segmentation in a complex environment, and the multithreading technology is used to improve the processing speed, so it is conducive to real-time fire recognition. Then, we extract the significant characteristics such as roundness, sharp corners, corrosion resistance, flame core features such as relative coordinates, the relative area in the suspected areas as fire classification, reduce the dimension of feature space and the amount of calculation. Radial basis function (RBF) neural network is adopted to complete the fire recognition, and it shortens the time of fire recognition and improves fire recognition accuracy. The experiment results show that the suspected area extraction accuracy is 90%, the fire recognition accuracy is 85%, the method can improve the precision and speed of the fire recognition.

Experiment Teaching Case of Digital Image Processing Course Based on ImageJ

ZHAO Yi-li1,2, XU Dan1, ZHANG Yan2

2016, 0(3): 64-67+73. doi:10.3969/j.issn.1006-2475.2016.03.013

Asbtract ( 271 )

References | Related Articles | Metrics

Aiming at the experimental teaching of digital image processing in computer science and technology, a new teaching method of digital image processing based on Java language and ImageJ platform is proposed. Experimental teaching of digital image processing course is designed by the students in the experimental course, so that students can complete the digital image processing problem independently. Although some students have learned the Java language, taking into account that not all students are familiar with ImageJ software, the starting point of the experimental task is usually the first to allow students to understand and test the existing ImageJ plug-in code template. Secondly, the teacher asks the students to expand the existing plug-in based on the experimental requirements in the existing digital image processing code. As the ImageJ software is open source, and itself is an open plug-in architecture system, the experimental teaching method of this kind of structure becomes possible.

An Approach of Image Tag Recommendation Based on Subspace Learning Model

QI Chao

2016, 0(3): 68-73. doi:10.3969/j.issn.1006-2475.2016.03.014

Asbtract ( 208 )

References | Related Articles | Metrics

Represented by Flickr and Picasa, online photo albums allow users to tag images, hoping to make it more convenient as well as efficient to organize and retrieve image resources. Recently, automatic tag recommendation system has become a hot research field considering the increasing request that high-quality tags be provided. In this thesis, a new method for tag recommendation system is proposed. Unlike the traditional one which only depends on frequency information or visual feature similarity while neglecting the relation between visual content and the semantic meaning contained in tags thus leading to unsatisfactory recommendations, the new method can find out a latent subspace shared by visual features and tag contents using matrix factorization. As for an untagged image, recommendations can be made when its visual features are projected into the latent subspace and the relevance level it has with others tags is figured out. This new method has been proved efficient after being tested on NUS-WIDE data set with more satisfactory results.

An Accurate Method for Extracting External Contours of Moving Objects

CHEN Hong-you, LI Yu-feng

2016, 0(3): 74-77. doi:10.3969/j.issn.1006-2475.2016.03.015

Asbtract ( 220 )

References | Related Articles | Metrics

The external contours of moving objects are important information sources for semantic analysis of moving objects in video surveillance system. For some defects of external contours through simple morphological processing of moving objects area, a more accurate method is proposed. Firstly, a roughly determined moving objects area is got through foreground detection. Secondly, the external contours area is roughly localized through the watershed method. Finally, the external contours area is accurately localized to extract accurate external contours through shadow removal and object reconstruction. By using the accuracy rate, recall rate and comprehensive performance norm, the experimental results show that it could get more accurate external contours.

A Measurement of Similarity Search on Heterogeneous Information Networks

WU Zhuan-hua

2016, 0(3): 78-84. doi:10.3969/j.issn.1006-2475.2016.03.016

Asbtract ( 271 )

References | Related Articles | Metrics

The emerging of large-scale interconnection networks such as social networks and bibliographic networks raised many challenges for similarity search, wherein the similarity measure is one of the key issues. The existing methods of similarity measure did not consider the different semantic of multiple paths on homogeneous networks. This paper proposes a novel similarity measure method based on meta path, which can search the same type of objects on heterogeneous networks. Meta path is path consisting of a series of relationships defined in the different object types, which provides a common basis for the similarity search engines on the networks. Experiments on real data sets show that compared to the disorder similarity measures, the proposed method supports fast path similarity query, and it can be widely used in social networks and e-commerce.

Universal Remote Online Status Monitoring Terminal

SHI Cheng-wei1, CAO Gui-ning1, CHEN Xiang-xian1, ZHOU Jie2

2016, 0(3): 85-90. doi:10.3969/j.issn.1006-2475.2016.03.017

Asbtract ( 240 )

References | Related Articles | Metrics

Aiming at the disadvantages of traditional remote online status monitoring terminals which are not universal and inflexible on data computing, a universal remote online status monitoring terminal is designed. By providing a variety of standard interface for hardware universality, using the configuration file for software universality, the monitoring terminal can access a variety of sensors to obtain state data, and upload them to a remote server via GPRS, to achieve the remote online status monitoring. After tests, this terminal is compatible with sensors which have GPIO, RS232 and RS485 interfaces, and can use several data computing methods including fast Fourier transform and mean function. The result proves that this terminal satisfies the universality, and can be used to monitor the online status of the remote devices.

Broken-point Continuingly-transferring Scheme of Large Files Based on HTML5

WANG Li-min, LIANG Zheng-he, DUAN Quan-feng

2016, 0(3): 91-95. doi:10.3969/j.issn.1006-2475.2016.03.018

Asbtract ( 228 )

References | Related Articles | Metrics

In Web applications, it is often needed to upload a file to the server. With current file upload methods, it is difficult to deal with large file uploading and user experience is also bad. Uploading big files often failed because of network interruption and the client had to reupload. With the development of HTML5 technology, a series of API about file operation emerged. This makes it possible to use JavaScript on the client side to slice local files and further achieve the function of file broken-point continuingly-transferring. On this basis, this paper solves timeout problem of merging files and correctness problem of the final file on the server side.

Communication Interface of Dots and Boxes Battle Platform in Computer Game

ZHANG Li-qun1, CAO Yang1, LI Sha2

2016, 0(3): 96-99+126. doi:10.3969/j.issn.1006-2475.2016.03.019

Asbtract ( 217 )

References | Related Articles | Metrics

In view of the deficiencies of traditional computer games, a dots and boxes battle platform in computer game was built. The dots and boxes battle platform in computer game and the clients of computer game program are placed in network environment. The transmission of all kinds of information is through the communication interface module of the battle platform in computer game to achieve the information interaction between the battle platform and game programs, thereby controlling both sides of the game programs and realizing the automatic game. Test results show that the design of this communication interface is safe, reliable and having a good communication performance.

RBAC-based Security Access Scheme for Demand Response

GUO Long-hua1, XIA Zheng-min1, ZHENG Sheng-jun2, WANG Hong-kai3, MA Zhi-cheng4

2016, 0(3): 100-104+110. doi:10.3969/j.issn.1006-2475.2016.03.020

Asbtract ( 229 )

References | Related Articles | Metrics

This paper studies role-based security access scheme for demand response. One time password (OTP) -based security authentication towards the participators is utilized after the access request. The security data of network devices is collected by active or passive data acquisition techniques. Security data are correlationally analyzed with historical data and role information. The participators will be granted roles and permission for the system and data resource according to the data analysis result. Comparisons and evaluation show the scheme enhances the security and flexibility at the price of additional overhead.

A Dynamic Recognition System of Unknown Malicious Programs Based on Host Characteristics

LIU Zhi-yong1, WANG Hong-kai2, LI Gao-lei3, WU Jun3, SU Ya-ting1

2016, 0(3): 105-110. doi:10.3969/j.issn.1006-2475.2016.03.021

Asbtract ( 188 )

References | Related Articles | Metrics

Characteristics of states changing before/after the execution of unknown malicious programs were analyzed, a novel host characteristics-based unknown malicious programs dynamic recognition system is developed by using virtual execution technology. All suspicious programs were redirected into the special sandbox and executed. The unknown malicious programs were recognized by real-timely monitoring and deeply analyzing files, regedits, processes, services and network systems of the virtual hosts in sandboxes. Next, according to the real-time records in the process of the execution of the unknown malicious programs, early warning strategies were produced to protect the files of the real-world scenarios from being altered or attacked. Experimental results show that the accuracy of this system for unknown malicious programs recognition has been improved significantly. Hence, it can high-efficiently prevent smart grid from being attacked by the unknown malicious programs.

Visual Simulation of Line-throwing Operation in Ship Rescue

CHEN Zhuo, ZHANG Xiao-lei, WANG Zhi-wen, XIONG Wei

2016, 0(3): 111-115+121. doi:10.3969/j.issn.1006-2475.2016.03.022

Asbtract ( 247 )

References | Related Articles | Metrics

According to requirements of the simulated training of the marine rescue, it is necessary for the visual system of rescue ship simulator to introduce the visual simulation of rescue operations in which the line-throwing operation is one of the most important works when doing the ship towing rescue process. In this study, OpenGL is introduced into the visual system developed with Vega Prime, and to achieve that, the hybrid programming is implemented by using the event issued subscription mechanism. On that basis, the visual simulation of the projectile's motion and ballistic trajectory is achieved by utilizing the internal trajectory model deduced from the energy conservation and the external trajectory model taking the air resistance into account. In consequence, the simulation effect is vivid and intuitional, and it overcomes deficiencies of the exclusive usage of Vega Prime.

A Distributed Multi-layer Context-aware Assembly Model

SONG Chen, LIU Hui-yi

2016, 0(3): 116-121. doi:10.3969/j.issn.1006-2475.2016.03.023

Asbtract ( 221 )

References | Related Articles | Metrics

An agent-based distributed multi-layer virtual assembly model is proposed. Firstly, we construct an assembly information model in the sight of each component, define the assembly knowledge layer based on a series of parameters used to describe feature object. The assembly decision layer treats the parts as individual agents, and defines a series of message templates for transforming parameters between agents. To help agents make proper decision, decision rules are summarized based on process of virtual assembly. Finally, the control layer is defined with a dynamic DOF updating algorithm, to lead users’ operation in a proper order. Experiment result shows that the proposed model can understand user behavior nicely.

Smart Controller for Homes Based on SOPC

SU Chang

2016, 0(3): 122-126. doi:10.3969/j.issn.1006-2475.2016.03.024

Asbtract ( 224 )

References | Related Articles | Metrics

The two-way interaction between users and grids tends to be appreciated more and more in the smart grid, but the intelligent level of current household electrical appliances can still not meet the requirement of smart power utilization. A design of smart controller for homes based on SOPC is proposed, with a hardware core of FPGA and more functional modules, such as energy measuring circuit, wireless communication circuit, and relay control circuit. With the help of the software system, the controller can collect power load data, communicate with user’s mobile client, and control power supply of household appliances. This system has been testified by sets of experiment to fulfill the acquirements of smart power utilization, and has the advantage of simple and compact characteristic, with a performance of intellectualizing our household appliances and optimizing rational electric power consumption.

Table of Content