Computer and Modernization

Construction of User Enneagram Model Based on BP Neural Network

HUANG Rong1, CHEN Yu-bing1, YUE Qing2, LIU Xing-lin1, WU Ming-fen1

2019, 0(03): 1. doi: 10.3969/j.issn.1006-2475.2019.03.001

Asbtract ( 213 )

PDF (1133KB) ( 108 )

References | Related Articles | Metrics

With the development of E-commerce, the study of users consumer behavior and personality is increasingly prevalent, but the personality classification is mainly based on Eysenck Personality theory and Big Five personality. Besides, consumer behavior is not combined with user personality traits when analyzing user consumption behavior. This paper puts forward an enneagram model of users based on BP neural network. Firstly, the model analyzes user consumption dimensions on the users shopping logs by the dimension of the consumer behavior that is extracted from consumer behaviours, then, analyzes enneagram for consumer behavior dimensions, finally, classifies enneagram of user by BP neural network. This paper tests the model and verifies its feasibility with the users’ shopping logs for six months provided by Tianchi big data.

Prediction of Blood Glucose Based on K-means Clustering Algorithm with RBF Neural Network

YU Li-ling， JIN Hao-yu

2019, 0(03): 9. doi:10.3969/j.issn.1006-2475.2019.03.002

Asbtract ( 245 )

PDF (1090KB) ( 179 )

References | Related Articles | Metrics

In view of the complexity and instability of blood glucose data in diabetic patients, this paper presents a short-term blood glucose prediction method based on K-means clustering algorithm using RBF neural network. Firstly, the blood glucose concentration time series collected by CGMS is filtered and normalized to improve the smoothness of the blood glucose data sequence and weaken the randomness of the original blood glucose data sequence. Then the RBF network is constructed on the processed blood glucose concentration time series. The K-means clustering is used to optimize, and the weights of the RBF network are adjusted by the least square method to obtain the predicted value of the future blood glucose concentration, thereby ensuring the accuracy of the prediction.

Progression Prediction Model of Chronic Kidney Disease Based on #br# Sparse Logistic Regression and Multiple Ensemble Algorithm

YANG Jin-shan, LI Zhi

2019, 0(03): 13. doi:10.3969/j.issn.1006-2475.2019.03.003

Asbtract ( 203 )

PDF (1178KB) ( 114 )

References | Related Articles | Metrics

Only a subset of the patients with stage 3 Chronic Kidney Disease (CKD) progresses to stage 4. By observing the clinical data, there are significant differences in physiological indicators between progressive and non-progressive patients. Firstly, a sparse logistic regression (SLR) with L1/2 regularization is proposed, and it is used to select the key factors that influence the progression of CKD. Then, the progression prediction model is built by SLR, Support Vector Machine (SVM) and Adaboost Decision Tree (BOOSTDT). In addition, stacking algorithm (STKSSD) is introduced to overcome the shortcomings of unstable generalization performance due to lack of samples. Finally, Artificial Neural Network (ANN) and Bidirectional Long Short-Term Memory Networks (BLSTM) are used to model the data respectively. The experimental results show that when 11 key features such as phosphorous,serum creatinine,and so on are selected by the SLR, the STKSSD algorithm achieves the best performance and obtains 86.97% recall rate， 92.86% precision rate, and 89.82% F1-score.

Neural Network Model Reference Control of FlexRay Vehicle Network

YANG Mei， WANG Yi， LIU Zhi-chao， ZHANG Liang-yu

2019, 0(03): 19. doi:10.3969/j.issn.1006-2475.2019.03.004

Asbtract ( 215 )

PDF (588KB) ( 84 )

References | Related Articles | Metrics

Aiming at the complexity and nonlinear characteristics of FlexRay vehicle network control system, the limited network bandwidth resources result in the uncertainty of data transmission and the delay of data transmission, which makes the FlexRay network control performance degrade when the data is transmitted at high speed. Neural network has the ability of self-learning, adaptive and global approximation. In order to improve the performance of FlexRay vehicle network control, a neural network control method based on network bandwidth utilization is proposed. First, the structure of the neural network model reference control system is analyzed. Secondly, the neural network model reference controller of the FlexRay vehicle network is designed.In the case of load, the performance of the controller is simulated by using Simulink in Matlab software. The simulation results show that the controller can effectively improve the performance of FlexRay vehicle network control and have good adaptability to the change of the parameters of the control object.

Regulatory Vehicle Recognition Method Based on Dual Camera Collaboration

WANG Xi-long, ZHANG Wei-wei, WU Xun-cheng

2019, 0(03): 23. doi: 10.3969/j.issn.1006-2475.2019.03.005

Asbtract ( 111 )

PDF (2141KB) ( 50 )

References | Related Articles | Metrics

Aiming at the difficult problem of urban traffic monitoring for regulated vehicles, an automatic capture system for regulated vehicles based on dual camera collaboration is studied. Firstly, according to the characteristics of the large side information entropy of the control model, which is easy to identify the inter-class target, the target side image is priorly selected as the feature carrier, and the convolutional neural network is used to identify the side of different models. Secondly, the coordinate mapping model between cameras is established. The BP neural network algorithm is used to locate the target vehicle’s face, which effectively simplifies the camera collaborative calibration process and quickly locates the target vehicle’s face. Finally, the license plate is identified based on the target vehicle’s facial image. The experimental results show that the detection accuracy of the proposed model is 89.94%, and the vertical height of the positioning face area and the ground truth is 0.912. The overall performance of the system reaches a high precision.

A Mobile Face Recognition Technology Based on Diffusion Speed

ZHAN Wei， ZHANG Ke， XU Huan， LUO Xian， LONG Fei, JI Yi-yi

2019, 0(03): 28. doi:10.3969/j.issn.1006-2475.2019.03.006

Asbtract ( 135 )

PDF (2004KB) ( 91 )

References | Related Articles | Metrics

This paper proposes a method for modeling the differences in illumination characteristics between living faces and fake faces using diffusion speed. Aiming at the need for face detection of mobile phone, according to the characteristics that the photo-reflection of the fake photo is more balanced and diffusion is slower than the living photo, a living detection method based on the diffusion speed model is proposed. The detection method obtains the diffusion velocity by introducing a total variation flow (TV), and uses the local velocity feature vector obtained by LSP coding (similar to LBP) as the input of the linear SVM classifier based on the obtained diffusion velocity map, and classifies the authenticity of the input image by classification. By designing multiple sets of comparative experiments, it is shown that the algorithm can obtain very good recognition results regardless of indoor environment or outdoor environment, multiple face poses and expressions, and various lighting conditions, and the LSP-based basic scheme is of high real-time and effective, can be deployed in a variety of mobile terminal devices, enable cross-platform one-button implant applications.

Action Recognition Technology Based on Improved C3D Neural Network

LIAO Xiao-dong, JIA Xiao-xia

2019, 0(03): 32. doi:10.3969/j.issn.1006-2475.2019.03.007

Asbtract ( 447 )

PDF (849KB) ( 150 )

References | Related Articles | Metrics

Although the C3D convolutional neural network proposed by Facebook can achieve good video action recognition accuracy, there is still much room for improvement in terms of speed, and the model obtained by training is too large to be used by mobile devices. This paper uses small convolutional kernels to reduce the characteristics of parameters, optimizes the existing network structure, and proposes a new action recognition scheme, which decomposes the 3×3×3 convolutional kernel commonly used in the original C3D neural network into deep convolution and point convolution (1×1×1 convolution kernel), and training tests on the UCF101 dataset and ActivityNet dataset. The results show that compared with the original C3D network, the improved C3D network accuracy is 2.4% higher than C3D, 12.9% faster than C3D in speed, and the model size is compressed to 25.8%.

A New High-performance Computer Platform Based on Material Science Research

WANG Kang1,2， YANG Yun-ping1,2, LIU Bo-ping1,2, FU Kang1,2, WU Zhi-ping1,2， YIN Min1,2

2019, 0(03): 39. doi:10.3969/j.issn.1006-2475.2019.03.008

Asbtract ( 162 )

References | Related Articles | Metrics

High-performance computing cluster as the “infrastructure” to ensure scientific research has been upgraded to national strategy. High-performance computing is widely used, especially in materials science. At present, high-performance computing platform which can provide high quality service of remoting, visualization and graphics for computing user has become the breakthrough point for the research of high-performance services. In this paper, a new type of high-performance computing platform system based on materials science research is proposed. The system is developed by Java language, B/S architecture is adopted to provide services. It implements the integration of the mainstream software of materials science research, provides a convenient way of accessing with a friendly user interface design, and combines with OpenPBS to optimize job scheduling method, so as to provide higher priority computing cases for platform users and ensure more efficient computing resources in material scientific research and applications.

Application of Hybrid Adaptive Particle Swarm Optimization #br# Algorithm in Power Economic Dispatch

CHEN Hong-wei, WANG Wan-cheng, WANG Ji-tuo

2019, 0(03): 45. doi:10.3969/j.issn.1006-2475.2019.03.009

Asbtract ( 170 )

PDF (880KB) ( 131 )

References | Related Articles | Metrics

Aiming at the lowest generation cost in power system, combining with the system balance constraints and unit operation constraints in actual power generation operation, a power economic dispatch (ED) model is established. Because the standard particle swarm optimization algorithm is easy to fall into local optimum, the final result of solving ED model by this method is not satisfactory. A nonlinear adaptive weight adjustment strategy is proposed to enhance the global search and local search ability of the algorithm, firstly, a niche optimization population strategy is introduced to make the algorithm jump out of the local optimum. Then the improved hybrid adaptive particle swarm optimization algorithm is applied to solve ED model. Finally, a numerical example shows that the efficiency of the proposed algorithm and the accuracy of the solution are improved.

Dynamic Feedback Load Balancing Algorithm of Streaming Media Cluster Based on HLS

YANG Bing-zhao, LI Ze-ping, LIU Jiang-tao

2019, 0(03): 51. doi:10.3969/j.issn.1006-2475.2019.03.010

Asbtract ( 171 )

PDF (1280KB) ( 120 )

References | Related Articles | Metrics

Aiming at large number of concurrent video request for mobile video users, in order to solve the problems that the load feedback of streaming media cluster nodes is not timely, the node selection is not accurate, and the node load is too heavy, which leads to the user satisfaction degrading, this paper puts forward an improved dynamic feedback load balancing algorithm. The static and dynamic load factors are considered to measure the service performance and current load of each node. The load weight vector is calculated by using analytic hierarchy process; meanwhile the load weight is calculated, the residual load is introduced and modified, which makes the description of cluster load conditions more accurate, improves the load gradient caused by the sudden increase of requests in the traditional dynamic feedback load balancing algorithm. The experimental results show that the algorithm achieves multi-node load dynamic balance, improves the efficiency of the nodes.

Trusted Mobile Application Market Based on Blockchain

LIAN Geng-xiong

2019, 0(03): 58. doi:10.3969/j.issn.1006-2475.2019.03.011

Asbtract ( 154 )

PDF (1085KB) ( 124 )

References | Related Articles | Metrics

It is impossible for ordinary users to identify whether the mobile application market is credible and the malicious behavior in the mobile market is difficult to trace. This paper proposes a solution for the trusted mobile application market based on blockchain. Blockchains have decentralized, tamper-proof, traceable security features based on which the publication and dissemination of applications are implemented. This solution builds a trusted bridge among developer, application market and user. It also ensures that the source of the mobile application can be verified, the integrity is verifiable, the trajectory is traceable, and malicious behavior is difficult to deny. The analysis shows that the scheme can establish a credible mobile application ecosystem and improve the security of mobile applications.

Abnormal Data Detection Method Based on Intelligent Bat Algorithm

SUN Yuan, LIAO Xiao-ping

2019, 0(03): 62. doi:10.3969/j.issn.1006-2475.2019.03.012

Asbtract ( 154 )

PDF (898KB) ( 109 )

References | Related Articles | Metrics

With the popularity of big data applications, network attacks become more serious and become the main network security problems. Aiming at the problem of network attack detection in large data environment, a network attack detection system is designed, which combines clustering with intelligent bat algorithm (DEBA). The system combines K-means algorithm with bat algorithm to classify data stream, and achieves efficient detection of abnormal data. The experimental results show that the clustering accuracy, algorithm time-consuming and false alarm rate of the system are obviously better than the K-means algorithm based on the traditional bat algorithm and the K-means algorithm based on the single network anomaly detection method.

A Method for Intercity Train Diagram Generation Based on Adaptive Genetic Algorithm

TANG Xue-qin

2019, 0(03): 68. doi:10.3969/j.issn.1006-2475.2019.03.013

Asbtract ( 217 )

PDF (826KB) ( 195 )

References | Related Articles | Metrics

By studing the intercity train diagram generation problem, a model is built to describe its solution space, determine optimal traveling order of trains and minimum total train running time. New mutation operator and crossover operator are adopted, and a modified adaptive genetic algorithm is proposed. The algorithm adopts the two-dimensional coding form of “train-interval traveling order” which can firstly determine the traveling order of trains in each interval.The algorithm is combined with the principle of breadth-priority cyclic, which can schedule train through the step of “time determine, discovery conflict, solve it”, determine the arrival and departure time of the train at each station. And then, the adaptive genetic algorithm is used for global optimization to obtain the best solution. Experiments show that the adaptive mechanism can improve the performance of the algorithm, get the best traveling order of the trains faster,and obtain a train diagram effectively.

An Improved Feature Selection Algorithm Based on Category Distinguished Words

LI Fu-xing，MENG Zu-qiang

2019, 0(03): 73. doi:10.3969/j.issn.1006-2475.2019.03.014

Asbtract ( 213 )

PDF (719KB) ( 137 )

References | Related Articles | Metrics

The traditional category distinguished words(CDW) feature selection algorithm, which takes inter-class dispersion degree and intra-class importance degree as comprehensive metrics, ignores the fact that contribution weights of the two indicators to feature scoring function are often different, and thus affects feature selection efficiency to some extent. A CDW feature selection algorithm combining with balance factor(ICDW) is proposed. During feature selection, the contribution weights of two indicators to feature scoring function are adjusted by continuously adjusting the value of the balance factor to complete more efficient feature selection. Using Nave Bayes classification algorithm for text categorization, experiments show that classification performance of ICDW algorithm not only outperforms that of CDW algorithm, but also exceeds that of ECE, IG and CHI, which are commonly used for feature selection.

Metadata Driven Distributed Data Resource Management Technology

JIAO Li1, SUN Song-zhou2, LIU Tian-xu2, ZHANG Xue-yang2, WANG Zhen-ji1， WANG Ya-jun1

2019, 0(03): 78. doi:10.3969/j.issn.1006-2475.2019.03.015

Asbtract ( 170 )

PDF (1122KB) ( 102 )

References | Related Articles | Metrics

By metadata driven method, the distributed data resource management technology is studied. The data unified organization is implemented. The high effective data sharing services are provided. After analyzing and classifying data, based on metadata definition and management, the data is indexed and cached in distributed data management. The key technologies are studied, including the distributed resource global directory construction for directory service, the high speed transportation in data cache management for reducing delay of dealing, collecting, and dispatching data. The dynamical and adaptive data scheduling way based on multi-backup is provided for resolving the data selection and dynamical refresh in distributed data resource management.

A Book Recommendation Algorithm Based on Improved Co-similarity Calculation

LI Yue, CHEN Li, WANG Huai-bin

2019, 0(03): 85. doi:10.3969/j.issn.1006-2475.2019.03.016

Asbtract ( 171 )

PDF (768KB) ( 132 )

References | Related Articles | Metrics

Recommendation system can solve the information overload in mass data and recommend content that users are interested in. User similarity calculation is a common recommendation algorithm, but the traditional algorithm only considers the similarity between user-item ratings and ignores the influence of users’ inherent characteristics. This paper presents an algorithm combining user feature similarity with user-item rating similarity, and uses F1 indicator to evaluate the efficiency of the recommendation algorithm. The experimental results show that the improved algorithm can effectively improve the recommendation effect.

Overlapping Community Detection Algorithm Based on LeaderRank

ZHU Shuai, XU Guo-yan, LI Min-jia, ZHANG Wang-juan

2019, 0(03): 90. doi:10.3969/j.issn.1006-2475.2019.03.017

Asbtract ( 293 )

PDF (691KB) ( 131 )

References | Related Articles | Metrics

There are lots of overlapping communities in the real social networks, better detection of overlapping communities in social networks is conducive to studying network characteristics and reflecting the real situation of the networks. In order to solve the problem of the randomness of COPRA in the overlapping community of multi-label propagation, this paper proposes a label propagation algorithm based on the importance of nodes. The algorithm uses LeaderRank to calculate the importance of each node in the network, and selects the nodes of high importance to expand into a group as the pretreatment of the initial label phase, uses reasonable label update order to prevent offset pretreatment phase, and then uses the contribution degree to weaken the randomness of the label selection stage. Experimental results on benchmark networks and real networks show that the algorithm improves the quality of community discovery results.

Text Feature Selection and Text Representation for Short Essays

MA Jian-hong， LIU Guang-sen， YAO Shuang， YANG Zhi

2019, 0(03): 95. doi:10.3969/j.issn.1006-2475.2019.03.018

Asbtract ( 264 )

PDF (1112KB) ( 158 )

References | Related Articles | Metrics

Due to its sparsity, real-time and non-standard features, short essay has many problems in text feature selection and text representation, which affects text classification accuracy. Aiming at the problem of high feature dimension disaster in text feature selection, a two-stage text feature selection algorithm is proposed. First, balance parameter, frequency, concentration, part of speech， and location are introduced into mutual information algorithm, and then the characteristic set with previous rank in the sorting result is initialized to train genetic algorithm to get optimal text feature set. Because the calculation of TFIDF aims at the whole corpus without considering the uneven distribution between classes, the variance is introduced when calculating the IDF formula. And the improved TFIDF formula is used to weight the Word2Vec word vector to represent the text vector. The improved algorithms are applied in the artificially constructed encyclopedic short essay corpus for experiments. Experiments show that the improved text feature selection algorithm and text representation algorithm have a 2%-5% improvement in classification effect.

Chinese Short Text Classification Based on Sentence-LDA Topic Model

ZHANG Hao1,2, ZHONG Min1,2

2019, 0(03): 102. doi:10.3969/j.issn.1006-2475.2019.03.019

Asbtract ( 388 )

PDF (629KB) ( 225 )

References | Related Articles | Metrics

The short text features are sparse and the context is strongly dependent, which leads to the traditional long text classification technology can’t be directly applied. In order to solve the problem of short text feature sparseness, a short text classification method based on Sentence-LDA topic model is proposed. The topic model is an extension of the LDA (Latent Dirichlet Allocation) model, it assumes that a sentence produces only one topic distribution. The trained Sentence-LDA topic model is used to predict the topic distribution of the original short text, thereby extend the obtained topic words into the original short text features, and complete the short text feature expansion. The SVM (Support Vector Machine) is finally used to classify the expanded short text. Experiments show that compared with the traditional method of directly representing short text based on VSM (Vector Space Model), the proposed method can effectively improve the accuracy of short text classification.

One-time Password Authentication Scheme Based on NFC #br# in the Internet of Things Environment

ZHAO Dong-hao, LU Yu, WANG Zeng-guang， ZHANG La

2019, 0(03): 107. doi:10.3969/j.issn.1006-2475.2019.03.020

Asbtract ( 191 )

PDF (697KB) ( 120 )

References | Related Articles | Metrics

Aiming at the defects of high energy consumption, large amount of calculation, and low efficiency for traditional identity authentication technology, this paper proposes an IoT(Internet of Things) terminal authentication scheme based on NFC(Near Field Communication)and OTP(One-Time Password). This scheme has the advantages of low cost and simple implementation of one-time password, and applies to terminals with partial storage and weak computing capabilities in the Internet of Things environment. At the same time, both parties of communication use NFC technology to interact, and NFC technology is used to perform conflict detection during device initialization and effectively solve the security problem of plaintext transmission of one-time passwords. Through the analysis of its performance, it can be known that the program can effectively prevent common attacks while ensuring a smaller amount of calculation and higher efficiency, which can be applied to the Internet of Things environment.

Data Security Protection of Electricity Marketing Information System

FANG Zhou, CHENG Qing, PEI Xu-bin

2019, 0(03): 111. doi:10.3969/j.issn.1006-2475.2019.03.021

Asbtract ( 209 )

PDF (1339KB) ( 123 )

References | Related Articles | Metrics

In recent years, with the improvement of cloud computing, Internet of Things, big data and mobile Internet, the classical electricity marketing system also develops the construction of big data platform. At the same time, the problem of data security is exposed during data mining. Due to there is much privacy information in the electricity marketing system, information leakage will bring huge loss to electricity company and will threaten user security, so many researches focus on the protection of data under using it normally. At present, most of the electricity companies still solve the security problem from the physical layer with authority management and firewall, but these methods are not enough in the society with highly developed information technology. This paper describes methods for electricity marketing system to ensure the safety of the database from data layer through the transformation on the original data, including data encryption and data masking. These methods reinforce the secrecy of the database by concealing sensitive information and eliminating data security risk fundamentally.

An Efficient Proxy Re-encryption Scheme with Keyword Search

HAN Xiao, ZENG Qi, CAO Yong-ming

2019, 0(03): 117. doi:10.3969/j.issn.1006-2475.2019.03.022

Asbtract ( 200 )

PDF (598KB) ( 243 )

References | Related Articles | Metrics

Public key encryption with keyword search(PEKS) is a practical cryptographic paradigm that enables one to search for the encrypted data without compromising the security of the original data. It provides a promising solution to the encrypted data retrieval issue in public key cryptosystems. As a combination of PEKS and proxy re-encryption(PRE), PRES allows a semi-trusted proxy to simultaneously re-encrypt and search a delegator’s encrypted data. In 2010, Shao et al. firstly presented a proxy re-encryption with keyword search scheme, but their scheme’s security is based on the premise of reducing computing efficiency. This paper solves this problem by presenting a new proxy re-encryption scheme with keyword search without bilinear pairings except the text algorithm. In the random oracle model, it is formally proved that the proposed scheme satisfies the trapdoor indistinguishability security and keyword ciphertext indistinguishability security. Comparison analysis shows that it is efficient and practical.

An Anonymous Network Information Publishing Scheme with Identity Traceable

CUI Yao

2019, 0(03): 122. doi:10.3969/j.issn.1006-2475.2019.03.023

Asbtract ( 162 )

PDF (698KB) ( 90 )

References | Related Articles | Metrics

In this paper, the generation algorithm for bitcoin address based on base58 encoding is studied, and based on that, an anonymous network information publishing scheme with identity traceable is proposed, on the one hand, the information of the publisher can be protected in our scheme, and the anonymous of information publisher is realized, on the other hand, once an information publisher publishes the bad information, the regulatory authorities can track down the information issuer. The efficiency and security of the scheme are analyzed.

Table of Content