International Journal of Autonomous and Adaptive Communications Systems (45 papers in press)
Hotspot Prediction Of e-Commerce Network Users Based On Improved K-Nearest Neighbor Algorithm
by Gang Qiao
Abstract: Aiming at the problems of low accuracy and poor prediction effect in traditional e-commerce network users' hot spot prediction, this paper proposes an e-commerce network users' hot spot prediction method based on improved k-nearest neighbor algorithm. According to the users hotspot prediction impact indicator, the K-nearest neighbor algorithm is improved in pattern matching process. The key point method is used to remove the noise interference of the original time series. The dynamic time warping algorithm is used to measure the similarity of time series of users hotspot. The distance weight and trend coefficient are introduced according to the difference of users hotspot time series to deduce the future users hotspot and realize hotspot prediction of e-commerce network users. Experimental results show that the method in this paper greatly reduces the deviation of prediction results, which fully shows that the method has better prediction effect.
Keywords: Improved K-nearest neighbor algorithm; E-commerce network; Hotspot prediction.
Research On The Detection Of Privacy Information Sharing Behavior Of e-Commerce Users Based On Big Data
by Wei Chen, Dongmie Xia, YingJi Li, Xuan Fu
Abstract: In order to solve the problems of behavior data dimensionality reduction and confidence skewness in the detection process of traditional e-commerce users' privacy information sharing, an e-commerce users' information behavior detection method based on big data technology was proposed.Big data technology is used to complete the data storage activities in combination with MYSQL text. According to the storage big database, the storage format is divided, and the big data reduction activities are carried out. The big data reduction and dimensionality reduction operations are used to realize the big data reduction. Based on the low-dimensional data, the density point comparison of shared information is carried out, and the abnormal IP is queried according to the comparison results to realize the detection of data behavior. Experimental results show that the detection method has better effect and higher confidence in reducing the dimension of privacy big data of e-commerce users.
Keywords: Big data; E-commerce users; Privacy information; Sharing behavior; Dimensionality reduction of big data; Information density point.
Research On Remote Monitoring Method Of Smart Classroom Based On Internet Of Things
by Dacong Jiang
Abstract: In view of the problems of insufficient security and poor user experience in traditional monitoring methods of smart classroom, a new remote monitoring scheme based on Internet of things technology is proposed. In the hardware part of the paper, the remote monitoring architecture of smart classroom is constructed, the network topology, the Internet of things gateway, the data forwarding network framework and the intelligent environment monitoring function module are designed, and the data collection is realized through the establishment of database entity to provide information data for monitoring and scheduling. In the software design part, the connection program and the running program of the server and the client are designed. The experimental results show that this method has good security, and the user interface function is obviously optimized after the application of the scheme, and the short response time also makes the method have a good application prospect.
Keywords: Internet of things technology; Smart classroom; Remote monitoring; Internet of things gateway;.
Security state monitoring method for perception node in power Internet of things based on low rank model
by Rongtao Liao, Zhihua Xiao, Yixi Wang, Dangdang Dai
Abstract: In order to overcome the problem of low precision and recall in the current power Internet of things security monitoring results, a low rank model based security monitoring method for power Internet of things sensor nodes is proposed. This method constructs the security monitoring platform of power Internet of things sensing node, designs the adaptive sensing mechanism of edge node data types under counting bloom filter, and realizes the adaptive recognition of sensing node data fields. The normal observation data is described according to the low rank part, and the abnormal data is described according to the sparse part. The augmented Lagrangian method is used to optimize the objective equation and realize anomaly detection. The experimental results show that the method has high accuracy and recall, and high reliability.
Keywords: Low rank model; Power Internet of things; Perception node; Security; Monitoring.
An Evaluation Model Of e-Commerce Credit Information Based On Social Big Data
by Yun Zhang
Abstract: In order to overcome the problems of low accuracy and poor stability in the evaluation of Internet trading activities, an evaluation model of e-commerce credit information based on social big data is proposed. The model will be composed of four layers: basic data layer, synthetic data layer, random model layer and integrated learning layer. The logical structure of the model is divided into social communication big data preprocessing, credit evaluation sub model establishment, evaluation sub model integration, so as to enhance the ability of model credit division. On this basis, the credit evaluation index system is established, and the e-commerce credit information is evaluated by BP neural network method. The results of model verification show that the model has good generalization ability and accuracy, can distinguish important variables effectively and stably, can acquire e-commerce credit situation more scientifically, and can control the security situation of e-commerce credit information under the social big data environment.
Keywords: Social big data; E-commerce; Credit; Information; Evaluation; Security situation.
An adaptive prediction model for sparse data forecasting
by Xuan Yao
Abstract: Sparse data generated by the limitation of data acquisition are ubiquitous for prediction. However, the general prediction model is challenging to deal with those sparse data. Therefore, this paper aimed to propose an adaptive sparse data prediction model. Firstly, we introduced the aXreme Gradient Boosting (XGBoost) algorithm to build an adaptive prediction model to correct sparse data constantly. Secondly, the sparsity perception of the XGBoost algorithm is used for parallel tree learning. Finally, we applied the model to the PM2.5 concentration forecasting of Nanjing, China. We trained the model and adjusted the parameters to get better prediction results, and compared the prediction results with actual data to prove the feasibility of the model.
Keywords: Adaptive Prediction model; XGBoost; Sparse data; PM2.5.
Dynamic Acquisition Method Of Users Implicit Information Demand Based On Association Rule Mining
by Xiang Li
Abstract: In order to overcome the problems of low precision and poor recall in the current research results of user demand mining, a dynamic method based on association rule mining is proposed. Using association rules to get user behavior related data, analyzing user behavior through the crawler system, using different association strategies according to different business, combining with user browsing time, user interest attenuation factor to calculate user interest, build user dynamic interest model. Based on the analysis of user interest, in the initial stage of mining, support and trust are input respectively, and association rule mining algorithm is called to realize the dynamic mining of user implicit information demand. The experimental results show that the mining accuracy and recall rate of this method are higher than 95%, the whole method has strong scalability and practicality.
Keywords: Association rules; Implicit information demand; Dynamic acquisition.
Layered Routing Algorithm For Wireless Sensor Networks Based On Energy Balance
by Danxia Luo, Changan Ren
Abstract: Aiming at the shortcomings of traditional LEACH Routing Protocol in wireless sensor network data transmission applications, such as high total energy consumption, low residual energy and short network life, a hierarchical routing algorithm based on energy balance is proposed. This algorithm is based on the energy consumption model of wireless sensor network communication, and adopts the non-uniform clustering algorithm to introduce the threshold. According to the relationship between the distance between cluster head node and sink node and the threshold, the implementation method of network communication is selected. In addition, a simple correlation multi-path route is designed to realize the multi hop communication between clusters. By considering the communication cost and the residual energy value of nodes, the hierarchical route with balanced energy consumption is realized. Experimental results show that the algorithm has obvious advantages in balancing network energy consumption and prolonging network lifetime.
Keywords: Library and Information Management; Coding Information; Automatic Extraction.
Personalized Recommendation Algorithm For e-Commerce Network Information Based On Two-Dimensional Correlation
by Enwei Cao
Abstract: In view of the poor accuracy and low efficiency of the traditional e-commerce personalized recommendation algorithm, a two-dimensional correlation based personalized recommendation algorithm for e-commerce network information was proposed. Using two-dimensional correlation, categorize e-commerce user relevancy analysis to measure the personality interests of users in the electronic commerce network, e-commerce project through the Jaccard similarity coefficient, the similarity calculation between the interest spread model was constructed, differentiate the importance of data push grades, numerical characteristics of e-commerce behavior is influenced by the importance level is calculated, using the sorting result to realize e-commerce personalized recommendation. The experimental results show that the proposed method has high accuracy, diversity and efficiency.
Keywords: Two-dimensional correlation; E-commerce network; Personalized recommendation of information; Interest dissemination.
Research On Coverage Holes Repair In Wireless Sensor Networks Based On Improved Artificial Fish Swarm Algorithm
by Dongliang Li
Abstract: In WSN, holes are formed when nodes become invalid. To resolve this problem, Holes Repairing Algorithm based on Fish Swarm Optimization in Wireless Sensor Network (HRFSO) is proposed in this paper. In the algorithm, network coverage is served as the objective function, and biological behaviors of artificial fish are used to simulate nodes movements. The new actions of jumping and survival of the fittest are defined besides foraging, rear-ending and grouping to improve the convergence of optimization. Self-adaptive vision and step length are used when updating the status of artificial fish. Failed holes are repaired by moving sensor node with the shortest distance. The simulation results show that the algorithm is suitable for repairing holes with fast speed by moving fewer nodes. It can increase WSN coverage with better repairing result, faster convergence, higher accuracy, efficiency, and robustness. The results also show lifetime of the network can be prolonged.
Keywords: Network Coverage; Artificial Fish Swarm Algorithm (AFSA); Wireless Sensor Network; Hybrid network; Robustness.
A Hole Repair Algorithm For Wireless Sensor Networks Based On Virtual Attractive Force Constraint
by Ting Hu
Abstract: There are some problems in the traditional algorithm, such as long running time and poor coverage effect. In this paper, a new algorithm based on virtual attractive force constraint is proposed. Based on the virtual attractive force model of intensity-based virtual force algorithm with boundary forces (IVFA-B), aiming at the particularity of ideal distance between heterogeneous network nodes, this paper analyzes the relationship between the perception radius of two heterogeneous nodes and the optimal distance between nodes when realizing the maximum coverage in grid. By combining the best distance and the best distance threshold of virtual force algorithm, the adaptability of heterogeneous network is provided. At the same time, the node moving probability is introduced into the nodes moving distance formula to repair the hole in wireless sensor network node. The simulation results show that the proposed algorithm can achieve better coverage effect and reduce the running time effectively, which proves that the proposed algorithm has better application performance.
Keywords: Virtual attractive force constraint; Wireless sensor; Hole repair of network node.
Vulnerability Detection Of The Authentication Protocol In The Iot Based On Improved Wavelet Packet
by Shihong Chen
Abstract: In order to overcome the problems of long detection time and large detection error in traditional vulnerability detection methods for the authentication protocol in the IOT, this paper proposes a new method based on improved wavelet packet for vulnerability detection of the authentication protocol in the IOT. This method uses the improved wavelet packet to preprocess the data packet and form a small amount of original data. Combined with the method of protocol state diagram, it improves the coverage of traversal path and the effectiveness of trial cases. At the same time, it uses the method of sending TCP detection packets to detect whether there is vulnerability in the IOT authentication protocol. The experimental results show that the proposed method can effectively reduce the detection time and improve the detection accuracy, with the highest detection accuracy of 98.2%.
Keywords: Improved wavelet packet; IOT; Authentication protocol; Vulnerability detection; Traversal path.
State Scheduling Method Of Redundant Nodes In Power Communication Network Based On Least Square Method
by Liang Ma, Jie Zhou, Bintai Xu, Youxiang Zhu, Yingjie Jiang
Abstract: In order to overcome the problem of large energy consumption in traditional scheduling methods, a state scheduling method based on least square method is proposed for redundant nodes in power communication network. This method can identify and mark redundant nodes and obtain the location information of adjacent nodes in power system environment. Using the least square method and iterative method to find the location coordinates of redundant nodes in the power communication network, building the basic power communication network model, according to the work requirements of redundant nodes in the power communication network, to achieve the scheduling of redundant nodes. The experimental results show that the average energy consumption is 0.16kj less than that of the traditional method, which has better performance of coverage quality in the monitoring process and can extend the network monitoring time in the later stage of operation.
Keywords: Power communication; Communication network; Redundant node; Node state; State scheduling.
Optimum Design of Distance Education Assistant System based on Wireless Network
by Zixiang Yan
Abstract: Due to the constraints of various environments, the existing distance education assistant system can not meet the needs of the present stage. Aiming at the above problems, a new distance education assistant system based on wireless network is designed. Firstly, the function of the hardware part of the distance education assistant system is designed, the functions of several subsystems are introduced, and the business process of the hardware part of the system is analyzed. Combining the video and audio signal coding technology in the software design, the characteristics of the editing code are analyzed, and the software part of the system is optimized by using MMX technology. The simulation results show that the proposed system effectively reduces the response time of the system, improves the stability of the system, lays a solid foundation for the stable operation of the system, and realizes the optimization of the distance education assistant system.
Keywords: Wireless network; Distance education; Assistant system; Optimization.
Collaborative Variational Factorization Machine For Recommender System
by Jiwei Qin, HongLin Dai
Abstract: At present, the recommendation systems are confronting the huge challenge of data sparsity and high complexity of algorithm. Like the traditional collaborative filtering recommendation methods, they are difficult to adapt to the data sparse environment, resulting in low prediction accuracy. To address the aforementioned issues, this paper presents a novel Factorization Machine based on Collaborative filtering framework called Collaborative variational Factorization Machine (CVFM) that considers the user-user relations with the interaction data for Recommender systems. First, the user-item explicit ratings are used to build the user-user relationship by the similarity calculation. Next, we develop a variational Factorization Machine with a linear process to exploit the inherent relationship of latent variables from interaction information. The experimental results on three different datasets show that the presented CVFM is superior to other popular methods in prediction accuracy, at the same time, maintain the stability of our algorithm with dealing with sparse data.
Keywords: Service recommendation; factorization machine; collaborative filtering; Service calculation.
Integrated Radar Radio: Enabling technology for Smart Vehicle of Smart Cities
by MITHUN CHAKRABORTY, Debdatta Kandar, Bansibadan Maji
Abstract: The growing technological development in the field of information and communication technology has evolved the futuristic concept of smart cities, wherein the objects, embedded with high speed processors and memory, would be intelligent in the sense that they are capable to communicate among each other and can take decision. The smart cities will ensure increased road safety, traffic mobility, sustain environment and economic development. To ensure these features smart vehicle becomes an integral component of the smart cities. The smart vehicles should be equipped with simultaneous intelligent sensing and communication technologies at the back end to enable for increased road safety, traffic mobility etc. This requires the joint operation of radar and communication without interference. The aim of the paper is to develop an integrated radar radio platform without interference between the radar and the radio, facilitating smart vehicles. The concept substantiated here for integrated radar radio
Keywords: OFDM; radar radio; UWB; IV; IVC; V 2 V; V 2 I; FMCW; ICI; mmW.
Network Dynamic Routing And Spectrum Allocation Algorithm Based On Blockchain Technology
by Jue Ma
Abstract: To overcome the problems of low resource utilization rate and high bandwidth blocking rate of traditional network dynamic routing and spectrum allocation, a network dynamic routing and spectrum allocation algorithm based on blockchain technology is proposed. In this algorithm, a hybrid integer linear model of network dynamic routing and spectrum allocation is constructed to minimize spectrum consumption and frequency. Based on the extended static heuristic algorithm of blockchain, the link with the largest load is selected to optimize the spectrum allocation, and the linear model and extended static heuristic algorithm are combined to update the frequency gap state of the link where the path is located, so as to achieve the purpose of dynamic routing and spectrum allocation of the network. The experimental results show that the spectrum utilization rate is as high as 99.66%, and the bandwidth blocking rate is as low as 0.
Keywords: Blockchain technology; Network dynamic routing; Spectrum allocation; Bandwidth blocking.
Security Key Distribution Method of Wireless Sensor Network Based on DV-Hop Algorithm
by Fei Gao
Abstract: In order to overcome the problems of low security connectivity and poor distribution accuracy of traditional key distribution methods for network security, this paper proposes a security key distribution method for wireless sensor networks based on DV-Hop algorithm. In this method, the improved DV-Hop algorithm is used to locate the network security key distribution points, and the distributable points are separated according to the location results. According to the separation results, the key management tree is introduced to manage the distributable points in a centralized way, and the key management tree is used to complete the authentication, key distribution and update of wireless sensor network equipment. The experimental results show that the energy consumption of key establishment and update is low, and the minimum energy consumption of key update is only 25 ? J, which has strong anti-attack performance and high overall security.
Keywords: DV-Hop; Wireless sensor network; Key management tree; Key distribution.
Detection Of Malicious Rank Attack Nodes In Communication Network Based On Windowed Frequency Shift Algorithm
by Hao Yang, Yibo Xia, Wen Cai, Xin Xie
Abstract: In order to overcome the problem of low detection efficiency and accuracy in the existing detection methods of malicious nodes in communication networks, a detection method of malicious Rank attack nodes in communication networks based on windowed frequency shift algorithm is proposed. The original signal samples are collected, and the windowed signal spectrum is obtained by windowed truncation and DTFT processing. The frequency shift of signal is calculate, the direction of frequency shift is judged, the amplitude and frequency parameters of sampling signal are calculated, and the abnormal detection of communication network signal is realized according to the calculation results of parameters. The experimental results show that compared with the traditional methods, the proposed method has higher detection efficiency and accuracy, the highest detection rate can reach more than 98%, which can effectively protect the security of the communication network.
Keywords: Windowed frequency shift; Communication network; Malicious Rank attack; Node detection.
Performance of RPL under various mobility models in IoT
by Spoorthi Shetty
Abstract: The Internet of Things is a system used primarily for communication where various devices are connected for the collection, analysis and execution of the task required The main challenge in IoT device is, they are resource-constrained Hence, usage of an effective data transmission routing protocol plays an vital role in IoT It is identified from the research that, IPv6 Routing Protocol for Low Power and Lossy Networks (RPL)is an effective routing protocol for static IoT network Along with static network, it is necessary to evaluate the effectiveness of the RPL for different mobility models The energy consumption of the Reference Point Mobility Model (RPGM) is compared in this document with the Column Mobility Model (CMM) for RPL at distinct concentrations of salability using Cooja simulator with Contiki operating system By the extensive experimental analysis, it is identified that the CMM is more energy efficient than the model of RPGM model.
Keywords: Reference Point Group Mobility Model; Column Mobility model; Internet of Things; Routing Protocol for Low power and Lossy networks.
PREDICTION OF BIRD SPECIES USING RANDOM FOREST ALGORITHM-INTERNET OF BIRDS
by VIMAL SHANMUGANATHAN, Kaliappan M, Vijayalakshmi K, Muthulakshmi S, Selva Ishwarya
Abstract: In our routine life, we tend to stumble upon several birds. Bird-watching may be an in-style hobby that offers relaxation in way of life. Infinite individuals look at the class of various bird species while visiting bird sanctuaries., to make the bird watchers easy tool for developed where we can assist birders to acknowledge 60 bird species however we tend to can not ready to acknowledge the kind of that bird species. To beat this downside we tend to stumble upon an answer of building a package as a project. From DCNN formula may be foreseen at 88. We can notice additional correct and stable prediction of the image exploitation random formula in Jupyter notebook.
Keywords: image recognition; random forest algorithm; internet of birds; deep learning; DCNN.
Research On Parallel Association Rules Mining Of Big Data Based On Improved K-Means Clustering Algorithm
by Li Hao, Tuanbu Wang, Chaoping Guo
Abstract: In order to overcome the problems of time-consuming, low precision and redundant rules in association rules mining of big data, a parallel association rule mining method based on improved K-means clustering algorithm is proposed. This paper introduces the matter-element theory of extension, combines matter-element theory and database, and constructs the matter-element relation database model of extension, to realize the mining of parallel association rules of big data on the basis of extension. Redundant algorithm and equivalent transformation are used to eliminate redundant association rules. The experimental results show that the proposed method has high mining efficiency, high mining accuracy and high rule association, which proves that the proposed method has better application performance.
Keywords: K-means clustering algorithm; Association rules; Data mining; Redundancy algorithm; Equivalence transformation.
Dynamic Key Distribution Method For Wireless Sensor Networks Based On Exponential Algorithm
by Yun ZHAO, Yong XIAO, Weibin LIN, Chao CUI, Di XU
Abstract: In order to overcome the problem of low robustness of key distribution of wireless sensor networks, a dynamic key distribution method for wireless sensor networks based on exponential algorithm is proposed in this paper. In this method, the collusion characteristics of newly added and cancelled nodes in wireless sensor networks are used to establish the wireless sensors security model. Based on the wireless sensors security model, the exponential algorithm is used to achieve dynamic key distribution through the five indicators of initialization, session key self-repair, session key mutual repair, joining node and withdrawing node. The experimental results show that when the number of dynamic key nodes is 600, the probability of communication failure is 47%; when the number of hops is 10, the energy cost is only 1.64mJ, and the network robustness is high.
Keywords: Exponential algorithm; Wireless sensor; Network; Dynamic key; Distribution; Method.
Heuristic Positioning Method Of Intrusion Nodes In Sensor Networks Based On Quantum Annealing Algorithm
by Yun ZHAO, Ziwen CAI, Tao HUANG, Bin QIAN, Mi ZHOU
Abstract: In order to overcome the problems of low positioning accuracy and long time-consuming in traditional heuristic positioning methods, a new heuristic positioning method of intrusion nodes in sensor network based on quantum annealing algorithm is proposed. This method analyzes the result graph of sensor network and node system, selects the multi-communication radius method to communicate and broadcast among each sensor node, at the same time, refines the number of hops of nodes, and selects the weighting factor to calculate the average hopping moment of unknown nodes. On the basis of the above, through quantum tunneling effect, combined with quantum annealing algorithm, the heuristic positioning of intrusion nodes in sensor network is completed. The simulation results show that the proposed method can effectively improve the positioning accuracy and reduce the running time. The maximum positioning time is only 0.2min.
Keywords: Quantum annealing algorithm; Sensor network; Intrusion node; Heuristic positioning.
Research on Abnormal Data Recognition Method of Optical Network Based on WIFI Triangular Location
by Bingchen Lin
Abstract: In order to overcome the problems of low recognition accuracy and poor reliability of traditional optical network abnormal data identification methods, a new optical network abnormal data recognition method based on WiFi triangulation positioning is proposed. Time series analysis method is used to analyze the channel model of optical network to obtain the temporal characteristics of abnormal data in optical network. Hyperbolic frequency modulation decomposition method is used to detect the time domain characteristics of abnormal data, and the total energy of abnormal data in time and frequency domain is obtained. The abnormal data signal model is established by the energy density characteristics of abnormal data, and the specific position of abnormal data in the abnormal data signal model after filtering is identified by using WiFi triangle positioning algorithm. The experimental results show that the accuracy of the method is higher than 95%, and the recognition performance is good.
Keywords: WiFi; triangulation; channel model; total time-frequency energy; energy density characteristics.
The Traffic Jam Management and Prediction using IoT & Deep Learning technique for Smart City Infrastructure
by Shubham Gupta, Shreyansh Dixit, Rohit Sharma
Abstract: In todays era, the roadways connect two cities and even countries. In the rapidly shifting world, we need to get our work done in a split of second but, due to the on growing population especially country like India has a plethora of vehicles and commuters who travel by road, these can be of different varieties, from four-wheeler to a two-wheeler. But, due to the huge population of vehicles around us, the traffic density on roads is increasing and this leads to traffic jams which are very frequent and can extend up to even hours to get into a smooth flow again. Thus, we require some specialized techniques to solve this issue which we have shown in our paper using YOLOv3 and some statistical-based analysis using images and video files for traffic prediction. This will help in ensuring traffic management by alerting the authorities on time to take necessary actions.
Keywords: IoT; Deep Learning; Artificial Intelligence; Big Data Analytics; Security; Smart City; Visualization; Traffic Jam.
Response Efficiency Optimization of Data Cube Online Analysis for Network user's behavior
by Hui Zhang, Su Zhang, Xiaoling Jiang
Abstract: Data cube plays an important role in online analysis and processing of multi-dimensional data warehouse. Aiming at the problem of long response time and poor compression performance of data query in current methods, the optimization performance of the method is reduced, and a response efficiency optimization method for online analysis data cube based on formal concept lattice is proposed. Firstly, the data of network user's behavior is analyzed and combined with the access frequency of network user. Secondly, the time-varying and stability of data warehouse are analyzed in detail. Finally, slicing and dicing operations in online analysis are analyzed. The experimental results show that the proposed method has a shorter query response time and can quickly retrieve data encoding with better compression performance when the number of fact tables and dimension tables is increasing.
Keywords: Network user's behavior; Data cube; Online analysis; Response efficiency optimization.
Social Media Based Deep Auto-Encoder Model for Clinical Recommendation
by Kretika Tiwari, Dileep Singh
Abstract: In recent years, systems that use deep learning and patient clinical information for drug and ADR recommendations have become one of the research hotspots in the medical community However, it is still a mind hunting task for the clinical communitys to establish a model that combines the recommendation system as a hybrid This paper proposed a hybrid model that expands deep self-decoder and Top-k co-patient information by constructing a joint optimisation function, as SAeCR For extracting implicit clinical semantic information, the network representation learning method is used To evaluate the SAeCR models performance, three experiments have been carried out on two real social network data sets The experimental results show that the proposed model performs better than the other classification technique in a more sparse and more extensive data set Furthermore, the social network information can better identify the clinical relationship between co-patient.
Keywords: Adverse Drug Reaction · Collaborative filtering · Deep Learning · Drug recommendation · Clinical Recommendation System · Recommendation System · Social Media.
IoT Based Vehicular Accident Detection Using Deep Learning Model
by Ishu Rani, Bhushan Thakre, K. Jairam Naik
Abstract: With increase of population and running valuable time, the demand for cars has skyrocketed creating an unprecedented condition in spite of traffic risks and road collisions. The crashes are growing at an unprecedented pace hence it causes death. Now, when Machine Learning has taken over, the previously complex problems have become feasible, and the real-life applications of these artificial ML models have been very promising. In this article, a learning model that learns over an image dataset, thereby classifying never before seen images and data has been proposed. It aims at classifying the real-time accidents based on the level of damage. For that an ANN is utilized to train the model and to learn the similarities among images and accident data. The proposed solution is efficient as it was tried to improve the efficiency of existing model using certain literature mentioned, augmenting different extractions and leaning techniques.
Keywords: Vehicles; Accident detection; Classification; Accuracy; Deep Learning; IoT; Training model; Image polarity.
Twitter Sentiment Analysis using Ensemble Classifiers on Tamil and Malayalam Languages
by Gokula Krishnan V, Deepa J, Pinagadi Venkateswara Rao, Divya V
Abstract: The proliferation of social network is generating a huge amount of texts and drawing attentions Sentiment Analysis (SA) extracts useful information from such data Maximum researches on SA have been done on the English language, but others main languages such as Tamil and Malaya requests obligation too It is pivotal to work on Tamil and Malaya social posts because it is the most spoken language by native speakers and heavily used in social media Although such a crowd, modest work has been done on different languages SA This paper proposes to automatically classify the overall polarity of sentiments expressed in Tamil and Malaya tweets posts by Twitter users in three classes: Positive, Negative and Neutral, and determine a fruitful approach to solve this problem Two samples of Tamil and Malaya languages are collected and later divided into two different types of corpuses Each sample in both corpuses is annotated
Keywords: English Language; Malaya; Polarity; Tamil; Twitter; Sentiment Analysis; Long-Short Term Memory; Sentiment Analysis; Ensemble Classifiers.
A multi-level autopoietic system to develop an artificial embryogenesis process
by Rima HIOUANI, Nour Eddine DJEDI, Sylvain Cussat-blanc, Yves Duthen
Abstract: This paper presents a new model for the self-creation of an artificial multicellular organism from one cell, which is inspired by The Autopoietic System Theory at different levels. This theory has been proposed to define the universal self-organization and the self-creation of living systems, the use of this concept allows the development of the artificial organism as a closed organization, and it has been widely used to understand the living systems and their capabilities such as self-organization, self-creation, autonomy, evolution, reproduction... We proposed MLAS Multi-level Autopoietic System" beside the self-organization to embody this autopoietic system. However, in contrast to the proposed system by Varela, we set it up according to various levels (organs autopoietic machine, tissues autopoietic machine, and cells autopoietic machine). Inside the level of cell autopoietic machine, we proposed the second contribution in this paper, which is a Boolean Artificial GRN with an epigenetic part; lead the cells to create its history during evolution.
Keywords: Autopoietic System; self-organization; self-adaptation; Artificial Gene regulatory network; evolution; diversity.
Unsupervised learning of local features for person re-identi?cation with loss funciton
by Lunzheng Tan, GuoLuan Chen, Rui Ding, Xia Limin
Abstract: Many methods for person re-identi?cation focus on making full use of local features, which typically requires either a comprehensive manual labeling or complex pretreatment. This paper proposes a novel loss function, termed feature channels dropout and de-similarity loss, which drives the autonomous learning of discriminative local features in Convolutional Neural Networks. The proposed loss function consists of two components. The first is a feature channels dropout component designed to compel each feature channel to be discriminative. A novel channel-dropout function and a cross-channel-element-max function are applied in this component. The second component is a de-similarity component that uses Pearson correlation coe?cient to constrain feature channels and ensure they differ from each other. This component is conducive to diverse local features mining. Extensive experiments on three large-scale re-identification datasets demonstrate that the feature channels dropout and de-similarity loss achieves superior performance compared with state-of-the-art methods.
Keywords: Person re-identi?cation; Local feature; Unsupervised learning; Loss function.
Fingerprint Liveness Detection Approaches: A SURVEY
by Mingyu Chen, Chengsheng Yuan, Ying Lv
Abstract: In contemporary society, with the popularity of smart wearable devices, people are more inclined to use convenient and efficient identity verification based on biometrics. Human fingerprints are one of the most commonly used biometric factors, which have the advantages of uniqueness, convenience and security. Compared with traditional password authentication, fingerprint authentication system has extremely high security, but it is still vulnerable to fingerprint spoofing attacks. Counterfeiters can imitate user fingerprints by using various human body simulation materials, thereby realizing illegal authentication and infringing user rights and interests, so liveness detection is quite necessary. According to the fingerprint image and biometric information obtained by the sensor, the fingerprint liveness detection (FLD) can distinguish whether the fingerprint is from a real person. This paper reviews the development history and the latest progress in the field of FLD. Both hardware and software based state-of-the-art methods are thoroughly presented to help researchers to carry out further research.
Keywords: Fingerprint Liveness Detection; Biometrics; Understand; Software; Hardware.
Face Forgery Detection with Cross-Level Attention
by Yaju Liu, Jianwei Fei, Peipeng Yu, Chengsheng Yuan, Haopeng Liang
Abstract: Currently, face videos manipulated using deep learning models are widely spread on social media, which violates personal privacy and may disturb social security. In this study, we start by discovering the essential differences between real and fake faces. To extract Multi-scale artifacts and increase the perceptual field of the downsampling layer, we introduce atrous spatial pyramid pooling (ASPP). Considering the drawback that ASPP does not use all pixels for computation and may lose information, we design a Cross-Level Attention(CLA) module to interact with the output of the ASPP block with the backbone. Our proposed attention mechanism allows the network to focus on locally manipulated areas without destroying other features of the model. Experimental results on the large publicly available facial manipulation database Faceforensics++ show that our method effectively improves detection accuracy and generalization, and confirms that great detection performance is achieved even for compressed images.
Keywords: Face forgery detection ASPP Attention mechanism.
Analysis and optimization of RON loss via compound variable selection and BP neural network
by Yunshu Dai, Jianwei Fei, Fei Gu, Chengsheng Yuan
Abstract: The loss of octane in gasoline refining process can cause huge economic losses. Reducing the loss of octane has high practical significance. However, octane loss involves many operations in gasoline refining process, which are coupled with each other and have a highly nonlinear relationship with octane loss. Therefore, the analyze and optimization of octane loss is a high-dimensional nonlinear programming problem. Therefore, this paper proposes a compound variable selection scheme. Based on the selection of independent variables by outlier filtering and high correlation filtering, the representative operations are selected by random forest and grey correlation analysis, and the octane loss is predicted by BP neural network and XGBoost algorithm. To optimize the octane loss, an operation optimization scheme based on fast gradient modification is proposed. Based on the octane loss prediction network, the main operations are gradually fine-tuned to reduce the octane loss.
Keywords: RON loss optimization; variable selection; XGBoost; BP neural network.
A Survey on Neural Network-based Image Data Hiding for Secure Communication
by Yue Wu, Peipeng Yu, Chengsheng Yuan
Abstract: Data hiding has always been a hot research topic in the field of information security, and has attracted more and more attention from the academic community. At the same time, the rise of deep learning has also injected new development directions into the field. According to the characteristics of data hiding for images, many scholars have made corresponding improvements to the neural network and achieved many creative results. This review summarizes the main methods and representative research results of data hiding for images based on neural network. The principles and methods of neural network-based steganography and watermarking are introduced in detail, Finally we discuss problems of existing research and point out the direction for further research.
Keywords: Data hiding; Steganography; Image watermarking; deep neural network.
Research and design of in-loop virtual simulation system of tread winding control software based on MCD
by Mingxia Chen, Jijing He, Haitao Zheng, Hanyu Shi
Abstract: To improve the efficiency of industrial equipment design and debugging, virtual debugging technology is used to save the cost of industrial equipment debugging and reduce the risk of physical debugging. In this paper, an MCD-based tread winding virtual simulation system is presented, and the Software-In-Loop Virtual Debug of this system is used to study the application effect of fuzzy PID control algorithm in winding control devices. The simulation results show that compared with the traditional PID control and open-loop control, under the closed-loop control formed by the PID control algorithm with fuzzy control, the speed of the roller head is closer to the expected speed, the operation is more stable, the operation trajectory is more smooth, and a good control effect is achieved. The feasibility and effectiveness of the MCD-based tread winding virtual simulation debugging scheme are verified, and an idea is provided for the design of industrial equipment.
Keywords: in-loop virtual simulation system; MCD; tread winding control software.
Augmented Data Control in Cipher Security using Functional Procedures
by Raghvan M, Krishnmoorthy Prabu
Abstract: Data security, integrity and confidentiality are the main challenges of todays digital world Even the highly secured data can be easily broken down by a simple hacking algorithm Though, data protection raises in terms of exponential growth, on the other side there is also a tremendous growth to break down the protection The scheme for data protection attracts more researchers and still the research is going on In the first part of this paper, we present a brief study of various techniques which supports data security, and the applications corresponding to the methods of cryptography.
Keywords: Cryptographic techniques; Genetic Algorithm; Data security; Cloud database.
A New Digital Material Sharing Method Of Multimedia Network Teaching Based On 5G Communication Technology
by Degang Lai, Ke Wang
Abstract: In order to solve the problem of low efficiency of traditional network teaching material sharing, a new method of multimedia network teaching digital material sharing based on 5G communication technology is proposed. XML technology is used to design the digital material development model of multimedia network teaching. In order to realize the development of digital materials for multimedia network teaching, the material is packaged in modules. Based on 5G communication technology, a multimedia network teaching digital material sharing platform is developed to construct the Shared service protocol and realize the efficient sharing of materials. The experimental results show that the proposed method has higher file transfer rate than the traditional method, better practical performance and better application prospect.
Keywords: 5G communication technology; Multimedia network teaching; Digitalization; Material sharing; Linux system.
The Security Storage Method Of Dynamic Data In Internet Of Things Based On Blockchain
by Li Sun
Abstract: This paper proposes a method based on dynamic storage chain to overcome the problem of large amount of data in the Internet of things. In this method, ECC and D-H are used as encryption tools of the whole architecture to realize encrypted communication between lot devices. Combined with the blockchain technology, the IOT node access and the IOT dynamic data are stored to promote the dynamic data between IOT nodes to be stored in the offline storage structure, so as to achieve the purpose of secure storage of IOT dynamic data. The experimental results show that the average data storage time is 0.40s, the maximum root mean square error is 0.06, and the cost is controlled within 34500 yuan, which can effectively realize the security of IOT nodes.
Keywords: Blockchain; Internet of things; Dynamic data; Security storage.
Unsupervised Clustering Algorithm For Database Based On Density Peak Optimization
by Xiaochuan Pu, Wonchul Seo, Ning QI
Abstract: In order to improve the clustering effect of traditional unsupervised clustering algorithm for database, an unsupervised clustering algorithm based on density peak optimization is designed and proposed. K-nearest neighbor is used to set a new method to measure the sample density and sample distance, and a decision diagram of sample distance relative to sample density is drawn. The selected sample is the initial cluster center, and the number of clusters is automatically determined. In order to further improve the clustering results, the improved K-means algorithm and particle swarm optimization algorithm are introduced to optimize the convergence process of the algorithm. In order to verify the effectiveness of the proposed algorithm, a simulation experiment is designed. Experimental results show that the proposed algorithm is effective and feasible.
Keywords: Density peak optimization; Database; Unsupervised clustering; Initial cluster center; K-means algorithm; Particle swarm optimization algorithm.
Low beacon node localization algorithm in sensor networks based on multi-space key distribution algorithm
by Jianguo Huang, Sihan Fu
Abstract: : In the traditional sensor network low beacon node localization method, it is easy to be attacked by false routing, which seriously affects the localization accuracy. A location algorithm based on multi - space key distribution algorithm is proposed. By using the key information of multiple key Spaces brought by sensor nodes, the paired keys in the information are accurately calculated. Routing messages are encrypted and authenticated to effectively prevent spurious routing attacks. The hopping number information and distance information between beacon nodes are used as training data sets. The least square method is learned to realize the localization of sensor network low beacon nodes. Simulation results show that the algorithm can locate low beacon nodes quickly and accurately.
Keywords: Multi-space key distribution algorithm; Sensor network; Low beacon node location.
Research On e-Business Requirement Information Resource Extraction Method In Network Big Data
by Yawen Li
Abstract: For the challenge of the data sparsity of user-behavior in the current e-business personalized recommendation system, an information resource extraction method for e-business requirements based on similar case analysis is proposed. A recommendation model for e-commerce users requirements information resources is built, including static information, browsing behavior information, selection information of network user, and user interest. According to the built user requirement information resource recommendation model, the method based on similar case analysis is introduced into the personalized recommendation of e-business under the background of the personalized recommendation of e-business considering the potential requirement. The feature attribute similarity and comprehensive similarity of customer registration information are calculated. Experimental results show that the proposed method has good effect on product coverage, product exposure rate, and feedback rate. It can overcome the behavior sparsity of user-product, and extract the dark information in e-business requirement information resources, and overcome the long tail recommendation.
Keywords: network big data; e-business; requirement information resources; extraction method.
Exposing deepfakes in online communication:detection based on ensemble strategy
by Jie Xu, Guoqiang Wang, Tianxiong Zhou
Abstract: In recent years, deepfake techniques appeared in people's lives. As a product of deep learning, it can generate realistic face-swapping videos. Due to high fidelity, deepfake is often used to produce porn videos and guide public opinion, so as to pose a great threat to social stability. Previous studies have been able to get better detection accuracy. This paper aims to improve the detection ability of existing schemes by using the ensemble learning scheme from the perspective of model learning. Specifically, our scheme includes feature extraction, feature selection, feature classification and combination strategy. The experimental results on several datasets demonstrate that our scheme can effectively improve the detection ability of the model.
Keywords: deepfake detection; ensemble strategy; online communication; video forensics; deep learning.
Special Issue on: Special Issue Internet of Things and Blockchain Systems for Industry 4 0
Public chain based crowd sourcing for blockchain systems
by Vishnu Batla, Mukesh Soni, Mohd Naved, Samrat Ray, Renuvij
Abstract: Blockchain technology can be widely used in various services, such as online micropayments, supply chain tracking, medical record sharing, and crowdsourcing. Applying this technology to crowdsourcing systems can get a decentralised, privacy-protected, a verify research on decentralised crowdsourcing systems: decentralised crowdsourcing platforms based on smart contracts, decentralised crowdsourcing platforms based on a blockchain architecture. This paper provides a detailed overview of the leading blockchain-based decentralised rowdsourcing platforms and summarised the problems that have occurred in the existing technology, such as the security of the blockchain system, the safety of smart contracts and related issues of privacy protection, and discussed these issues in detail. Finally, I looked forward to the future of the field. The researchable problems and the amount of reference able literature is provided.
Keywords: blockchain; crowdsourcing; privacy protection; smart contract; public chain; private chain; user privacy.