Earticle

Sentiment analysis of reviews has been the focus of recent research, which also has been attempted in different domains such as product reviews, movie reviews, and customer feedback reviews. Most sentiment analysis of reviews focused on extracting overall evaluation for a single product which makes difficult for a customer to know all the features of product and make a decision. Thus, mining this data, identifying the user opinions about different features and classify them is an important task. This paper is devoted to identify opinion object from short comments, and analyze sentiment of product based on features-level. CRFs model based on word embedding feature is adopted by identifying opinion object, which obtains a satisfied results. In addition, calculate rules based on syntax parsing are proposed to accomplish features-level sentiment analysis which extracts user’s opinion on many aspects. Experimental results using short comments of movies show the effectiveness of our approach.

Security is playing a very important role in the field of authentication system,network system and Internet. The main goal of cryptography is to secure the data so that it cannot be accessed by any unauthorized user. Cryptography is an emerging tool, which is important for authentication. The AES is a 128 bit Symmetric block Cipher. This paper include enhancing reliability and security, use modified AES (Advanced Encryption Technique) which will be implemented step by step. For the process of encryption various classical techniques are used. These are substitution technique, rearrangement and transformation technique. Key expansion module is introduced in the encryption and decryption modules, which generates key for all iterations. In each iterative rounds addition of an arithmetic operator and a route transposition cipher is introduced in this modification. To increase the immunity against unauthorized attacks, the Key extended module doubles the number of iterative processing rounds.

As one of the most popular data mining techniques, clustering is an important way of exploratory data analysis and pattern discovery given the explosive increasing amount of dataset. Applying clustering in customer segmentation is a common method to discover high-value customers. However, traditional clustering methods such as k-means are performing single direction (either row or column) on the data matrix, and thus the clustering results might involve cases that are irrelevant to specific dimensions. Besides, traditional clustering is achieved upon the whole set of attributes or variables, and therefore only capable of discovering global information. Along this line, in order to reduce the dimensions and find out potential local patterns in the data matrix, we proposed a bi-clustering algorithm for customer segmentation. Our experiments using supermarket customer dataset improve the effectiveness and efficiency of proposed bi-clustering algorithm.

Information retrieval is fast becoming the prevailing form of information access, surpassing traditional database style searching. Ontologies have become the tool of choice employed in many information retrieval systems and more prominently in semantic information retrieval. In order to overcome the disadvantages in key word based information retrieval systems, which transfer irrelevant information, ontology has been designed. A system with ontology mimics the real world, where every task is laced with certain meaning as this is basic idea behind knowledge processing. Hadoop, which is an open source frame work for storing and processing large datasets, is used for pre-processing the text documents. First, a set of text documents are considered. Pre-processing is performed on a large domain of data using Hadoop MapReduce. This includes the removal of the stop words along with stemming and excluding less frequency words. Despite this pre-processing, owing to the colossal number of index terms still floating in the considered domain data, the problem of high dimensionality is encountered. Therefore the dimensionality of such a group of terms is reduced by identifying it as a concept and those concepts can be viewed as a single dimension in a ontology based information retrieval system. Now ontology is constructed by assigning synonym set to each concept in this structure using tools like word net. Thus constructed ontology can be mapped on to the processed query which gives us the relevant information from the data pool considered.

Principal Component Analysis (PCA) is a classical method for dimensionality reduction, data pre-processing, compression and visualization of multivariate data for different applications in biology, social science and engineering. The limitation of PCA is lacking of interpretation due to the non-zero loadings and the inconsistence for highdimensional data. Sparse principal component analysis (sparse PCA) is proposed mainly for the challenges of PCA above. For the past decades, many works of the development methods and theoretical analysis for sparse PCA have been presented. The goal of this paper is to give a comprehensive literatures review to recent progress in highdimensional sparse PCA from algorithm and statistical theory. Firstly we give the overview for PCA and sparse PCA. Secondly the algorithms of sparse PCA are categorized into different classes and provide detailed descriptions for typical formulations and methods in each category, and the typical packages of sparse PCA are also given. Considering that statistical analysis in high dimension becomes more involved in sparse PCA, and then the survey of theoretical analysis of sparse PCA is also presented. Finally the future trends as well as challenges are given.

Traditional Association Rule Mining has been extensively used to discover interesting rules or relationships between items in large databases but it has limitations that it solely deals with the items or products that are sold but avoids the items that are nearly sold. These nearly sold things carry hesitation data since customers are indecisive to shop for them. In this paper, with the help of vague set theory, we describe that item’s hesitation information is precious knowledge for the design of profitable selling strategies. This work proposed Genetic Algorithm based on evolution principles that has found its strong base in mining or maximize the rules for the items that customers mostly hesitate to purchase or has a high percentage of hesitation because of some reasons like price of an item, quality of an item, etc. Fitness function, crossover, and mutation are the main parameters involved in Genetic Algorithm which we used in our work. This work describes that if the reason of giving up the items is identified and resolved, we can easily remove this hesitation status of a customer and considering newly evolved rules as the interesting ones for boosting the sales of the item.

Microblog retrieval has received much attention in recent years. In microblog retrieval, the content linked by URLs is one of the most important information of a microblog. We present a Hyperlink-extended model for microblog retrieval that combines content of microblogs and the content of embedded hyperlinks webpages using a probabilistic ranking function based on language model. Hyperlink-extended language model incorporates the users' information retrieval requirements and the microblog author’s expression needs. Using standard TREC 2011 and TREC 2012 microblog retrieval collection, various aspects of our microblog retrieval model are evaluated. Results show our model significantly outperform the art-of-the-state URL-based approaches and the best performance of TREC 2012 microblog retrieval.

Focused on the problem of imperfect information in the process of reconstruction from single time series, a new technology for phase space reconstruction from multi-time series based on the data fusion is proposed. Firstly, the methods Cao and mutual information are used to select the reconstruction parameters, time delay and embedded dimension; secondly, the social cognitive optimization algorithm is brought to calculate the weights for each variable; thirdly, an adaptive weighted fusion estimating method is applied for data fusion; lastly, the effectiveness of the methods mentioned in this paper is demonstrated by the analysis results of one case study of real chemical plant data sets, and the proposed methods in this paper can improve the completeness of the information of the reconstructed phase space, which is also a good foundation for further analysis of complex system.

In this paper various methods of sequence analysis which include the alignment based and alignment free methods of tree generation are reviewed and these find distance/similarity among the sequences of different species. Alignment free method based on tuple count and set theory is proposed and the results are compared with the guide tree obtained using alignment based method. The proposed method is tested on DNA sequence of length below 1000bp (dataset1) and Sequence of length above 16000bp (dataset2). It achieves the similar performance as that of the alignment based method but without the alignment phase.

Pedestrian detection is one of the hot research problems in computer vision field. The Cascade AdaBoost System is a commonly used algorithm in this region. However, when the training datasets become larger, it is still a time consuming process to build one Adaboost classifier. In this paper we detail an implementation of the AdaBoost algorithm using the NVIDIA CUDA framework based on the haar features as feature vectors, and downscaling with integral image. The result shows that we can get nearly 6x from the standard code to with our CPU implementation to achieve a near real-time performance and ensure better classification results in misjudgment.

Multi Query Optimization is an important process in database and it becomes the commonplace due to the frequent usage of decision support systems in almost all the multinational enterprises. The multiple queries from different users that have been addressed to one schema often have a lot of common sub-expressions and it is the function of the multi query optimization algorithms such as Basic Volcano, Volcano RU and Volcano SH algorithms to optimize such multiple queries together and executes the common operation once and share the output among the queries. In this work, a multi query shareability algorithm which can efficiently detect the common sub- expressions among the multiple queries and share the output among those queries was proposed and algorithm for optimal order of those queries was also proposed. The Algorithm has a time complexity of O(n2 + 9n +6) while the most recent basic algorithm thus Volcano RU Algorithm has O(2n2 +20n +12), both the algorithms have O(n2) time complexity which is quadratic in nature. However, the Proposed Algorithm is more efficient and better than Volcano RU algorithm even if n approach to infinity.

Due to various reasons, there are generally missing data in datasets. Usually the missing data in these incomplete datasets need to be filled. In this paper, the drawbacks of some existing data filling approaches for incomplete information systems are analyzed based on Rough Set theory. Several similarity relation models are discussed and the Valued Limited Tolerance Relation model is proposed. A data filling algorithm based on the Valued Limited Tolerance Relation model is put forward. This approach makes full use of the similarity of objects and selects the object which is the most similar to the incomplete object. More missing data can be filled scientifically. The experimental results show that this approach is effective.

The potential of cloud computing for overriding the needs for deploying various infrastructures for running a server based services brought up a revolutionary change in the way the traditional demands of the people use to be handled. Cloud computing provides the rental service for the user in which a user can use the particular software by paying for that on the cloud server. Since the whole scenario is beneficial to big industries like Facebook, Google, Orkut etc., various other fields are also getting dependent on cloud computing. Since tons of data is uploading every second to the cloud server does need to be mined properly for efficient data storage. In this paper we try to integrate the data preprocessing technique with data classification technique to mine big data’s of asthma based patients. We have used simulation tool called eclipse to run the API’s of weka and cloudsim for setting up the experimental environment.

Sentiment analysis based on the aspects of products or services is designed to explore subjective information such as attitudes and opinions in user-generated reviews. Although a great many of approaches have been proposed in detecting aspects and the relevant aspect-specific sentiments, most of them detect the latent aspects with no proper classifying them or classify them employing unsupervised topic modeling without predicting the sentiment towards these aspects. This paper proposes a novel sentiment-aspect analysis probabilistic modeling framework consisting of Seeding words extraction and semi-supervised topic (SST) model based on Sentence-LDA. More specifically, the proposed methodology starts by capturing seeding words from the websites inherent semi-structured information about products or services description. Then, it employs the captured seeding words to instruct the discovery of aspects and relevant sentiment of products or services simultaneously. Experimental results show that significant improvements have been achieved by the proposed method with respect to other state-of-the-art methods.

Due to the low computational precision, local optimal solution and slow convergence speed of particle swarm optimization (PSO) algorithm, an improved PSO (SAQPSO) algorithm based on simulated annealing (SA) and quantum theory is proposed in this paper. The first, quantum theory is used to change the updating mode of the particles in order to improve the search speed and the convergence precision, and guarantee the simplification and effectiveness. Then the SA with probability and local search ability is introduced into quantum PSO (QPSO) in order to keep the diversity of the population, avoid falling into local optimum and enhance the global search ability. The SAQPSO algorithm keeps the characteristics of the simple and easy implementation, improves the global optimization ability and the convergence speed and the accuracy. Finally, some benchmark functions are used to prove the validity of the proposed SAQPSO algorithm. The computational results show that the proposed SAQPSO algorithm takes on the fast convergence speed, the better robustness and global search ability.

Big data processing is currently becoming increasingly important research field in computer technology professionals due to the continuous growth of the amount of data generated by various fields. However, the processing of large-scale research data requires cluster technology infrastructure which causes huge investments for educational institutions. Hadoop is an open-source framework that allows for distributed storage and processing of very large data sets on computer clusters built from commodity hardware. This technology is being widely used for the analysis of large datasets. This paper focuses on proposes a low cost scalable hadoop virtual cluster platform and the performance of hadoop irtual cluster. We first describe the design and implementation of a virtual datacenter using hadoop framework. Then we perform a set of experiments to investigate the performance of virtual datacenter with standard datasets. For experiment and evaluation, the Cloudera's distribution of Apache Hadoop (CDH) is installed and configured on the DIU Cloud and Big Data Lab as a prototype implementation of a virtual data center and tested with 40 workstations. In this paper, we use TPC BENCHMARK ™ DS” by-Transaction Processing Performance Council (TPC) benchmarks for evaluation virtual data center performance. The contributions of this paper, is to design model and implement a cost effective elastic virtual data center with hadoop framework and resource utilizations for educational institutions to provide high performance for distributed and parallel processing; as well as, Identify the bottlenecks of this systems e.g. bandwidth of network connectivity with all nodes in the lab.

Aimed at improving the efficiency of the development of complex simulation system and the problem of lower development costs, a validation development framework is proposed based on combination model. The system development process can be broken in the discovery of decomposition for the model, combination and combination of validation phase. This thesis gives the validation method based on model of tag transfer system behavior, judging by strong simulation and semantic similarity relations after the combination model of dynamic behavior and request model conforms to the degree. Simulation model combined the experimental results show that the method can automatically finish the combination of the simulation model validation, the validity of the model validation can meet the requirements of practical application.

In order to overcome the blindness of chaotic search, improve the convergence speed and global solving ability of the basic ant colony optimization(ACO) algorithm, an improved ACO algorithm based on combining multi-population strategy, adaptive adjustment pheromone strategy, chaotic search method and min-max ant strategy (MPCSMACO)is proposed in this paper. In the proposed MPCSMACO algorithm, the multi-population strategy is introduced to realize the information exchange and cooperation among the various types of ant colony. The chaotic search method with the ergodicity, randomness and regularity by using the logistic mapping is used to overcome too long search time, avoid falling into the local extremum in the initial stage and improve the search accuracy in the late search. The min-max ant strategy is used to avoid the local optimization solution and the stagnation. And the ants with different probability search different area according to the concentration of pheromone, so as to reduce the search number of the blindness of chaotic search method. Several Benchmark functions are selected to testify the performance of the MPCSMACO algorithm. The experiment results show that the MPCSMACO algorithm takes on the better global search ability and convergence performance.

In order to overcome the drawbacks of falling into local extremum and lower optimization precision of standard particle swarm optimization (PSO) algorithm, multi-population strategy, adaptive dynamic adjustment strategy and co-evolution mode are introduced into the standard PSO algorithm in order to propose an improved co-evolutionary PSO(MPACEPSO) algorithm based on multi-strategy evolution mode and multi-population co-evolutionary mechanism. In the evolutionary process of MPACEPSO algorithm, the multi-population strategy is used to divide the population into several sub-populations, which use different co-evolutionary model to evolve. These sub-populations are influenced and promoted each other in order to realize the exchange of information and co-evolution among the sub-populations, improve the convergence speed and search precision of MPACEPSO algorithm, and effectively suppress the appearance of the local optimum. The adaptive dynamic adjustment strategy of inertia weight is used to keep the diversity of population, reduce the probability of falling into the local extremum. Finally, the ZDT functions are selected to test the optimization performance of proposed MPACEPSO algorithm. The experimental results show that the proposed MPACEPSO algorithm has faster convergence speed, stronger global search ability, higher solving precision and better dynamic optimization performance. The experimental result analysis shows that the proposed MPACEPSO algorithm is insensitive to parameters and easy to be used in solving the complex optimization problems.

Data privacy preservation is one of the most disturbed issues on the current industry. Data privacy issues need to be addressed urgently before data sets are shared on cloud. Data anonymization refers to as hiding complex data for owners of data records. In this paper investigate the problem of big data anonymization for privacy preservation from the perspectives of scalability and time factor etc. At present, the scale of data in many cloud applications increases tremendously in accordance with the big data trend. Here propose a scalable Two Phase Top-Down Specialization (TPTDS) approach to anonymize large-scale data sets using the MapReduce framework on cloud. For the data anonymization-45,222 records of adults information with 15 attribute values was taken as the input big data. With the help of multidimensional anonymization on map reducing framework, here implemented the proposed Two-Phase Top-Down Specialization anonymization algorithm on hadoop will increases the efficiency of the big data processing system. In both phases of the approach, deliberately design multidientional MapReduce jobs to concretely accomplish the specialization computation in a highly scalable way. Data sets are generalized in a top-down manner and the better result was shown in multidmientional MapReduce framework by compairing the onedimentional MapReduce framework anonymization job. The anonymization was performed with specialization operation on the taxonomy tree. The experiment demonstrates that the solutions can significantly improve the scalability and efficiency of big data privacy preservation compared to existing approaches. This work has great applications to both public and private sectors that share information to the society.

Word sense disambiguation (WSD) is important for natural language processing. It plays important roles in information retrieval, machine translation, text categorization and topic tracking. In this paper, the transition among senses of words is considered. For an ambiguous word, its semantic codes and its left word’s semantic codes are taken as disambiguation features. At the same time, a new method based on hidden Markov model (HMM) is proposed for Chinese word sense disambiguation. Chinese Tongyici Cilin is used to determine semantic codes of words. HMM is optimized in training corpus. The WSD classifiers based on HMM is tested. Experimental results show that the accuracy of word sense disambiguation is improved.

Faultage is a specialized term in geology, but it can be used to describe some characteristics vividly in data resource. In this paper, we set up the preliminary theoretical system of data faultage to lay the foundation of later research and make contribution to the structure standardization of data resource. More concretely, data faultage in six areas has been enumerated firstly. Then, the conception of data faultage is presented on the theory of geological faultage, and the details of data faultage are discussed on the microscopic view. Finally, we make a verification case based on data faultage, some information from Shanghai Media Group are used to analyze the distribution of its listeners, and the theoretical system of data faultage are verified.

This paper proposes a Multiclass Least Squares Twin Support Vector Machine (MLSTSVM) classifier for multi-class classification problems. The formulation of MLSTSVM is obtained by extending the formulation of recently proposed binary Least Squares Twin Support Vector Machine (LSTSVM) classifier. For M-class classification problem, the proposed classifier seeks M-non parallel hyper-planes, one for each class, by solving M-linear equations. A regularization term is also added to improve the generalization ability. MLSTSVM works well for both linear and non-linear type of datasets. It is relatively simple and fast algorithm as compared to the other existing approaches. The performance of proposed approach has been evaluated on twelve benchmark datasets. The experimental result demonstrates the validity of proposed MLSTSVM classifier as compared to the typical multi-classifiers based on ‘Support Vector Machine’ and ‘Twin Support Vector Machine’. Statistical analysis of the proposed classifier with existing classifiers is also performed by using Friedman’s Test statistic and Nemenyi post hoc techniques.

It is an important to quickly and accurately forecasting road network traffic flow in intelligent transportation systems, Aiming at the forecasting problem of short-term traffic flow, this paper proposed a traffic flow prediction algorithm, which based on traffic flow sequence partition and neural network model. Firstly, the algorithm divided the traffic flow into different patterns and time sequence by clustering, secondly, described and predicted traffic flow model according to BP neural network. Finally, the experiment shows that based on combined model is much accurate.

In order to effectively integrate multimodal information and multilayer constraints, we present a unified probabilistic framework for sports video analysis. Based the framework, three instances of statistical models are constructed and compared. Experimental results indicate our method with multimodal fusion processes semantic events in sports video more effectively.

Six Degrees of Separation is a theory that has recently been popularized due to the emergence of various social networking platforms. This theory states that any two random people in the world can be associated with each other with no more than six intermediate links. Though this theory has not been verified yet it has a great potential for practical implementation. In this paper we use this theory to create an application which is going to help the user in finding other individuals who share some common links and can be of benefit to the user. This is achieved through maintaining a central database which contains the details of all the users using our application. A user can search for other people on our database based on certain parameters which would then display the resultant links between the user and the selected target. This establishing of links is achieved using two dimensional bi-direction search algorithm.