Search

Anomaly detection in a dynamic data stream is a challenging task. The endless bound and high arriving rate of data prohibits anomaly detection models to store all observations in memory for processing. In addition, the dynamically moving properties of the data stream exhibit concept drift. While recent studies focus on feature extraction for anomaly detection, majority of them assume data stream are static ignoring the possibility of concept drift occurring. Anomaly detection models must operate efficiently in order to deal with high volume and velocity data, that is to have low complexity and to learn incrementally from each arriving observation. Incremental learning allows the model to adapt to concept drift. In cases where drifting rate is higher than adaptation rate, the capability to detect concept drift and retraining a new model is much preferable to minimize the performance losses. In this paper, we propose MIR_MAD, an approach based on multiple incremental robust Mahalanobis estimators that is efficient, learns incrementally and has the capability to detect concept drift. MIR_MAD is fast, can be initialized with small amount of data, and is able to estimate the drift location on the data stream. Our empirical results show that MIR_MAD achieves state-of-the-art performance and is significantly faster. We also performed a case study to show that detecting concept drift is critical to minimize the reduction in model's performance.

Secure multi-party computation (SMC) allows parties to jointly compute a function over their inputs, while keeping every input confidential. SMC has been extensively applied in tasks with privacy requirements, such as privacy-preserving data mining (PPDM), to learn task output and at the same time protect input data privacy. However, existing SMC-based solutions are ad-hoc – they are proposed for specific applications, and thus cannot be applied to other applications directly. To address this issue, we propose a privacy model DAG (Directed Acyclic Graph) that consists of a set of fundamental secure operators (e.g., +, −, ×, /, and power). Our model is general – its operators, if pipelined together, can implement various functions, even complicated ones. The experimental results also show that our DAG model can run in acceptable time.

Preterm newborns are vulnerable and easily infected due to the immature immune system. Late-onset neonatal sepsis occurring 48 hours after birth is a widespread disease among preterm newborns leading to high mortality and morbidity rates. The diagnosis is primarily based on biochemistry test, and the prescribed treatment is to use antibiotics. Risk averse clinicians, often applied overdose to reduce the mortality. A non-invasive method on monitoring vital sign signals deterioration to predict late-onset neonatal sepsis is proposed in this paper. First, we set up collectors within the local networks in Neonatal Intensive Care Units (NICUs) where bedside monitoring machines locate to capture the necessary data. Then they were transformed to images according to specific rules. Finally, a convolutional neural network was built to predict the onset of sepsis. Pilot experiments conducted on data we have collected demonstrated the feasibility of this deep learning model. This method could be incorporated into the current clinical workflow as a decision support system and provide useful information for clinicians.

Predicting student academic performance can be done by using educational data mining. Machine learning techniques play an important role for predicting academic performance from the large-scale data like the internet access log files from a university. Current data sources are mainly manual collections of data or data from a single unit of study. This study highlights the use of a new data source by transforming a university log file to predict academic performance. The log file comprises student internet access activities and browsing categories. To detect overall student academic performance, we select the best prediction accuracy by enhancing two datasets and comparing different weights in the time and frequency domains. We found that the random forest technique provides the best way in these datasets to predict students at risk-of-failure. We also found that data from internet access activities reveals a better accuracy than data from browsing categories. The combination of two datasets reveals a better picture of students' internet utilization and thus indicates how students at risk-of-failure can be detected by their internet access activities and browsing behavior.

As a prevalent disease of preterm infants, late-onset neonatal sepsis has taken up a huge proportion of morbidity and mortality of newborn babies. We have been continuously capturing vital signs of preterm infants in NICU, and proposed a non-invasive method based on machine learning techniques to predict the clinicians' treatment on them. Then we provide evaluation of predictive models and prove their feasibility. Our models could help the pediatricians make wiser clinical decision, such as more accurate treatment, avoiding the abuse of antibiotics to some extent.

Trend and outlier are frequently used to derive early warning predictive signal to decision maker in order to achieve ultimate quality decision outcome in domain specific (e.g. commercial, scientific, biomedical and engineering, just to name a few) applications. We develop a gradient-based algorithm using sample entropy gradient(SEG) for trend and outlier prediction in high frequency time series data streams. L2 similarity measure (Euclidean distance between two linearized gradient curves is then computed and used to quantify the degree of similarity and compared with a threshold L2 value to judge the extend of dissimilarity that would be classified as outlier. SEG algorithm which circumvents the need to pre-specify tolerance parameter in those cross sample entropy (CSE)-based algorithms that invariably involve real domain expert to set the tolerance threshold. We conduct real data experiments on SEG algorithm to two application areas: dynamic wind speed data stream; and financial time series data. Our experiments demonstrated that SEG algorithm can be feasibly used in online implementation to derive predictive early warning signals to domain-specific decision maker.

Bitcoin is gaining more popularity because it is designed initially to reduce the transaction cost and improve the online payment security. It is expected to replace the traditional currencies among online transactions. Building on that, we believe that the security issues during Bitcoin payment must be put in high priority. Therefore, we performed some researches on Double-spending issues. In this paper, the root causes of double spending are discussed, we focus on one possible method of double spending under the circumstance that the network connection status is not stable. We further simulate the attacking in a local area with network impairment. In terms of the reasons and specific conditions of this type of attacking, we propose a new mechanism to help these community merchants to recognize the double spending transactions and defend the related attacking. Finally, an analysis is performed among this defending strategies and it proves that the countermeasures are effective to reduce the double-spending risks.

We propose a structural health monitoring framework that is based on the hybrid intelligent systems to compute composite structure health index as structure health indicator. With our proposed health structure monitor framework, it provides a decision-level analysis for structure health monitoring. In contrast, recent studies focus on signal processing, mechanical modelling and computer based systems, which are not robust enough to provide decision-level analysis confronted with heterogeneous data sources. Therefore, data fusion technique is required to alleviate possibilities of data conflicts, lack of conciseness and incompleteness that arise from the heterogeneous data sources to compute composite structure health index, which is essential for risk analysis. Our proposed intelligent data fusion framework is based on hybrid adaptive resonant theory neural network (superior learning from dynamic data) and adaptive fuzzy inference (superior reasoning with dynamic fuzzy rules derived from time series data formulated membership functions) systems. The outcome of our proposed framework is a structure composite health index, an aggregated index via optimal weighted variables. We implement our framework using a set of GPS and Wind velocity sensors output from a real bridge to evaluate our framework. These sensors data used in our study are collected from long cable-bridge. Our case study showed that our proposed intelligent data fusing framework is capable of reporting the structural status correctly.

The potential of Gov 2.0 to facilitate greater communication, participation, and collaboration with citizens has been debated repeatedly in recent literature. However, Gov 2.0 fails to attract a satisfying level of citizen participation and has not lived up to expectations. Therefore, governments need to rethink how to engage the digital population via Gov 2.0 to achieve public value. To engage effectively, there needs to be a synergy between citizens and governments. The proposed research model in this paper aims to explain the synergy between citizens and governments that is needed to create public value via Gov 2.0. This synergistic interaction has emergent properties to create substantial public value, which is expected to result in many benefits including the increase in citizen participation via Gov 2. The theoretical foundation of the proposed research model builds on theories such as co-creation and public value. The paper concludes by suggesting further empirical research to evaluate and refine the research model.

Secure multi-party computation (SMC) allows parties to jointly compute a function over their inputs, while keeping every input confidential. It has been extensively applied in privacy-preserving computation, such as privacy-preserving data mining (PPDM), to protect data privacy. However, most SMC-based solutions are ad-hoc. They are proposed for specific applications, and thus cannot be applied to other applications directly. To address this issue, we propose a privacy model DAG (Directed A cyclic Graph) that consists of a set of secure operators (e.g., Multiplication and division). Our DAG model is general -- its operators, if pipelined together, can implement various functions. It is also extendable -- secure operators can be defined to add new features to the model. As an application study, we have applied our DAG to kernel regression. Experiments on datasets of more than 680,000 tuples show that our DAG model is effective and its running time is nearly thrice that of non-privacy setting, where parties directly disclose data.

Algorithms such as OPUS and Apriori-based Mapreduce for enhancing the efficiency of mining frequent item-set for pattern recognition application from transactional dataset have been proposed in the literature. Most of these algorithms are, however, evaluated offline on relatively small data size. When confronting with larger data size, which is inevitable for todays organisation, most if not all algorithms performed not as efficient as required to meet the real time big data driven decision making needs. We therefore attempt to solve these efficiency problems by proposing a VAMR (Vertical-Apriori Map-reduce) algorithm. VAMR is based on data attribute identifier which is exploited as capability metric for mining frequency item-set from large dataset in a single node (for example in a single site enterprise) that has no distributed and parallel computing system environment. Our evaluations using synthetic datasets and data from public repository suggest that VAMR algorithm can offer superior efficiency in mining frequent item-sets from large transaction dataset.

Some illustrative examples are provided to identify the ineffective and unrealistic characteristics of existing approaches to solving fuzzy linear programming (FLP) problems (with single or multiple objectives). We point out the error in existing methods concerning the ranking of fuzzy numbers and thence suggest an effective method to solve the FLP. Based on the consistent centroid-based ranking of fuzzy numbers, the FLP problems are transformed into non-fuzzy single (or multiple) objective linear programming. Solutions of FLP are then crisp single or multiple objective programming problems, which can respectively be obtained by conventional methods.

Four main challenges (volume, velocity, variety, veracity) have confronted computation algorithm designers in big data mining. Homomorphic cryptosystem with secured multi-party computation of matrix operations has been shown to yield high privacy preserving while data miners perform information retrieval from big data. This research concerns with the computation complexity of the big data with specific focus on computational load reduction while preserving data privacy. We propose a Teo-Han-Lee (THL) algorithm with various matrix operations to reduce the cryptographic cost significantly by cutting off at least one-third or more total computational operations. In THL, a pre-generated random key technique that we propose to apply here can decrease the computational time in which the random keys can be retrieved from memory without being generated on the fly. We further develop a collusion-resistant secure sum product protocol (CRSSPP) which is integrated in THL algorithm over arbitrary partitioned data. Experimental results demonstrated that THL-CRSSPP algorithm is more efficient than Vaidya et al SVM method [2] (state-of-the-art SVM method) and hence would be more applicable to the cloud-based big data mining. The THL-CRSSPP algorithm can also be integrated into Hadoop Mahout with a minimal effort.

The objective of this paper is to reveal the factors that determine customers' satisfaction of eMarketplaces in Saudi Arabia. Six hypotheses were postulated based on an extensive literature review in terms of security, complementary services, strong infrastructures, reliability and regulatory requirements of eMarketplaces. A survey (n=337) was carried out using convenience online tools. The sample comprised participants who have involved on eMarketplaces activities. Participants then been divided into two different groups based on their age, experience and their involvement in the technology. The first group was students (n=156) and the second group was non-students (n=181). An exploratory factor analysis followed by a confirmatory factor analysis run in SPSS which been used to test the validity of the model for each group. The survey results support all five factors from the original model with differences between the two groups regarding the strong infrastructures and regulatory requirements factors, indicating that the modified, model labeled M2 can be utilized in enhancing customer satisfaction with eMarketplaces among Saudi users.

Many studies have suggested that service system service value creation is considered crucial for electronic and online systems. However few studies have focused on the success factors that drive it. This paper proposes a research model for determining the relationship between complementary services and the service system value of eMarketplaces in Saudi Arabia (SA). It was tested by Saudi online customers (n=337) divided into two groups: student (N=156) and non-student (N=181) based on their age, experience and their involvement in the technology. The findings have proven that there is a positive correlation between complementary services and service system value creation moderated by cultural values of SA.

We review pseudo-relevance feedback as a mechanism for expanding short texts. Where short texts exhibit evolving concepts, topics and other characteristics, Web-based feedback systems were touted as the most ideal way of enriching the feature space of short texts. However, we note from a recent implementation of a Web-based pseudo-relevance feedback that it would only perform well under clinical situations. Further improvements to address fundamental noise in Web documents did not show significant improvements leading us to conclude that relevance feedback using Web documents directly are unsuitable for real-world conditions. In this paper, we present Eddi, which is a recent system that provides an exemplar of a typical pseudo-relevance feedback system. We first show the conditions in which Eddi will work and then discuss the situations where it would fail. We then present the variations to Eddi from our attempt to improve the robustness of Eddi's algorithm when dealing with complex Web documents. We then present the results from all variations to show the lack of robustness for pseudo-relevance feedback with Web documents.

A power grid has four segments: generation, transmission, distribution and demand. Until now, utilities have been focusing on streamlining their generation, transmission and distribution operations for energy efficiency. While loads have traditionally been a passive part of a grid, with rapid advances in ICT, demand-side technologies now play an increasingly important role in the energy efficiency of power grids. This paper starts by introducing the key concepts of demand-side management and demand-side load management. Classical demand-side management defines six load shape objectives, of which “peak clipping” and “load shifting” are most widely applicable and most relevant to energy efficiency. At present, the predominant demand-side management activity is demand response (DR). This paper surveys DR architectures, which are ICT architectures for enabling DR programs as well as load management. This paper also surveys load management solutions for responding to DR programs, in the form of load reduction and load shifting algorithms. A taxonomy for “group load shifting” is proposed. Research challenges and opportunities are identified and linked to ambient intelligence, wireless sensor networks, nonintrusive load monitoring, virtual power plants, etc.

This paper proposes and tests a methodology for selecting features and test cases with the goal of improving medium term bankruptcy prediction accuracy in large uncontrolled datasets of financial records. We propose a Genetic Programming and Neural Network based objective feature selection methodology to identify key inputs, and then use those inputs to combine multi-level Self-Organising Maps with Spectral Clustering to build clusters. Performing objective feature selection within each of those clusters, this research was able to increase out-of-sample classification accuracy from 71.3% and 69.8% on the Genetic Programming and Neural Network models respectively to 80.0% and 77.3%.

An analysis of the accuracy and efficiency of multiparty secured protocols is carried out so that both measures can be optimally exploited in the design of malicious party and semi-honest party. Finding efficient protocols of the Secure Multiparty Computation(SMC) is one active research area in the field of privacy preserving data mining (PPDM). The efficiency of privacy preserving data mining is crucial to many times-sensitive applications. In this paper, we study various efficient fundamental secure building blocks such as Fast Secure Matrix Multiplication(FSMP), Secure Scalar Product (SSP), and Secure Inverse of Matrix Sum (SIMS). They are supportively embedded the enhanced features into conventional data mining. We evaluate time/space efficiency on the different protocols. Experimental results are shown that there is a trade-off of accuracy and efficiency in the secured multiparty protocols targeted on semi honest party PPDM. It is therefore articulated that dimensionality reduction techniques such as Fisher Discriminant, Graph, Lapalician, and Support Vector Machine, should be used to preprocess the data. Key contributions of this paper include, besides providing some analyses of accuracy and efficiency, are commendation on further directions for computational efficiency improvement for multiparty online real data PPDM in cloud computing platforms (private and public).

This research presents a logical approach for implementing dynamic business rules based on a context-aware integrated BPM-SOA framework [18]. Specifically, a system is developed to demonstrate how Internet-based commercial lending operations with intelligent decision making can deliver faster speed, improved consistency and reliability throughout the entire lending process with scalability. With Internet-based BPM-SOA and BI engine implementation, it is possible to reassess the credit rating via key financial health indicators (i.e. liquidity, capital structure, profitability, and cash flow ratios) during the lending assessment process. If the likelihood of a credit rating change is high, the credit rating will be placed on Watching, which implies that some factors would have changed to impact on the credit rating of an applicant, hereinafter obligor. Our experimental implementation of business loan application process has demonstrated the timely and improved potential loan default possibility while revealing the scalability of our proposed context-aware integrated BPM-SOA framework.

Accurate forecasting of electricity load has been one of the most important issues in the electricity industry. Recently, along with power system privatization and deregulation, accurate forecast of electricity load has received increasing attention. In this paper, For overcoming difficulties in application of the method of BP neural network, such as it is difficult to define the network structure and the network is easy to fall into local solution. At first, By giving the undefined relation between learning ability and generalization ability of BP neural network, the hidden notes are obtained. Secondly, it poses to optimize the neural network structure and connection weights and defines the original weights and bias by means of genetic algorithm. Meanwhile, it reserves the best individual in evolution process, so that to build up a genetic algorithms Neural Networks model. This new model has high convergent speed and qualification. In order to prove the rationality of the improving GA-BP model, it analyses the network load with a area. Compare with BP neural network, it can be found that the new model has higher accuracy for power load forecasting.

This paper is to apply Automatic Speech Recognition (ASR) technique to the meal service robot in order to the user can be easier to order the meal and increase interaction between the robot and human. So far, there is not any speech recognition module for the robot in the market. In general, modify slightly by using the voice technology and modules directly to be used for the robot. Therefore, this paper is based on Mel-frequency Cepstral coefficients (MFCC) to scratch the characteristic parameters and make use of Hidden Markov Model (HMM) as the recognition technology and create speech model by spce3200 for meal service robot. The main purpose of this paper is to achieve the design of embedded HMM of speech recognition and reducing the small size, low power, short time consumption. In addition, it can fast finish training speech data and real-time recognition. This paper presents to train 25 sets of Chinese speech data for the meal service robot. The voice output is using record. We test a sentence and multi sentences under the three different environments. The highest recognition rate of a sentence has a successful rate 95% and multi-sentences can have 85% with record. In summary, the integrated test results have a relevant reliability.

The Discrete Fourier Transform (DFT) is still a widely-used tool for analyzing and measuring both stationary and transient signals in power system harmonics. However, the misapplications of DFT can lead to incorrect results caused by some problems such as aliasing effect, spectral leakage and picket-fence effect. A strategy of Recursive Group-harmonic Power Minimizing (RGPM) algorithm is developed for system-wide harmonic/inter-harmonic evaluation in power systems. The proposed algorithm can restore the dispersing spectral leakage energy caused by the DFT, and regain its harmonic/inter-harmonic magnitude and respective frequency. Every iteration loop for harmonic/interharmonic evaluation can guarantee to be convergent using the proposed Group-Harmonic Bin Power (GBP) algorithm. Consequently, not only high-precision in integer harmonic measurement can be retained, but also the inter-harmonics can be identified accurately, particularly under system frequency drift. The numerical example is presented to verify the proposed algorithm in term of robust, fast and precise performance.

This paper proposes a new approach to validating and visualising cluster structure by combining fuzzy membership functions and spectral clustering. By modifying the Visual Cluster Validity algorithm (VCV) to use an external fuzzy membership function as the distance measure and using sum of cluster membership as the sorting function, computational experiments on both the Zelnik-Manor synthetic and UCI real datasets show the proposed method, SpecVCMV, more clearly identifies the underlying cluster structure in the data.

Timely detection of hidden patterns is the key for the analysis and estimating of driving determinants for mission critical decision making. This study applies Cheong and Lee’s “context-aware” content analysis framework to extract latent properties from Twitter messages (tweets). In addition, we incorporate an unsupervised Self-organizing Feature Map (SOM) as a machine learning-based clustering tool that has not been investigated in the context of opinion mining and sentimental analysis using microblogging. Our experimental results reveal the detection of interesting patterns for topics of interest which are latent and cannot be easily detected from the observed tweets without the aid of machine learning tools.

Based on a successful implementation of a Collaborative Digital Network (CDN) the West Midlands Collaborative Commerce Marketplace (WMCCM, http://www.wmccm.com), this paper presents a breeding environment to enable stakeholders to innovate and gain competitive advantage in the global market. Stakeholders in a CDN form Communities of Practice (CoPs) and they collectively belong to an Electronic Network of Practice (E-NoP) formed via ICT. Business opportunities are gained via sharing of resources and core competences, including innovative ideas and capabilities. Three cases from the WMCCM are presented to illustrate the concept. Though aimed at rejuvenating failing SMEs in the UK, the WMCCM concept can be applied to other regions of the world. Current work involves developing theories for CDNs as well as extending the WMCCM concept to other regions of the world.

MTECS 2009 workshop Proceedings publish the state-of-the-art investigation from international researchers in pursuing and discovering new knowledge on how mobile technologies are applied in mobile enterprise computing systems. The proceedings start with a brief explanation of the motivation for the workshop and subsequently a short description of the peer-reviewed papers and a brief introduction to the workshop discussion session. We conclude with special acknowledgment to the participating authors and MTECS 2009 Program Committee members.

The paper describes a multiagent system (MAS) based architecture developed as a part of a multiagent simulation model to manage the evolution of corporate culture amongst participating members on supply chain networks. While current researches have suggested a single evolution path for corporate culture evolution, this research in contrast argued that a co-evolution path can provide more appropriate corporate culture in supporting of intelligent decision making. Thus the MAS helps all participating members in achieving the desired level of a common corporate culture understanding. Monte Carlo simulation results demonstrated the co-evolution approach has advantages in achieving corporate culture fit over single path approaches. The research is a starting point for the researchers to investigate the corporate culture roles on supply chains.

Technical analysis is aimed at devising trading rules capable of exploiting short-term fluctuations on the financial markets. The application of genetic programming (GP) as a means to automatically generate such trading rules on the stock markets has been studied. Computational results, based on historical pricing and transaction volume data, are reported for the thirty component stocks of the Dow Jones Industrial Average index. Statistical evidence shows that for the stocks that were studied, the use of GP based trading rules ensures a positive dollar return in all market scenarios. The performance of the GP based trading rules was also evaluated against the performance of the popularly used MACD technical indicator. In general, GP based trading rules offer greater returns over the simple buy and hold approach than the MACD trading signal.

Enterprise resource planning (ERP) systems are currently involved into every aspect of organization as they provide a highly integrated solution to meet the information system needs. ERP systems have attracted a large amount of researchers and practitioners attention and received a variety of investigate and study. In this paper, we have selected a certain number of papers concerning ERP systems between 1998 and 2006, and this is by no means a comprehensive review. The literature is further classified by its topic and the major outcomes and research methods of each study are addressed. Following implications for future research are provided.

This paper focuses on the application of evolutionary finance to long-term equity portfolio management. Based on Charles Darwin's theory of evolution and the associated theory of natural selection, evolutionary finance offers a promising third paradigm after traditional finance and behavioural finance. From an evolutionary finance perspective, capital markets are made up by investors who use their portfolio strategies that interact frequently with each other to compete for market capital, much alike the competition amongst biological species in ecology for finite resources. By adopting an evolutionary investment model, we provide some evidence for the evolutionary finance approach to long-term equity portfolio strategies. We simulate the competition for capitals amongst some simple rebalancing rules under repeated interactions in a market model for many periods, using historic dividend data of firms listed on the Australian Stock Exchange. Our results illustrate a long term winner of market share known as the evolutionary investment rule, as well as some market phenomena like excess volatility and an irrational rule winning market share over a rational rule.