Artificial Intelligence Applications and Innovations

In this paper we propose a new service delivery platform (SDP), named Future SDP that incorporates principles of cloud computing and service oriented architecture (SOA). Future SDP allows resources, services and middleware infrastructure deployed in diverse clouds to be delivered to users through a common cloud Broker. This cloud Broker is enhanced with policy, management, security and mediation functionalities that ensures proper use of development tools, resources and services only to certified participants of the federated clouds. This work focuses on solving interoperability problems both into infrastructure and middleware layer of the conventional SDPs. Finally, the operational attributes of Future SDP are depicted through a delivery scenario of healthcare telemonitoring services.

Geriatric depression is a highly frequent medical condition that influences independent living and social life of senior citizens. It also affects their medical condition due to reduced commitment to the appropriate treatment. Coexistence of depressive symptoms in Mild Cognitive Impairment (MCI) and lack of objective tools towards their reliable distinction from neurodegeneration, motivated this study to propose a computerized approach of depression recognition. Resting state electroencephalographic data of both rhythmic activity and synchronization features were extracted and the Mahalanobis Distance (MD) classifier was adopted in order to differentiate 33 depressive patients from an equal number of age-matched controls. Both groups demonstrated cognitive decline within the context of MCI. The promising results (89.39% overall classification accuracy, 93.94% sensitivity and 84.85% specificity) imply that combination of neurophysiological (EEG) and neuropsychological tools with pattern recognition techniques may provide an integrative diagnosis of geriatric depression with high accuracy.

The diagnosis of brain neoplasms has been facilitated by the emerging of high-quality imaging techniques, such as Magnetic Resonance Imaging (MRI), while the combination of several sequences from conventional and advanced protocols has increased the diagnostic information. Treatment planning and therapy follow-up require the detection of neoplastic and edematous tissue boundaries, a very time consuming task when manually performed by medical experts based on the 3D MRI data. Automating the detection process is challenging, due to the high diversity in appearance of neoplastic tissue among different patients and, in many cases, similarity between neoplastic and normal tissue. In this paper, we propose an automatic brain tumor segmentation method based on a multilabel multiparametric random walks approach initialized by an outlier detection scheme. Segmentation assessment is performed by measuring spatial overlap between automatic segmentation and manual segmentation performed by medical experts. Good agreement is observed in most of the 26 cases for both neoplastic and edematous tissue. The highest achieved overlapping values were 0.74 and 0.79 for neoplastic and edematous tissue, respectively.

In this work, we study the behavior of a feature selection algorithm (backwards selection) using random forests, by fusing multi-modal data from different subjects. Two separate datasets related to cutaneous melanoma, obtained from image (dermoscopy) and non-image (microarray) sources are used. Imputations are applied in order to acquire a unified dataset, prior the effect of machine learning algorithms. The results suggest that application of the normal random imputation method acts as an additional variation factor, helping towards stability of potential recommended biomarkers. In addition, microarray-derived features were favorably selected as best predictors compared to image-derived features.

A selection of artificial neural network models were built and implemented for systematically study the contribution and the sensitivity of the main influencing parameters as important contributing factors for the non-invasive prediction of chromosomal abnormalities. The parameters that had been investigated are: the previous medical history of the pregnant mother, the nasal bone, the tricuspid flow, the ductus venosus flow, the PAPP-A value, the b-hCG value, the crown rump length (CRL), the changes in nuchal translucency (deltaNT) and the mother’s age. The main conclusions drawn are: 1) The deltaNT is the most significant factor for the overall prediction, while the CRL the least significant. 2) The previous medical history of the pregnant mother is not a significant factor for the prediction of the abnormal cases. 3) The nasal bone, the tricuspid flow and the ductus venosus flow contribute significantly in the prediction of trisomy 21 but not in the prediction of the “normal” cases. 4) The PAPP-A, the b-hCG and the mother’s age are of intermediate importance. Also, a sensitivity analysis of the attributes PAPP-A, b-hCG, CRL, deltaNT and of the mother’s age was done. This analysis showed that the CRL and deltaNT are more sensitive when their values are decreased, the PAPP-A is more sensitive when its values are increased and the b-hCG is insensitive to variations in its values.

The complete hospital information system supports the electronic patient record, which, with the history registration, its laboratory and depiction examinations through the use of expert systems, leads to the immediate and effective diagnosis and treatment of the illness. In this present study with the use of a data flow diagram, which consists of a small indication of the way an expert system based on Artificial Intelligence can be made applicable, the steps which lead to the diagnosis of thyroid cancer will be mentioned, when the patient is admitted to the outpatient Endocrinology department. With the data flow diagram, the users can visualize the way in which the system will operate and what it can achieve. Also it will present the course which the medical professional follows in order to reach the diagnosis of thyroid cancer with the slightest error percentage, using the medical information in the extensive hospital information system.

A large scale analysis of gene expression data, performed by Segal and colleagues, identified sets of genes named Cancer Modules (CMs), involved in the onset and progression of cancer. By using functional interaction network data derived from different sources of biomolecular information, we show that random walks and label propagation algorithms are able to correctly rank genes with respect to CMs. In particular, the random walk with restart algorithm (RWR), by exploiting both the global topology of the functional interaction network, and local functional connections between genes relatively close to CM genes, achieves significantly better results than the other compared methods, suggesting that RWR could be applied to discover novel genes involved in the biological processes underlying tumoral diseases.

Quantitative analysis of brain lesions and ischemic infarcts is becoming very important due to their association with cardiovascular disease and normal aging. In this paper, we present a semi-supervised segmentation methodology that detects and classifies cerebrovascular disease in multi-channel magnetic resonance (MR) images. The method combines intensity based fuzzy c-means (FCM) segmentation with spatial probability maps calculated from a normative set of images from healthy individuals. Unlike common FCM-based methods which segment only healthy tissue, we have extended the fuzzy segmentation to include patient-specific spatial priors for both pathological conditions (lesions and infarcts). These priors are calculated by estimating the statistical voxel-wise variation of the healthy anatomy, and identifying abnormalities as deviations from normality. False detection is reduced by knowledge-based rules. Assessment on a population of 47 patients from different imaging sites illustrates the potential of the proposed method in segmenting both hyperintense lesions and necrotic infarcts.

This paper deals with the deployment of personalized healthcare (pHealth) services upon new networks supporting pervasive connectivity. A profiling scheme of two levels is introduced, where health status related attributes and preferences are included in the upper-level, named Service User Profile and communication attributes compose the lower-level, named Transport User Profiles. We manage pervasive connective by means of Always Best Connected paradigm for enhancing personalization within Next Generation Network architecture.

One of the most important tasks in the process of designing educational material for distance learning is the representation and modeling of the cognitive domain to which the material refers. However, this representation should be formal, complete and reusable in order to be used by intelligent tutoring system applications, other knowledge domains or tutors. In the context of this work, we propose a methodology that relies on the notion of ontology so as to represent the knowledge domain. Moreover, this methodology has been applied to the educational material of the Hellenic Open University.

Many researchers have focused on the mining of educational data stored in databases of educational software and Learning Management Systems. The goal is the knowledge discovery that can help educators to support their students by managing effectively educational units, redesigning student’s activities and finally improving the learning outcome. A basic data mining technique concerns the discovery of hidden associations that exist in data stored in educational software Databases. In this paper, we present the KDD process which includes the application of the Apriori algorithm for the association rules mining from the educational data of ESOG Web-based application.

The importance of e-learning and e-commerce applications has significantly increased in the past few years. Seeking better design and implementation principles is a research goal with, potentially, a significant impact. One of the commonalities of both applications is user-centricity. Understanding user behavior is critical especially in user-centered applications such as e-commerce and e-learning. In this work we discuss some of the fundamental similarities and differences in e-commerce and formal e-learning adaptation and discuss lessons that could be learned. We argue that current user pattern mining techniques should take into account behavioral and educational theories for distance learning in order to be efficient.

In this work, we propose and evaluate a Higher Order Singular Value Decomposition (HOSVD) of a tensor as a means to classify player behavior and adjust game difficulty dynamically. Applying this method to player data collected during a plethora of game sessions resulted in a reduction of the dimensionality of the classification problem and a robust classification of player behavior. Simultaneously HOSVD was able to perform significant data compression without significant reduction as regards to the accuracy of the classification outcome and furthermore, was able to alleviate the data sparseness problem common within data collected from game sessions.

The increased demand of Web services and diverse characteristics of users have resulted in a plethora of applications that aim to provide personalized services based on the heterogeneous needs and preferences of users. With the aim to enhance and support the personalization process of Web applications, an innovative adaptation mechanism is proposed. The mechanism is based on a series of psychometric measures which capture the cognitive style of users and a Computational Intelligence technique embracing Artificial Neural Networks and Fuzzy Logic. The proposed mechanism decides on the adaptation effects of Web applications and provides a personalized user experience. The proposed method has been evaluated with a user study and provides interesting insights with respect to the effect of adaptation in terms of task accuracy, performance and satisfaction of users while interacting with an adapted and a non-adapted version of the same Web environment.

Lives are saved by utilization and application of the latest technologies in hospitals around the world to improve patient treatments and well being. Secure, accurate, near real time data acquisition and analysis of patient data and the ability to update such data will reduce cost and improve the quality of patient’s care. This paper considers a wireless framework based on radio frequency identification (RFID) that uses wireless networks for fast data acquisition and transmission. This paper discusses the development of an intelligent multi-agent system in a framework in which RFID can be used for patient data collection. An approach to make the data communications more secure in a hospital environment is proposed. A new method for data classification and access authorization is also developed which will assist in preserving privacy and security of data.

Development and management of disaster recovery plan for IT systems are complex, demanding, and yet crucial to an organization success and its competitive position in the marketplace. Due to rapid changes in emerging technologies there is a need for constant improvement and adjustment to disaster recovery plans for IT systems. There are a large number of processes involved in disaster recovery planning for IT system. The interdependencies of these processes make it very difficult for Chief Information Officers (CIOs) to comprehend and be aware of effect of inefficiencies that may exist in development of these processes in the disaster recovery plan of their organization. This paper considers the implementation of a Fuzzy Cognitive Maps (FCM) to provide facilities to capture and represent complex relationships in implementing a disaster recovery plan for IT systems and their related processes to improve the understanding of CIOs about the systems and its associated risks.

QSAR modeling is a method for predicting properties, e.g. the solubility or toxicity, of chemical compounds using statistical learning techniques. QSAR is in widespread use within the pharmaceutical industry to prioritize compounds for experimental testing or to alert for potential toxicity. However, predictions from a QSAR model are difficult to assess if their prediction intervals are unknown. In this paper we introduce conformal prediction into the QSAR field to address this issue. We apply support vector machine regression in combination with two nonconformity measures to five datasets of different sizes to demonstrate the usefulness of conformal prediction in QSAR modeling. One of the nonconformity measures provides prediction intervals with almost the same width as the size of the QSAR models’ prediction errors, showing that the prediction intervals obtained by conformal prediction are efficient and useful.

Given an example-feature set, representing the information context present in a dataset, is it possible to reconstruct the information context in the form of clusters to a certain degree of compromise, if the examples are processed randomly without repetition in a sequential online manner? A general transductive inductive learning strategy which uses constraint based multivariate Chebyshev inequality is proposed. Theoretical convergence in the reconstruction error to a finite value with increasing number of (a) processed examples and (b) generated clusters, respectively, is shown. Upper bounds for these error rates are also proved. Nonparametric estimates of these error from a sample of random sequences of example set, empirically point to a stable number of clusters.

Venn Predictors (VPs) are machine learning algorithms that can provide well calibrated multiprobability outputs for their predictions. The only drawback of Venn Predictors is their computational inefficiency, especially in the case of large datasets. In this work, we propose an Inductive Venn Predictor (IVP) which overcomes the computational inefficiency problem of the original Venn Prediction framework. Each VP is defined by a taxonomy which separates the data into categories. We develop an IVP with a taxonomy derived from a multiclass Support Vector Machine (SVM), and we compare our method with other probabilistic methods for SVMs, namely Platt’s method, SVM Binning, and SVM with Isotonic Regression. We show that these methods do not always provide well calibrated outputs, while our IVP will always guarantee this property under the i.i.d. assumption.

Automated detection of anomalous trajectories is an important problem in the surveillance domain. Various algorithms based on learning of normal trajectory patterns have been proposed for this problem. Yet, these algorithms suffer from one or more of the following limitations: First, they are essentially designed for offline anomaly detection in databases. Second, they are insensitive to local sub-trajectory anomalies. Third, they involve tuning of many parameters and may suffer from high false alarm rates. The main contribution of this paper is the proposal and discussion of the Sliding Window Local Outlier Conformal Anomaly Detector (SWLO-CAD), which is an algorithm for online detection of local sub-trajectory anomalies. It is an instance of the previously proposed Conformal anomaly detector and, hence, operates online with well-calibrated false alarm rate. Moreover, SWLO-CAD is based on Local outlier factor, which is a previously proposed outlier measure that is sensitive to local anomalies. Thus, SWLO-CAD has a unique set of properties that address the issues above.

In this paper, an introduction to the main steps required to develop conformal predictors based on fuzzy logic classifiers is provided. The more delicate aspect is the definition of an appropriate nonconformity score, which has to be based on the membership function to preserve the specificities of Fuzzy Logic. Various examples are introduced, to describe the main properties of fuzzy logic based conformal predictors and to compare their performance with alternative approaches. The obtained results are quite promising, since conformal predictors based on fuzzy classifiers show the potential to outperform solutions based on the nearest neighbour in terms of ambiguity, robustness and interpretability

Indoor localisation is the state-of-the-art to identify and observe a moving human or object inside a building. Location Fingerprinting is a cost-effective software-based solution utilising the built-in wireless signal of the building to estimate the most probable position of a real-time signal data. In this paper, we apply the Conformal Prediction (CP) algorithm to further enhance the Fingerprinting method. We design a new nonconformity measure with the Weighted K-nearest neighbours (W-KNN) as the underlying algorithm. Empirical results show good performance of the CP algorithm.

This paper describes the methodology of providing multiprobability predictions for proteomic mass spectrometry data. The methodology is based on a newly developed machine learning framework called Venn machines. They allow us to output a valid probability interval. We apply this methodology to mass spectrometry data sets in order to predict the diagnosis of heart disease and early diagnoses of ovarian cancer. The experiments show that probability intervals are valid and narrow. In addition, probability intervals were compared with the output of a corresponding probability predictor.

Current classification algorithms focus on vectorial data, given in euclidean or kernel spaces. Many real world data, like biological sequences are not vectorial and often non-euclidean, given by (dis-)similarities only, requesting for efficient and interpretable models. Current classifiers for such data require complex transformations and provide only crisp classification without any measure of confidence, which is a standard requirement in the life sciences. In this paper we propose a prototype-based conformal classifier for dissimilarity data. It effectively deals with dissimilarity data. The model complexity is automatically adjusted and confidence measures are provided. In experiments on dissimilarity data we investigate the effectiveness with respect to accuracy and model complexity in comparison to different state of the art classifiers.

Pattern recognition is becoming an increasingly important tool for making inferences from the massive amounts of data produced in magnetic confinement fusion experiments. However, the measurements obtained from the various plasma diagnostics are typically affected by a considerable statistical uncertainty. In this work, we consider the inherent stochastic nature of the data by modeling the measurements by probability distributions in a metric space. Information geometry permits the calculation of the geodesic distances on such manifolds, which we apply to the important problem of the classification of plasma confinement regimes. We use a distance-based conformal predictor, which we first apply to a synthetic data set. Next, the method yields an excellent classification performance with measurements from an international database. The conformal predictor also returns confidence and credibility measures, which are particularly important for interpretation of pattern recognition results in stochastic fusion data.

In order to improve the computational efficiency of conformal predictor, distance metric learning methods were used in the algorithm. The process of learning was divided into two stages: offline learning and online learning. Firstly, part of the training data was used in distance metric learning to get a space transformation matrix in the offline learning stage; Secondly, standard CP-KNN was conducted on the remaining training data with a nonconformity measure function defined by K nearest neighbors classifier in the transformed space. Experimental results on three UCI datasets demonstrate the efficiency of the new algorithm.

In typical video-to-video transmissions, security and confidentiality is becoming an issue of greater importance, but these features come at a cost. In mobile environments, where CPU time is a valuable resource, such features should be thoroughly thought over as they usually require heavy computational resources. In this paper a short analysis on existing streaming solutions, standardised and otherwise, is performed while taking into consideration the scope of the LiveCity project of developing applications destined to the end-user. An analysis of different transmission protocols and their specifications, as well as encryption protocols designed to work on top of streamed data, is performed as a means to access which specifications better fit LiveCity requirements.

The LiveCity Project addresses a number of communities where the citizens of a city can have specific challenges and derive immediate benefits or advantages from the proper and effective use of live interactive high definition video-to-video (v2v) communications over the Internet, for various sectors of applications and/or related services. One of these communities is the educational community, where major benefits and impacts may raise, as a result of the innovative v2v implementation and use, to support teaching and learning activities. In this paper, we propose a specific context to satisfy this challenge, based on the formation of a “network” of involved schools, where we also identify essential users’ needs and propose distinct educational use cases (i.e.: v2v for school-to-school communication and v2v for an interactive homework channel). In addition, we discuss potential benefits from this action and we identify legal and ethical issues that may affect further progress.

The LiveCity Project effort intends to create a city-based “Living Lab” and associated ecosystem to pilot live interactive high-definition video-to-video (v2v) on ultrafast wireless and wireline Internet infrastructure for the support of appropriate public service use cases among a number of city user communities initially in four major European cities (Dublin, Athens, Luxembourg (city) and Valladolid). The essential aim is to empower the citizens of a city to interact with each other in a more productive, efficient and socially useful way by using v2v over the Internet, as the latter can be considered to improve city administration, reduce fuel costs and carbon footprint, enhance education, improve city experiences for tourists/cultural consumers and save patients’ lives. LiveCity underpins technology which has the ability to massively scale while it integrates the necessary ingredients in an efficient low-cost manner and provides a proper testing ground for a mass market deployment to the cities in Europe.

The LiveCity Project effort intends to create a city-based “Living Lab” and associated ecosystem to pilot live interactive high-definition video-to-video (v2v) on ultrafast wireless and wireline Internet infrastructure for the support of appropriate public service use cases among a number of city user communities initially in four major European cities. The essential aim is to empower the citizens of a city to interact with each other in a more productive, efficient and socially useful way by using v2v over the Internet, as the latter can be considered to improve city administration, enhance education and city experiences for tourists/cultural consumers and save patients’ lives. We discuss the role that stakeholders can play in identifying appropriate KPIs to assess the progress of the LiveCity concept, covering the underlying network infrastructure, the intended services-facilities per specific case, as well as users’ satisfaction and requirements.

Recent advancements of IP networks pave the way for Over-the-Top (OTT) applications. Evolved telecom platforms provide revenue potentials via Service Gateways (APIs) on top of VoIP/RCS (IMS), Machine Type Communication (MTC) and Smart Bit pipe approaches. QoS is achieved through over-provisioning in today’s access and core networks since there are no flexible mechanisms that are available for end-users to influence the QoS level. Processes for user-demanded and operator-controlled QoS management as well as mechanisms for applications signaling their requirements on the data path into the network are far from being adequate. Novel approaches regarding end-to-end inter-domain flow-control architectures, i.e. network slicing, as well as machine-to-machine (M2M) virtualization platforms that handle such functions as device/communication management, session management and bearer and charging management are emerging promising enhanced multimedia communications and efficient utilization of network resources. They promote cloud services and they integrate the computer word into next generation telecommunications.

Future Internet and smart cities are creating a very promising paradigm for providing advanced services to citizens. The paradigm of e-Health forms a valuable yet demanding use case for design, develop, deploy and provide related services. The aim of LiveCity project is to empower the citizens of a city to interact with each other in a more productive efficient and socially useful way by using high quality video-to-video (v2v); v2v can be used to improve medical services. This paper presents the related concepts, the scenario and the pilot set for the tele-monitoring service realization, deployment and provision.

New technologies, viewing paradigms and content distribution approaches are about to take the TV/video services industry by storm. Five emerging trends are observable,

among which is the worldwide deployment of IP Version 6 (IPv6),

that are all related to the next-generation delivery of entertainment-quality video and can be capitalized upon by progressive service providers, telcos, cable operators, and ISPs. This work aims at exploring the IPv6-based evolving trends and offering practical suggestions of how it could support effective growth of video-to-video and mobile video communications. It also addresses an overview of IPv6; the rapid expansion of video-based solutions in the ICT market sector; IPv6 advantages for enhanced video communications as well as QoS issues from the use of IPv6 and IPv6 multicast approaches.

Current research and improvements in the field of wireless sensor networks are focused on decreasing the power consumption and miniaturization, improved smartness and better wearability of the sensor, and especially with their capability for environmental sensing. Today, the survival of these kinds of networks is a critical issue especially in order to keep environmental information updated. This paper presents, an improvement of the environmental sensing acquisition system shown in [1], by applying more sensors to gather data. It was found a novel method of reading sensor data using smartphones and also the structure of sensors themselves helps to decrease the power consumption of the network.

We illustrate the application of automated representation of knowledge acquired from sensor network data to quality of life services. Specifically, for a sensor network used to monitor a residential building we acquire knowledge about events of interest to occupants and represent such knowledge in ontology. An event of particular interest to quality of life which we discuss is ‘unhealthy’ exposure to carbon monoxide. Hence, we aim at reducing the considerable gap between raw sensor data and abstract domain terminology. Our results support the claim that computational techniques in signal processing, machine learning, and ontology engineering are important elements to systems that make use of environmental sensing, including systems for quality of life information services.

Environmental and meteorological conditions are of utmost importance for the population, as they are strongly related to the quality of life. Citizens are increasingly aware of this importance. This awareness results in an increasing demand for environmental information tailored to their specific needs and background. We present an environmental information platform that supports submission of user queries related to environmental conditions and orchestrates results from complementary services to generate personalized suggestions. From the technical viewpoint, the system discovers and processes reliable data in the web in order to convert them into knowledge. At run time, this information is transferred into an ontology-structured knowledge base, from which then information relevant to the specific user is deduced and communicated in the language of their preference. The platform is demonstrated with real world use cases in the south area of Finland showing the impact it can have on the quality of everyday life.

Analysis of environmental information is considered of utmost importance for humans, since environmental conditions are strongly related to health issues and to a variety of everyday activities. Despite the fact that there are already many free on-line services providing environmental information, there are several cases, in which the presentation format complicates the extraction and processing of such data. A very characteristic example is the air quality forecasts, which are usually encoded in image maps of heterogeneous formats, while the initial (numerical) pollutant concentrations, calculated and predicted by a relevant model, remain unavailable. This work addresses the task of semi-automatic extraction of such information based on a template configuration tool, on methodologies for data reconstruction from images, as well as on Optical Character Recognition (OCR) techniques. The framework is tested with a number of air quality forecast heatmaps demonstrating satisfactory results.

Air quality is one of the main priorities for the improvement of the life quality in urban regions, as air pollution is usually, concentrated in such densely populated areas. Most of the countries have a national air quality monitoring network that allow an analysis of the air quality status, especially for urban regions that are nodes in this network. As the network is geographically distributed, it can be mapped in a natural way on an intelligent agents based system. The paper describes the modeling framework of an air quality monitoring and analysis multiagent system for urban regions.

The quality of life is influenced by the environment quality and one of the major factors that require a continuous monitoring is the level of radiation. This paper presents a microcontroller-based system designated for monitoring the release of subatomic high frequency particles in the gamma ray area. Unlike the related scientific literature, where experiments are based on desktop computers for data processing and remote transmission, this work presents an independent standalone microcontroller-based system which incorporates the standard internet protocols. The measurement data, as well as warning signals, are sent to the decision making factors via a communication channel (such as internet, mobile phone or radio amateur band). Some experimental results are also discussed in the paper.

Air pollution can affect health and well-being of people and eco-systems. Due to the health risk posed for sensitive population groups, it is important to provide with hourly and daily forecasts of air pollution. One way to assess air pollution is to make use of the Common Air Quality Index (CAQI) of the European Environment Agency (EEA). In this paper we employ a number of Computational Intelligence algorithms to study the forecasting of the hourly and daily CAQI. These algorithms include artificial neural networks, decision trees and regression models combined with different datasets. The results provide with a satisfactory CAQI forecasting performance that may be the basis of an operational forecasting system.

The contribution of data mining to education as well as research in this area is done on a variety of levels and can affect the instructors’ approach to learning. This particular study focuses on problems associated with classification and attribute selection. An effort to forecast the results takes place before the educational process ends in order to prevent a potential learning failure.

The methodology used during the experiments excluded the case of overfitting and ensured the completion of the study. Particular emphasis was placed on analyzing the results, which demonstrated the superiority of the Pearson VII function kernel using the Support Vector Machines algorithm to the Bagging meta-learning method. We also determined the appropriate point in the course timeline in order to get reliable results regarding students’ outcome and finally, attribute selection gave us interesting results, in terms of students’ data.

In this study, the goal is multifold. At first we present a summarized review of terms and facts regarding the branch of ecoinformatics, web mining and the semantic web. In Section 2 we provide some related work derived from the current literature upon the web mining and the production of semantic content. The main part of our work follows presenting a notional model for building semantic content through 2-level web mining. This is achieved in web sites containing environmental data. We conclude mentioning the importance of this contribution from different points of view.

This paper presents a case-study of the effectiveness of a trained system into classifying Greek songs according to their audio characteristics or/and their lyrics into moods. We examine how the usage of different algorithms, featureset combinations and pre-processing parameters affect the precision and recall percentages of the classification process for each mood model characteristic. Experimental results indicate that the current selection of features offers accuracy results, the superiority of lyrics content over generic audio features as well as potential caveats with current research in Greek language stemming pre-processing methods.

Current uncontrolled growth of online, digital multimedia content emphasizes research work on identifying trends on how this content popularity may grow over time wrt identifiable user events and interests. In this paper we analyze user-generated photos uploaded to Flickr in order to extract meaningful semantic trends covering specific geographical areas of interest. Initially, we cluster photos based on their geo-tagging metadata information and divide large areas into smaller “first level geo-clusters” of fixed size, allowing them to overlap if necessary. Within these first level geo-clusters, we identify semantically meaningful “places” of user interest, by analyzing additional textual metadata, i.e. user selected tags that characterize each place’s photos. By post-processing them, we select the most appropriate tags that are able to describe landmarks and events occurring within these places of interest and examine their temporal dynamics over a long period of time.

In our days there is no single test or biomarker that can predict whether a particular person will develop Parkinson Disease and a definitive diagnosis is only possible after death, with postmortem analysis. Recent discoveries have highlighted that defects in mitochondrial dynamics are associated with neurodegenerative disease. In this paper the general architecture of an integrated multi-scale ontology-based modeling technology for early diagnosis of PD and for progress monitoring is proposed. The proposed model will be used to identify physics and biology-based biomarkers of processes occurring at radically different scales, from cell to whole body.

Web 2.0 has facilitated interactive information sharing on the WWW, allowing users the opportunity to articulate their opinions on different topics. In this framework, certain practices implement information monitoring systems so as digests, reports on keywords and thematic queries regarding opinions on government decisions to be created. Analysis of rubrics associations, primary semantic and statistical interpretation of the texts is usually carried out. It is, on the other hand, rather difficult to get punctual predicts and estimate sufficiently forum users’ opinion strength. In this work we present a methodology which automatically mines and estimates the strength of users’ opinions on text forums regarding government decisions. According to our methodology, quantitative features are automatically mined from forum posts and then passed to a Support Vector Machine based classifier where the users’ opinion strength is estimated. The proposed methodology has been validated in real data and initial experimental results are presented.

In this paper, we address the problem of melodic string matching that enables identification of varied (ornamented) instances of a given melodic pattern. To this aim, a new set of edit distance operations adequate for

pitch interval

strings is introduced. Insertion, deletion and replacement operations are abolished as irrelevant. Consolidation and fragmentation are retained, but adapted to the pitch interval domain, i.e., two or more intervals of one string may be matched to an interval from a second string through consolidation or fragmentation. The melodic interval string matching problem consists of finding all occurrences of a given pattern in a melodic sequence that takes into account exact matches, consolidations and fragmentations of intervals in both the sequence and the pattern. We show some properties of the problem and an algorithm that solves this problem is proposed.

In this contribution, we relate the question of discernability of sound structures to the properties of the underlying analysis tools. In particular, we argue, that classical tools that are mainly used in sound processing and lead to features as prominent as the MFCC may be replaced by more accurate methods that are based on rather recent mathematical signal processing tools. In particular, we focus on adaptive representations that lend themselves to efficient computation and, on the other hand, on sparsity-promoting methods which are able to adapt to the structures present in a particular signal class.

We study the cutting problems of meanders using 2-Motzkin words. These words uniquely define elevated peakless Motzkin paths, which under specific conditions correspond to meanders. A procedure for the determination of the set of meanders with a given sequence of cutting degrees, or with a given cutting degree, is presented by using proper conditions.

In this work, we study collective intelligence behavior of Web users that share and watch video content. We propose that the aggregated users’ video activity exhibits characteristic patterns that may be used in order to infer important video scenes thus leading to collective intelligence concerning the video content. In particular, we have utilised a controlled user experiment with information-rich videos for which users’ interactions (e.g., pause, seek/scrub) have been gathered. Modeling the collective information seeking behavior by means of the corresponding probability distribution function we argue that bell-shaped reference patterns are shown to significantly correlate with the predefined scenes of interest for each video, as annotated by the users. In this way, the observed collective intelligence may be used to provide a video-segment ranking tool that detects the importance of video scene. In practice, the proposed techniques might improve navigation within videos on the web and have also the potential to improve video search results with personalised video thumbnails.

This paper aims to analyze human navigation behavior and identify similarities of cognitive styles using measures obtained from psychometric tests. Specific navigation metrics are utilized to find identifiable groups of users that have similar navigation patterns in relation to their cognitive style. The proposed work has been evaluated with a user study that entails a psychometric-based survey for extracting the users’ cognitive styles, combined with a real usage scenario of users navigating in a controlled Web environment. A total of 84 participants of age between 17 and 25 participated in the study providing interesting insights with respect to cognitive styles and navigation behavior of users. Studies like the reported one can be useful for assisting adaptive interactive systems to organize and present information and functionalities in an adaptive format to diverse user groups.

The advent and wide proliferation of Social Web in the recent years has promoted the concept of social interaction as an important influencing factor of the way enterprises and organizations conduct business. Among the fields influenced is that of Enterprise Knowledge Management, where adoption of social computing approaches aims at increasing and maintaining at high levels the active participation of users in the organization’s knowledge management activities. An important challenge towards this is the achievement of the right balance between informalities of socially generated data and the required formality of enterprise knowledge. In this context, we focus on the problem of mining vague knowledge from social content generated within an enterprise framework and we propose a learning framework based on microblogging and fuzzy ontologies.

As mobile devices are always with users and music listening is a very personal and situational behaviour, contextual information could be used to greatly enhance music recommendations. However, making such use of context, while learning user profiles, is still a challenging problem. We present a system for collecting context and usage data from mobile devices, but targeted at recommending music according to learned user profiles and specific situations. The developed data flow system requires supporting both short enough response times and longer asynchronous reasoning on the collected data. Furthermore, the mobile phone acts not only as sensor, but is directly related to the effectiveness of the music service experience. Thus, this paper provides a description of our approach to the system and the initial results of a usability test of the mobile application and its backend system.

The present work aims at identifying relations between the morphosyntactic and semantic properties of an author’s writings and his/her personality traits. Machine learning schemata are used to classify an author according to the values of the Big Five traits, or predict their numerical value. Unlike related work, the current approach focuses on Modern Greek text, and makes use of limited data and resources, available at its disposal. Meta-learning and synthetic oversampling help overcome the small dataset and its imbalanced class distribution.

An implementation of the digital music stand (DMS) is presented in this invited tutorial. The DMS constitutes an optical music recognition (OMR) system whose principal aim is the custom-made on-screen presentation of music manuscripts (MMs). Aiming at performing the minimum necessary amount of MM processing, the DMS constitutes middle ground between full-processing classical commercial OMR systems/platforms and pure MM presentation systems. The DMS is intended for musicians and choir singers at all levels of music erudition, in both cases including those with visual impairments which cannot be corrected with glasses or contact lenses; music activities ranging from study, rehearsal and performance of individuals as well as of orchestras are accommodated. The DMS implementation stages are detailed, with performance accuracy, which is necessary for correct MM presentation, also being reported.

In the wake of gene-oriented data analysis in large-scale bioinformatics studies, focus in research is currently shifting towards the analysis of the functional association of genes, namely the metabolic pathways in which genes participate. The goal of this paper is to attempt to identify the core genes in a specific pathway, based on a user-defined selection of genomes. To this end, a novel methodology has been developed that uses data from the KEGG database, and through the application of the MCL clustering algorithm, identifies clusters that correspond to different “layers” of genes, either on a phylogenetic or a functional level. The algorithm’s complexity, evaluated experimentally, is presented and the results on a characteristic case study are discussed.

Nowadays, people frequently use search engines in order to find the information they need on the Web. Especially Web search constitutes a basic tool used by million researchers in their everyday work. A very popular indexing engine, concerning life sciences and biomedical research is PubMed. PubMed is a free database accessing primarily the MEDLINE database of references and abstracts on life sciences and biomedical topics. The present search engines usually return search results in a global ranking making it difficult to the users to browse in different topics or subtopics that they query. Because of this mixing of results belonging to different topics, the average users spend a lot of time to find Web pages, best matching their query. In this paper, we propose a novel system to address this problem. We present and evaluate a methodology that exploits semantic text clustering techniques in order to group biomedical document collections in homogeneous topics. In order to provide more accurate clustering results, we utilize various biomedical ontologies, like MeSH and GeneOntology. Finally, we embed the proposed methodology in an online system that post-processes the PubMed online database in order to provide to users the retrieved results according to well formed topics.

The integration of gene expression datasets with gene function information provides valuable insights in unraveling the molecular mechanisms of the brain. In this paper, gene expression maps, acquired by the technique of voxelation, are analyzed using an atlas-based framework and the extracted spatial information is employed to organize genes in significant clusters. Moreover, gene function enrichment analysis of clusters enables exploring the relationships among brain regions, gene expressions and gene functions. Our work confirms the hypothesis that genes of similar spatial expression patterns display similar functions indicating that our methodology could assist in the functional identification of unannotated genes.

One of the most significant issues of the computational biology is the multiple pattern matching for locating nucleotides and amino acid sequence patterns into biological databases. Sequential implementations for these processes have become inadequate, due to an increasing demand for more computational power. Graphic cards offer a high parallelism computational power improving the performance of applications. This paper evaluates the performance of the Wu-Manber algorithm implemented with the OpenCL framework, by presenting the running time of the experiments compared with the corresponding sequential time.

In the post-genomic era there has been an explosion in the amount of genomic data available and the primary research problems have moved from being able to produce interesting biological data to being able to efficiently process and store this information. In this paper we present efficient data structures and algorithms for the

High Similarity Sequencing Problem

. In the

High Similarity Sequencing Problem

we are given the sequences

S

0

,

S

1

, …,

S

k

where

S

j

=

$e_{j_1} I_{\sigma_1}e_{j_2} I_{\sigma_2} e_{j_3} I_{\sigma_3}, \dots,e_{j_\ell} I_{\sigma_\ell}$

and must perform pattern matching on the set of sequences. In this paper we present time and memory efficient datastructures by exploiting their extensive similarity, our solution leads to a query time of

$O(m + vk \log \ell + \frac{m occ_v v}{w} + \frac{PSC(p)m}{w})$

with a memory usage of

O

(

N

log

N

 + 

vk

log

vk

).

Pairwise sequence alignment has received a new motivation due to the advent of next-generation sequencing technologies, particularly so for the application of

re-sequencing

—the assembly of a genome directed by a reference sequence. After the fast alignment between a factor of the reference sequence and a high-quality fragment of a short read by a short-read alignment programme, an important problem is to find the alignment between a relatively short succeeding factor of the reference sequence and the remaining low-quality fragment of the read allowing a number of mismatches and the insertion of a single gap in the alignment. In this article, we present

GapMis-OMP

, a tool for pairwise short-read alignment that works on multi-core architectures. It is designed to compute the alignments between all the sequences in a first set of sequences and all those from a second one in parallel. The presented experimental results demonstrate that

GapMis-OMP

is more efficient than most popular tools.

National/Ethnic population Mutation databases (NEMDBs) are online mutation depositories recording extensive information about the described genetic heterogeneity in populations and ethnic groups worldwide. FINDbase (

http://www.findbase.org

) is a database containing causative mutations and pharmacogenomic markers allele frequencies in various populations and ethnic groups. In this paper, we experiment with designing and applying new automated data mining techniques on the original FINDbase causative mutations data collection in an attempt to identify genomic relationships between populations. Furthermore, we have developed an interactive web-based visualization tool that enables users to correlate the information, determine the relationships and gain insight into the underlying data collection in a novel and meaningful way.

Proteins and their interactions are considered to play a significant role in many cellular processes. The identification of Protein-Protein interactions (PPIs) in human is an open research area. Many Databases, which contain information about experimentally and computationally detected human PPIs as well as their corresponding annotation data, have been developed. However, these databases contain many false positive interactions, are partial and only a few of them incorporate data from various sources. To overcome these limitations, we have developed HINT-KB (http://150.140.142.24:84/Default.aspx) which is a knowledge base that integrates data from various sources, provides a user-friendly interface for their retrieval, estimates a set of features of interest and computes a confidence score for every candidate protein interaction using a modern computational hybrid methodology.

Protein structure similarity is one of the most important aims pursued by bioinformatics and structural biology, nowadays. Although quite a few similarity methods have been proposed lately, yet fresh algorithms that fulfill new preconditions are needed to serve this purpose. In this paper, we provide a new similarity measure for 3D protein structures that detects not only similar structures but also similar substructures to a query protein, supporting both multiple and pairwise comparison procedures and combining many comparison characteristics. In order to handle similarity queries we utilize efficient and effective indexing techniques such as M-trees and we provide interesting results using real, previously tested protein data sets.

Until recently, it was commonly accepted that most genetic information is transacted by proteins. Recent evidence suggests that the majority of the genomes of mammals and other complex organisms are in fact transcribed into non-coding RNAs (ncRNAs), many of which are alternatively spliced and/or processed into smaller products. Non coding RNA genes analysis requires the calculation of several sequential, thermodynamical and structural features. Many independent tools have already been developed for the efficient calculation of such features but to the best of our knowledge there does not exist any integrative approach for this task. The most significant amount of existing work is related to the miRNA class of non-coding RNAs. MicroRNAs (miRNAs) are small non-coding RNAs that play a significant role in gene regulation and their prediction is a challenging bioinformatics problem. Non-coding RNA feature extraction and pre-miRNA classification Web Tool (ncRNA-class Web Tool) is a publicly available web tool (

http://150.140.142.24:82/Default.aspx

) which provides a user friendly and efficient environment for the effective calculation of a set of 58 sequential, thermodynamical and structural features of non-coding RNAs, plus a tool for the accurate prediction of miRNAs.

Muscle-specific kinase is a crucial receptor tyrosine kinase required for the development and function of neuromuscular junction. Although many protein domains have already been modeled with crystallographic techniques in various organisms, a single model for the whole human structure is not yet available. A model of the entire protein was constructed by using two parallel Homology Modeling approaches, one unsupervised and one driven by the user. In addition, by applying Molecular Dynamics simulations the present study provides further insights on the structure, and the intermolecular interactions of the protein were examined. The expected semi rigid globular form of the protein was confirmed and in addition a hydrophobic core and a hydrogen bond network that enhances the stability of the molecule were observed. Furthermore, these calculations identified an intriguing rotation of Ig domains and this finding sets the base for additional hypothesis and further investigation.