Dimitris Kanellopoulos

Combining Bagging, Boosting and Dagging for Classification Problems

Bagging, boosting and dagging are well known re-sampling ensemble methods that generate and combine a diversity of classifiers using the same learning algorithm for the base-classifiers. Boosting algorithms are considered stronger than bagging and dagging on noise-free data. However, there are strong empirical indications that bagging and dagging are much more robust than boosting in noisy settings. For this reason, in this work we built an ensemble using a voting methodology of bagging, boosting and dagging ensembles with 8 sub-classifiers in each one. We performed a comparison with simple bagging, boosting and dagging ensembles with 25 sub-classifiers, as well as other well known combining methods, on standard benchmark datasets and the proposed technique had better accuracy in most cases.

Exploiting tourism destinations' knowledge in an RDF-based P2P network

Destination Management Systems (DMS) is a perfect application area for Semantic Web and P2P technologies since tourism information dissemination and exchange are the key-backbones of tourism destination management. DMS should take advantage of P2P technologies and semantic web services, interoperability, ontologies and semantic annotation. RDF-based P2P networks allow complex and extendable descriptions of resources instead of fixed and limited ones, and they provide query facilities against these metadata instead of simple keyword-based searches. The layered adaptive semantic-based DMS (LA_DMS) and Peer-to-Peer (P2P) project aims at providing semantic-based tourism destination information by combining the P2P paradigm with Semantic Web technologies. In this paper, we propose a metadata model encoding semantic tourism destination information in an RDF-based P2P network architecture. The model combines ontological structures with information for tourism destinations and peers.

How can teleworking be pro‐poor?

Purpose – The purpose of the paper is to examine evidence in order to discover if teleworking has a pro‐poor growth impact – reducing inequality. For this reason, the paper seeks to propose a telework taxonomy for the poor and research questions that trigger future empirical research on poor teleworkers.Design/methodology/approach – The papers approach is a literature review. The focused literature includes articles that analyze telework issues with a potential for the poor. Such issues are mainly workforce and organizational issues.Findings – There is some evidence that provision of teleworking infrastructure has a dramatic effect on the income and quality of life of the rural poor. Special knowledge management tasks and types of telework can be proper for poor people. Economic and organizational aspects of telecentres for poor workers must be analyzed in depth.Research limitations/implications – The paper provides a foundation for future research directions in the teleworking domain for the poor. For i...

A Novel Contention Window Control Scheme for IEEE 802.11 WLANs

Abstract In the IEEE 802.11 standard, network nodes experiencing collisions on the shared medium need a mechanism that can prevent collisions and improve the throughput. Furthermore, a backoff mechanism is used that uniformly selects a random period of time from the contention window (cw) that is dynamically controlled by the Binary Exponential Backoff (BEB) algorithm. Prior research has proved that the BEB scheme suffers from a fairness problem and low throughput, especially under high traffic load. In this paper, we present a new backoff control mechanism that is used with the IEEE 802.11 distributed coordination function (DCF). In particular, we propose a dynamic, deterministic contention window control (DDCWC) scheme, in which the backoff range is divided into several small backoff sub-ranges. In the proposed scheme, several network levels are introduced, based on an introduced channel state vector that keeps network history. After successful transmissions and collisions, network nodes change their cw based on their network levels. Our extensive simulation studies show that the DDCWC scheme outperforms four other well-known schemes: Multiplicative Increase and Linear Decrease, Double Increment Double Decrement, Exponential Increase Exponential Decrease, and Linear/Multiplicative Increase and Linear Decrease. Moreover, the proposed scheme, compared with the IEEE 802.11 DCF, gives 30.77% improvement in packet delivery ratio, 31.76% in delay, and 30.81% in throughput.

Local Boosting of Decision Stumps for Regression and Classification Problems

Numerous data mining problems involve an investigation of associations between features in heterogeneous datasets, where different prediction models can be more suitable for different regions. We propose a technique of boosting localized weak learners; rather than having constant weights attached to each learner (as in standard boosting approaches), we allow weights to be functions over the input domain. In order to find out these functions, we recognize local regions having similar characteristics and then build local experts on each of these regions describing the association between the data characteristics and the target value. We performed a comparison with other well known combining methods on standard classification and regression benchmark datasets using decision stump as based learner, and the proposed technique produced the most accurate results.

MAC layer misbehavior in MANETs

Abstract In mobile ad hoc networks (MANETs), the IEEE 802.11 CSMA/CA is deployed as the primary medium-access control (MAC) layer protocol to schedule the access to the wireless medium. The IEEE 802.11 standard was designed with the assumption that nodes would never deviate from the protocol. However, MANET nodes may purposefully show misbehavior at the MAC layer to obtain more bandwidth or degrade the network performance and disrupt the network services. This paper reviews and classifies the most important strategies generating MAC layer misbehavior based on their objectives and operating principles. Then, it examines some of the recent proposed solutions and mechanisms for detecting and preventing MAC layer misbehavior. A comparison of the studied solutions is carried out using a set of critical evaluation metrics. Finally, the paper concludes with a brief summary of key ideas and a general direction that can provide a basis for future work.

Evaluating and Recommending Greek Newspapers' Websites Using Clustering.

Purpose – The aim of this work is to evaluate Greek newspaper websites using clustering and a number of criteria obtained from the Alexa search engine. Furthermore, a recommendation approach is proposed for matching Greek online newspapers with the profiles of potential readers. The paper presents the implementation and validation of a recommender tool that suggests to a user (based on age, education and income) an optional Greek newspapers website to read.Design/methodology/approach – A total of 25 newspaper websites were selected from the Greek information bank http://edoellada.gr. After investigating these websites one by one, this number was decreased to 16 websites due to their printing prevention, cessation or no coverage by Alexa.Findings – Based on data obtained from Alexa, the Naftemporiki newspaper has the highest traffic rank and the Eleftherotypia newspaper the largest number of links among others. The Macedonia newspaper has the largest number of foreign users. The results of the study also ...

Editorial: special issue on ubiquitous electronic commerce systems

Ubiquitous computing is a post-desktop model of human-computer interaction in which information processing has been thoroughly integrated into everyday objects and activities. Emerging ubiquitous electronic commerce systems (UECS) are expected to be available anytime, anywhere, and using different official or personal computing devices. Systems and services such as digital libraries, on-line business transactions, mobile office and mobile TV are widely deployed. Users will be able to access these services anytime, anywhere, while using any computing device in a pervasive way. For example, a user may bring a PDA on a field trip, carry a laptop (with both wireless and wired network cards) on a business trip, use high performance workstations at work, and use desktop PCs at home (with dial-up, cable, or DSL connection). In another example, a user continues to watch the soccer game over home TV, while he watches the game through a mobile device out of home. Some other

Bagged Averaging of Regression Models

Linear regression and regression tree models are among the most known regression models used in the machine learning community and recently many researchers have examined their sufficiency in ensembles. Although many methods of ensemble design have been proposed, there is as yet no obvious picture of which method is best. One notable successful adoption of ensemble learning is the distributed scenario. In this work, we propose an efficient distributed method that uses different subsets of the same training set with the parallel usage of an averaging methodology that combines linear regression and regression tree models. We performed a comparison of the presented ensemble with other ensembles that use either the linear regression or the regression trees as base learner and the performance of the proposed method was better in most cases.

Intelligent Multimedia Technologies for Networking Applications: Techniques and Tools

As ubiquitous multimedia applications benefit from the rapid development of intelligent multimedia technologies, there is an inherent need to present frameworks, techniques and tools that adopt these technologies to a range of networking applications.Intelligent Multimedia Technologies for Networking Applications: Techniques and Tools promotes the discussion of specific solutions for improving the quality of multimedia experience while investigating issues arising from the deployment of techniques for adaptive video streaming. This reference source provides relevant theoretical frameworks and leading empirical research findings and is suitable for practitioners and researchers in the area of multimedia technology.