Suchergebnisse

pam is a universal problem with which everyone is familiar. Figures published in 2005 state that about 75% of all email sent today is spam. In spite of significant new legal and technical approaches to combat it, spam remains a big problem that is costing companies meaningful amounts of money in lost productivity, clogged email systems, bandwidth and technical support. A number of approaches are used to combat spam including legislative measures, authentication approaches and email filtering. The most common filtering technique is content-based filtering which uses the actual text of the message to determine whether it is spam or not. One of the main challenges of content based spam filtering is concept drift; the concept or the characteristics used by the filter to identify spam email are constantly changing over time. Concept drift is very evident in email and spam, in part due to the arms race that exists between spammers and the filter producers. The spammers continually change the content and structure of the spam emails as the filters are modified to catch them. In this thesis we present Email Classification Using Examples (ECUE) a content based approach to spam filtering that can handle the concept drift inherent in spam email. We apply the machine learning technique of case-based reasoning which models the email as cases in a knowledge-base or case-base. The approach used in ECUE involves two components; a case-base editing stage and a case-base update policy. We present a new technique for case-base editing called Competence-Based Editing which uses the competence properties of the cases in the case-base to determine which cases are harmful to the predictive power of the case-base and should be removed. The update policy allows new examples of spam and legitimate emails to be added to the case-base as they are encountered allowing ECUE to track the concept drift. We compare the case-based approach to an ensemble approach which is a more standard technique for handling concept drift and present a prototype email filtering application that demonstrates how the ECUE approach to spam filtering can handle the concept drift.

The availability of the full text of Irish parliamentary questions offers opportunities for using machine learning techniques to examine the currently much discussed role of elected representatives (TDs) in the Irish parliamentary system. Bluntly, are TDs mainly national legislators or "constituency messenger boys"? This paper presents an initial investigation into the use of automated text classification techniques to categorise parliamentary questions from 1922 up to 2008 as national or local. The approach uses a bag of words representation, standard feature reduction methods and an SVM classifier. Initial results show there is very little evidence in the corpus of parliamentary questions in Dail Eireann to support the view that the role of the TD is determined by mainly clientelist/parochial imperatives.

In this paper we propose a framework to generate artificial data that can be used to simulate credit risk scenarios. Artificial data is useful in the credit scoring domain for two reasons. Firstly, the use of artificial data allows for the introduction and control of variability that can realistically be expected to occur, but has yet to materialise in practice. The ability to control parameters allows for a thorough exploration of the performance of classification models under different conditions. Secondly, due to non-disclosure agreements and commercial sensitivities, obtaining real credit scoring data is a problematic and time consuming task. By the provision of publicly available artificial data, credit scoring is opened to the wider data mining community. This in turn could help enable greater participation, promote replicable experimental findings, and give rise to solution proposals to outstanding credit scoring problems. To ensure that our framework is sufficiently grounded in reality, data distributions are generated using a troika of sources: demographic information from the Central Statistics Office, Ireland; housing statistics published by the Irish Government Department of the Environment, Heritage and Local Government; and a profile of loan defaulters developed using a recent report published by a credit rating agency. By engaging with a credit scoring expert we select characteristics that are typical of most application scorecard models including, amongst others: age, income, loan value, and occupation. Through user controlled settings the conditional prior probabilities of the characteristics can be adjusted over time to simulate differing scenarios. In order to assign class labels to the generated data a credit risk score is estimated based on the non-linear interactions between various characteristics. Based on the desired number of defaulters a cut-off score is placed on this monotonic ordering of credit scores to distinguish between those likely to repay and those likely to default on their financial obligation. The classification complexity is controlled by adding user-defined random Gaussian noise. After discussing the desirable characteristics of artificial data we describe a pseudo-random data generator for credit scoring and provide illustrations on how the framework can be used to generate population drift.

While in recent decades a number of efforts have been coordinated to address the issue of gender imbalance in STEM (science, technology, engineering and mathematics) disciplines, the problem still persists. Many authors speak of the 'leaky' pipeline metaphor that describes the loss of women in STEM areas before reaching senior roles. Research shows that women who leave are unlikely to return. The issue is particularly severe in the area of computer science, where women represent less than 20% of the labour force across the EU. This poster introduces a summary of findings from the literature on how to effectively recruit and retain women in computer science education. The focus was to identify initiatives with demonstrated impact that targeted mainly female undergraduate computing, computer science and technology students. This work considers the initiatives and interventions reported by the academic community, but also includes success stories from the non-academic sources around the globe, such as international equality awards submissions and online reports from universities, non-profit organisations and personal practices. Sources in languages other than English have also been considered. Practical initiatives that showed impact (which we call 'Actions') were identified and categorised into four groups comprising Policy, Pedagogy, Promotion and Influence. Each category is arranged into sub-groups and examples of actions that showed impact are provided for each of these groups. This research aims to help the wider community to get one step closer towards gender balance in computer science.