• European University of Technology
• External research report close

Clearly, machine learning techiques can play an important role in filtering spam email because ample training data is available to build a robust classifier. However, spam filtering is a particularly challenging task as the data distribution and concept being learned changeds over time. This is a particularly awkward form of concept drift as the change is driven by spammers wishing to circumvent the spam filters. In this paper we show that lazy learning techniques are appropriate for such dynamically changing contexts. We present a case-based system for spam filtering called ECUE that can learn dynamically. We evaluate its performance as the case-base is updated with new cases. We also explore the benefit of periodically redoing the feature selection process to bring new features into play. Our evaluation shows that these two levels of model update are effective in tracking concept drift.

Producing estimates of classification confidence is surprisingly difficult. One might expect that classifiers that can produce numeric classification scores (e.g. k-Nearest Neighbour or Naive Bayes) could readily produce confidence estimates based on thresholds. In fact, this proves not to be the case, probably because these are not probabilistic classifiers in the strict sense. The numeric scores coming from k-Nearest Neighbour or Naive Bayes classifiers are not well correlated with classification confidence. In this paper we describe a case-based spam filtering application that would benefit significantly from an ability to attach confidence predictions to positive classifications (i.e. messages classified as spam). We show that `obvious? confidence metrics for a case-based classifier are not effective. We propose an ensemble-like solution that aggregates a collection of confidence metrics and show that this offers an effective solution in this spam filtering domain.

Because of the changing nature of spam, a spam filtering system that uses machine learning will need to be dynamic. This suggests that a case-based (memory-based) approach may work well. Case-Based Reasoning (CBR) is a lazy approach to machine learning where induction is delayed to run time. This means that the case base can be updated continuously and new training data is immediately available to the induction process. In this paper we present a detailed description of such a system called ECUE and evaluate design decisions concerning the case representation. We compare its performance with an alternative system that uses Naive Bayes (NB). We find that there is little to choose between the two alternatives in cross-validation tests on data sets. However, ECUE does appear to have some advantages in tracking concept drift over time.

Following the recent terrorist attacks in Paris, the European media emphatically pronounced that billions of euros were wiped from tourism related stocks. This comes at a troublesome time for the tourism industry, in the midst of a global financial crisis, and the unpredictable rise of radical Islamic ideologies, which have caused chaos in the Middle East and Europe. The relationship and vulnerability of the industry to non-macro incidents have been well documented in the literature, mostly in theoretical terms. Nevertheless, the quantifiable impact of such events on tourism-specific stock values, both in terms of returns and volatility, received much less attention. With the use of an econometric methodology, the paper aims to enhance our conceptual capital pertaining to the effects of such possibilities on five hospitality and tourism stock indices. The empirical findings are of interest to stakeholders at all echelons of the spectra of the tourism and financial industries.

Welcome to ICMS: Global Consistency in Presenting Construction Costs ICMS aims to provide global consistency in classifying, defining, analysing and presenting construction costs at a project, regional, state, national or international level. This project is the first of its kind, bringing together numerous organisations from around the world to create shared international standards for presenting construction costs. This first edition of ICMS focuses on capital costs; however, future editions of ICMS may incorporate other matters such as costs in use. Consistent practice in presenting construction costs globally would bring significant benefits to construction cost management. Globalisation of the construction business has increased the need to make meaningful comparative analysis between countries, not least by international organisations such as the World Bank Group, the International Monetary Fund, various regional development banks, non-governmental organisations and the United Nations. The Coalition did not identify any existing standard that was suitable for international adoption. For this reason, the Coalition has come together to create shared standards. At a meeting at the International Monetary Fund in June 2015, Coalition members confirmed they were committed to promoting the implementation of ICMS to encourage world markets to accept and adopt ICMS as the primary standard for presenting construction costs across different nations in a consistent way. An independent Standard Setting Committee (the SSC) was formed. The SSC includes technical experts from 11 countries and a combined expertise covering 47 different markets. The SSC worked virtually and also met three times, once in Brussels and twice in London. A wide range of professional organisations are represented in the Coalition and the SSC. They were generous in providing their national standards, which provided the basis for the early deliberations of the SSC. The SSC produced the complete consultation draft of ICMS within a year, in July 2016. Following the private consultation period in October 2016, two public consultations took place between November 2016 and April 2017. The completed first edition was published in July 2017. The Coalition accepts that standard setting is a continuous and dynamic process; it will be listening closely to the global construction cost management community to ensure necessary updates are captured for continued improvement. In addition to preparing further editions of ICMS for additional categories of civil engineering projects, the SSC will also monitor all guidance notes on ICMS to ensure that they are consistent with the principles and intent of ICMS. All local, regional or worldwide approaches will be documented to allow coordination, expansion and consistency of ICMS guidance whenever required. The Coalition is beginning the important work of implementation by liaising with governments on a national, regional or state and local level to seek adoption of ICMS. Many key stakeholders are being engaged in the process of implementation. A list of ICMS-supporting partners is shown on the ICMS Coalition website (www.icms-coalition.org) – these organisations are committed to the adoption of ICMS. The Coalition, the SSC and the numerous participants in the consultation are proud to present ICMS. For further information on ICMS please visit the website (www.icms-coalition.org). On behalf of the ICMS Coalition Trustees: Ken Creighton – (Royal Institution of Chartered Surveyors) - Chair Martin Darley – (Association for the Advancement of Cost Engineering) - Vice Chair Julie Dela Cruz – (Philippine Institute of Certified Quantity Surveyors) – General Secretary Craig Bye – (Canadian Institute of Quantity Surveyors) - General Secretary

Presented at 1998 German Workshop on Case-Based Reasoning Software project cost estimation is difficult because of problems of quantifying project size and because of the continual emergence of new technology. This presents as a classic example of a weak theory domain where experience is key and appears well suited to CBR. Indeed there are several reports in the literature on the use of CBR in project cost estimation. Research described to date has focused on applications late in the development life cycle and in a narrow domain. In this paper we describe existing research and explore the use of CBR beyond these limits. We look at the use of CBR for estimation earlier in the life cycle and for use in broader domains where more abstract remindings need to be supported.

The problem of concept drift has recently received considerable attention in machine learning research. One important practical problem where concept drift needs to be addressed is spam filtering. The literature on concept drift shows that among the most promising approaches are ensembles and a variety of techniques for ensemble construction has been proposed. In this paper we consider an alternative lazy learning approach to concept drift whereby a single case-based classifier for spam filtering keeps itself up-to-date through a case-base maintenance protocol. We present an evaluation that shows that the case-base maintenance approach is more effective than a variety of ensemble techniques. The evaluation is complicated by the overriding importance of False Positives (FPs) in spam filtering. The ensemble approaches can have very good performance on FPs because it is possible to bias an ensemble more strongly away from FPs than it is to bias the single classifer. However this comes at considerable cost in overall accuracy. This research was supported by funding from Enterprise Ireland under grant no. CFTD/03/219 and funding from Science Foundation Ireland under grant no. SFI-02IN.1I111

Because of the volume of spam email and its evolving nature, any deployed Machine Learning-based spam filtering system will need to have procedures for case-base maintenance. Key to this will be procedures to edit the case-base to remove noise and eliminate redundancy. In this paper we present a two stage process to do this. We present a new noise reduction algorithm called Blame-Based Noise Reduction that removes cases that are observed to cause misclassification. We also present an algorithm called Conservative Redundancy Reduction that is much less aggressive than the state-of-the-art alternatives and has significantly better generalisation performance in this domain. These new techniques are evaluated against the alternatives in the literature on four datasets of 1000 emails each (50% spam and 50% non spam). This research was supported by funding from Enterprise Ireland under grant no. CFTD/03/219 and funding from Science Foundation Ireland under grant no. SFI-02IN.1I111.