Research and Enterprise Projects
| Home | Publications | Research & CV | Projects | Teaching |
| |
Research and Enterprise Projects |
||||
|
This project focuses on the use of bespoke process modelling methods and approaches which have been developed specifically for modelling multi-agency public sector projects. Such projects require a number of perspectives, for a variety of stakeholders, yet still need to have integrated and whole models, which are sufficiently rigorous to provide genuine process control. The project is carried out in conjunction with Dorset Police, though the modelling also involves a number of other agencies.
Rich Internet Application products are an important and growing area, yet both current software development and web development methods do not address sufficiently the particular needs of these kinds of development projects. In addition, the opportunities afforded by rich internet technologies, in terms of what they might offer to customers, particularly those upgrading from traditional applications may not be realised fully with current approaches. This project is funded through the Technology Strategy Board as a Knowledge Transfer Partnership project with Hark Solutions. The project is carried out by a team from the Software Systems Research Centre, Coles, Phalp, Jeary, and Fakorede, in conjunction with Hark.
This project involves reverse engineering a complex and major product which has been developed over a number of years. Aside from the architectural re-engineering an important challenge is to discover, extract and describe the business rules, the structure and the processes from the product and to document them in an understandable manner, so a new upgraded product based on this tacit knowledge can be built. In essence, not throwing out the baby with the bath water. This project is funded through the Technology Strategy Board as a Knowledge Transfer Partnership project with Morning Data. The project is carried out by a team from the Software Systems Research Centre, Coles, Jeary and Phalp in conjunction with the team at Morning Data.
The VIDE project, was funded, as part of framework 6, by the European Commission, to a total of 2.3 million Euros. The project's final review was completed in February 2009, with the project being rated as very good. The project overall provides a rich and accessible modelling toolset, allowing those who are not IT specialists to be involved in the description and specification of system behaviour, as part of a model driven software development process. Bournemouth’s work contributes strongly in the modelling, visualisation and interface design aspects of the project. Dr Phalp led the BU team at the ‘kick off’ meeting, and played an active role, both in steering academic direction, and contributing directly to deliverables, utilising his expertise of business process and software modelling.
Model Driven Development should support collaboration, such that different
team members situated at different sites should be able to work with the same
models and diagrams. However, for distributed development projects of software
product lines, such as found within Bosch, collaboration at the model layer
is often difficult to achieve because the layout of a diagram can be modified
in different ways by different team members working independently on the same
diagrams. Our industrial studies have confirmed that modellers attach significant
meaning to diagram layout, hence, when automatic layout algorithms are used
for merging much of the meaning (to the modellers) is lost. Bosch considers
the resolution of such issues important to their future software development,
and has funded fully a PhD student (Frank Grimm) to carry out the work, supervised
by Dr Phalp. We have introduced a semi-automatic approach to model and diagram
merging, which uses a novel (meta model driven) approach to ensure that much
of the meaning (from the modellers’ perspectives) is preserved, thus allowing
an effective model based development process across distributed sites.
Understanding the requirements of stakeholders is a common issue. Solution
systems are provided by developers, whose values are not always in line with
the wants or needs of various stakeholders or the organisation. Solutions need
to be akin to company values, available resources and the environment in which
the problem exists. Current research stems around the theme of the Model Driven
Architecture (MDA). Focus in the MDA is given to software engineering paradigms
being applied to the business user. In this research, ways to realise business
paradigms within the MDA via methodology and transformation are examined, with
a view to bridging the gap between business and software models. Requirements
Engineering ideals are suggested to significantly extend the MDA, thereby aligning
IT with business strategy. This allows for organisations to re-evaluate the
business process in terms of IT and / or even the business strategies themselves,
whilst enabling the system specification be a direct output product of the business
users involvement by visualisation means. This work was funded as one of the
Bournemouth University studentships.
Sherry is towards the final stage of her PhD on the requirements for and the use of web development methods. This has proved an interesting topic area where previous research has found that web development methods are not used in practice. Sherry's work has confirmed the assertion that academic web development methods are too difficult to use. In addition, the scope of the methods have not covered the full lifecycle, they are written in academic language and make incorrect assumptions as to the background and experience of prospective users. She has created a method which particularly explores the requirements and design phase of the lifecycle which has been used successfully in practice.
The use of microarray gene expression data to diagnose cancer patients has increased dramatically over the past decade, indicating an urgent need for the development of treatment measures for potential causes of diseases at a genomic level. However, due to the large amount of noise generated during microarray experiments, techniques for extracting informative genes that underlying the pathogenesis of the tumour cell from microarray data become necessary and the need of computing algorithm to undertaking such complex task emerge naturally. The goal of this research is to devise a more effective way for extracting significant features using genetic algorithms and artificial neural networks due to their learning abilities to construct hypotheses that can explain complex relationships in microarray data. This research explores the effectiveness of genetic algorithm-neural network (GANN) hybrid in analysing gene expression activities based on a specific tumour disease and identifying informative genes underlie different phases of tumour proliferation which could be vital for designing treatment strategies for patients.
This work, which was submitted in 2009, provides an investigation into the use of natural search strategies for improving the performance of autonomous vehicles operating in a search role. Research is reported that explores whether autonomous vehicle search can be enhanced by applying natural search behaviours for a variety of search targets. Having identified useful search behaviours, the work considers scenarios where detection is lost and whether natural strategies for re-detection can improve overall systemic performance. Analysis of empirical results indicates that search strategies exploiting behaviours found in nature can improve performance over random search and commonly applied systematic searches, across a variety of relative target speeds and against various target movement types. Experiments with target occlusion also reveal that natural reacquisition strategies could improve the probability of target redetection.
This research project considers the application of machine learning to intrusion detection, which in this context refers to detecting malicious behaviour in computer systems or computer networks. In particular for computer networks, the magnitude of data is generally so immense that it becomes an impossible task for a human being to handle; the amount of data will grow quicker than it can be analysed. Thus, one of the main goals of IDSs can be considered as automating the detection process. The project finds that the many of the results obtained in previous research are contradictory, and attempts to shed more light on why this might be so, finding that the performance of the techniques varies significantly in different circumstances, as use of the data set is altered. Empirical investigation reveals what impact various issues with the data set has on the performance of the techniques, explores ways of dealing with these issues, and discusses implications. Although not previously acknowledged in this domain, class imbalance has been found to be a significant challenge to most of the machine learning classifiers adopted in the literature. One part of this research focused on demonstrating that this is indeed an issue for intrusion detection, affecting the performance of the classifiers. Further from this, a novel approach to learning from imbalanced data has been proposed, using multi-objective genetic algorithms to evolve artificial neural networks and classifier ensembles.
This project produced enhanced and ‘enactable’ use case descriptions and a supporting software tool. The work followed from our project on analysis of requirements techniques, where we found that use cases lacked information about the dependencies amongst events. However, standard ways of providing such information required the production of a further model or models. This project developed tool support to ease the analysis of use cases, by using simple state information (from an annotated description) to control the logic of an enaction, and thus highlight such dependencies. In addition, the work also supported alignment of business goals and requirements by allowing the ‘enhanced’ specifications to capture the richness of process models, thus minimising the information loss in moving from process model to specification. The work, which is an ongoing theme, has led to number of paper outputs (including a recent paper in the requirements engineering journal in 2009) and a PhD completion (Kanyaru). In addition, our experience in both the production of such tools was important in our acceptance as part of the successful VIDE consortium (see VIDE project).
This project sought to gauge the utility of modern approaches to requirements engineering. In particular, the use case approach was investigated and the project analysed existing use case description rules before finally proposing an improved set of writing guidelines. One PhD student was employed, and completed successfully in 2002 (Karl Cox). The work used empirical software engineering approaches, such as experiment and case study, and led to a number of publications at relevant conferences and journals. This research theme continues within the group, for example, in 2007 three journal papers were published in this area - one in the Journal of Systems and Software (JSS) and two in the Software Quality Journal (SQJ), and a paper on comprehension, use cases and requirements, was presented by Phalp at the Software Quality Management Conference in 2009. In addition, lessons learned led to our successful work on enactable use case specifications and on software tools to support these approaches.
