Chris Price

A Link-Based Approach to the Cluster Ensemble Problem

Cluster ensembles have recently emerged as a powerful alternative to standard cluster analysis, aggregating several input data clusterings to generate a single output clustering, with improved robustness and stability. From the early work, these techniques held great promise; however, most of them generate the final solution based on incomplete information of a cluster ensemble. The underlying ensemble-information matrix reflects only cluster-data point relations, while those among clusters are generally overlooked. This paper presents a new link-based approach to improve the conventional matrix. It achieves this using the similarity between clusters that are estimated from a link network model of the ensemble. In particular, three new link-based algorithms are proposed for the underlying similarity assessment. The final clustering result is generated from the refined matrix using two different consensus functions of feature-based and graph-based partitioning. This approach is the first to address and explicitly employ the relationship between input partitions, which has not been emphasized by recent studies of matrix refinement. The effectiveness of the link-based approach is empirically demonstrated over 10 data sets (synthetic and real) and three benchmark evaluation measures. The results suggest the new approach is able to efficiently extract information embedded in the input clusterings, and regularly illustrate higher clustering quality in comparison to several state-of-the-art techniques.

Model-based systems in the automotive industry

The automotive industry was the first to promote the development of applications of model-based systems technology on a broad scale and, as a result, has produced some of the most advanced prototypes and products. In this article, we illustrate the features and benefits of model-based systems and qualitative modeling by prototypes and application systems that were developed in the automotive industry to support on-board diagnosis, design for diagnosability, and failure modes and effects analysis.

Automated multiple failure FMEA

Abstract Failure mode and effects analysis (FMEA) is typically performed by a team of engineers working together. In general, they will only consider single point failures in a system. Consideration of all possible combinations of failures is impractical for all but the simplest example systems. Even if the task of producing the FMEA report for the full multiple failure scenario were automated, it would still be impractical for the engineers to read, understand and act on all of the results. This paper shows how approximate failure rates for components can be used to select the most likely combinations of failures for automated investigation using simulation. The important information can be automatically identified from the resulting report, making it practical for engineers to study and act on the results. The strategy described in the paper has been applied to a range of electrical subsystems, and the results have confirmed that the strategy described here works well for realistically complex systems.

A Link-Based Cluster Ensemble Approach for Categorical Data Clustering

Although attempts have been made to solve the problem of clustering categorical data via cluster ensembles, with the results being competitive to conventional algorithms, it is observed that these techniques unfortunately generate a final data partition based on incomplete information. The underlying ensemble-information matrix presents only cluster-data point relations, with many entries being left unknown. The paper presents an analysis that suggests this problem degrades the quality of the clustering result, and it presents a new link-based approach, which improves the conventional matrix by discovering unknown entries through similarity between clusters in an ensemble. In particular, an efficient link-based algorithm is proposed for the underlying similarity assessment. Afterward, to obtain the final clustering result, a graph partitioning technique is applied to a weighted bipartite graph that is formulated from the refined matrix. Experimental results on multiple real data sets suggest that the proposed link-based method almost always outperforms both conventional clustering algorithms for categorical data and well-known cluster ensemble techniques.

Function-directed electrical design analysis

Abstract Functional labels provide a simple, but very reusable way for defining the functionality of a system and for making use of that knowledge. Unlike more complex functional representation schemes, these labels can be efficiently linked to a behavioral simulator to interpret the simulation in a way that is meaningful to the user. They are also simple to specify, and highly reusable with different behavioral implementations of the systems functions. This claim has been substantiated by the development of the FLAME application, a practical automated design analysis tool in regular use at several automotive manufacturers. The combination of functional labels and behavioral simulator can be employed for a variety of tasks—simulation, failure mode and effects analysis (FMEA), sneak circuit analysis, design verification, diagnostic candidate generation—producing results that are very valuable to engineers and presented in terms that are easily understood by them. The utility of functional labels is illustrated in this paper for the domain of car electrical systems, with links to a qualitative circuit simulator. In this domain, functional labels provide a powerful way of interpreting the behavior of the circuit simulator in terms an engineer can understand.

The Flame system: automating electrical failure mode and effects analysis (FMEA)

It is well known that FMEA is both tedious and time consuming-so much so, that an FMEA analysis on the design of a system is often only completed after a first prototype has been constructed. This situation can lead to time, effort and money being wasted. Automating the FMEA process will improve the speed and consistency with which an FMEA analysis can be performed. The Flame system aims to provide engineers with a knowledge based system (KBS) which is capable of performing automated FMEA. At present, we are concentrating our efforts on electrical design FMEA, however mechanical and software FMEA will be the subjects of future study. The input to the Flame system consists of a physical description of a particular circuit and a description of that circuits functionality. The output from Flame will be a complete (or near complete) FMEA form which can be checked, annotated and signed off by an engineer. The Flame system demonstrates that it is indeed possible to provide engineers with a means of performing automated electrical FMEA. The application considered is automobile systems.

FAILURE MODE EFFECTS ANALYSIS: A PRACTICAL APPLICATION OF FUNCTIONAL MODELING

Knowledge of how a device works is important for many tasks. Yet, systems that attempt to base their reasoning on the use of a functional model fail to capture such knowledge or only capture it implicitly. Instead they rely solely on the knowledge of the purpose of the system and provide causal explanations of how this purpose is achieved. This type of model only represents knowledge of what the system is for, not how the system works. However, engineers also rely on knowledge of how a device works to complete tasks successfully. One such task is failure mode effects analysis (FMEA). FMEA involves investigation and assessment of the effects of all possible failure modes on a system. This process is both tedious and time consuming, and it requires detailed expert knowledge of the system under consideration, including information about the structure of the system and its purpose or function. This means that any attempt to automate the whole of the FMEA process must involve both the structural and functional...

Effortless incremental design FMEA

Design FMEA of electrical systems is a costly and labour intensive process. Ideally it would be done when the electrical system is first designed, and repeated whenever any change is made to the design. Because of the cost, this has not been possible in the past. This paper describes how an existing tool for automating electrical design failure mode and effects analysis (FMEA) can be augmented to make incremental design FMEA much less of a burden for the engineer. The tool is able to generate the effects for each failure mode and to assign significance values to the effects. The first time that it is run on a design, the engineer still has quite a lot of work to do, examining the results and deciding what actions need to be taken because of the FMEA. When a change is made to the circuit, the engineer runs the FMEA tool again and receives a new report. Because of the uniformity of the reports provided by the FMEA tool, it has proved possible to write software which sorts out the failure effects which have changed from the previous analysis and only report those results to the engineer. This makes examination of the repercussions of the incremental FMEA much less effort for the engineer, and makes it feasible to perform an incremental FMEA every time the design is amended.

Applications of deep knowledge

Abstract This paper contrasts the shallow representational methods used in many of todays commercial expert systems with methods which reason about the function and structure of the objects under consideration. These methods are used to build ‘deep knowledge’ systems and are still being researched. The paper gives examples of the experimental application of these methods in different domains.

Disclosing false identity through hybrid link analysis

Combating the identity problem is crucial and urgent as false identity has become a common denominator of many serious crimes, including mafia trafficking and terrorism. Without correct identification, it is very difficult for law enforcement authority to intervene, or even trace terrorists’ activities. Amongst several identity attributes, personal names are commonly, and effortlessly, falsified or aliased by most criminals. Typical approaches to detecting the use of false identity rely on the similarity measure of textual and other content-based characteristics, which are usually not applicable in the case of highly deceptive, erroneous and unknown descriptions. This barrier can be overcome through analysis of link information displayed by the individual in communication behaviours, financial interactions and social networks. In particular, this paper presents a novel link-based approach that improves existing techniques by integrating multiple link properties in the process of similarity evaluation. It is utilised in a hybrid model that proficiently combines both text-based and link-based measures of examined names to refine the justification of their similarity. This approach is experimentally evaluated against other link-based and text-based techniques, over a terrorist-related dataset, with further generalization to a similar problem occurring in publication databases. The empirical study demonstrates the great potential of this work towards developing an effective identity verification system.