Advances in Case-Based Reasoning

This paper describes two different approaches for incorporating background knowledge into nearest-neighbor text classification. Our first approach uses background text to assess the similarity between training and test documents rather than assessing their similarity directly. The second method redescribes examples using Latent Semantic Indexing on the background knowledge, assessing document similarities in this redescribed space. Our experimental results show that both approaches can improve the performance of nearest-neighbor text classification. These methods are especially useful when labeling text is a labor-intensive job and when there is a large amount of information available about a specific problem on the World Wide Web.

Over the previous two years I have collected case-studies of successfully fielded commercial knowledge management systems that use case-based reasoning (CBR). These case-studies have been collated into a book to be published by Morgan Kaufmann in November 20021. This paper summarises the findings of the book, showing that CBR is ideally suited to the creation of knowledge management systems. This is because of the close match between the activities of the CBR-cycle and the process requirements of a knowledge management system. The nature of knowledge within an organisation is briefly discussed and the paper illustrates the dynamic relationship between data, information and knowledge, showing how CBR can be used to help manage the acquisition and reuse of knowledge.

In this paper we present an algorithm called DerUCP, which can be regarded as a general model for plan adaptation using Derivational Analogy. Using DerUCP, we show that previous results on the complexity of plan adaptation do not apply to Derivational Analogy. We also show that Derivational Analogy can potentially produce exponential reductions in the size of the search space generated by a planning system.

We present adapted inductive methods for learning similarities, parameter weights and diagnostic profiles for case-based reasoning. All of these methods can be refined incrementally by applying different types of background knowledge. Diagnostic profiles are used for extending the conventional CBR to solve cases with multiple faults. The context of our work is to supplement a medical documentation and consultation system by CBR techniques, and we present an evaluation with a real-world case base.

We describe Order-Based Retrieval, which is an approach to case retrieval based on the application of partial orders to the case base. We argue that it is well-suited to product recommender applications because, as well as retrieving products that best match customer-specified ‘ideal’ attribute-values, it also: allows the customer to specify soft constraints; gives a natural semantics and implementation to tweaks; and delivers an inherently diverse result set.

The amount and availability of high-quality geo-spatial image data, such as digital satellite and aerial photographs, is increasing dramatically. Task-based management of such visual information and associated knowledge is a central concern for organisations that rely on digital imagery. We are developing geo-spatial knowledge management techniques that employ case-based reasoning as the core methodology. In order to provide effective retrieval of task-based experiences that center around geo-spatial imagery, we need to forward novel similarity metrics for directly comparing the image components of experience cases. Based on work in geo-spatial image database retrieval, we are building an effective similarity metric for geo-spatial imagery that makes comparisons based on derived image features, their shapes, and the spatial relations between them. This paper gives an overview of the geo-spatial knowledge management context, describes our image similarity metric, and provides an initial evaluation of the work.

CBROnto is an ontology that incorporates common Case-Based Reasoning (CBR) terminology and serves as a domain-independent framework to design CBR applications. It is the core of COLIBRI, an environment to assist during the design of knowledge intensive CBR systems that combine cases with various knowledge types and reasoning methods. CBROnto captures knowledge about CBR tasks and methods, and aims to unify case specific and general domain knowledge representational needs. CBROnto specifies a modelling framework to describe reusable CBR Problem Solving Methods based on the CBR tasks they solve. This paper describes CBROnto’s main ideas and exemplifies them with an application to generate Spanish poetry versions of texts provided by the user.

A model for adaptation in case-based reasoning (cbr) is presented. Similarity assessment is based on the computation and the iterated estimation of structural relationships among representations, and adaptation is given as a special case of the general process.Compared to traditional approaches to adaptation within cbr, the presented model has the advantage of using a uniform declarative model for both case representation, similarity assessment and adaptation. As a consequence, adaptation knowledge can be made directly available during similarity assessment and for explanation purposes. The use of a uniform model also provides the possibility of a cbr approach to adaptation.The model is compared with other approaches to adaptation within cbr.

We present an information visualisation tool, The Cube, as a solution to the problem of visualising cases derived from large amounts of clinical data. The Cube is based on the idea of dynamic 3D parallel diagrams, an idea similar to the notion of 3D parallel coordinate plots. The Cube was developed to provide interactive visualisation of the case base in terms of relationships between and within cases, in order to enhance the clinician’s ability to intelligibly analyse existing patient material and to allow for pattern recognition and statistical analysis.The design and use of The Cube are presented and discussed. We show how the declarative model used and the tight coupling between different visualisation tools directly led to a similarity assessment-based solution to the problem of finding a proper arrangement of dimensions in 3D parallel coordinate displays. The declarative, user-centered nature of The Cube makes it suitable for interactive case-based reasoning (cbr) and opens up for the possibility of case-based visualisation for cbr.

Software systems are becoming increasingly complex, demanding for more computational resources and better software development methodologies. The software engineer and the CASE tool must work like a team. For this to happen, the CASE tool must be able to understand the user, and to provide new functionalities, such as flexible retrieval of old designs. We think that Case-Based Reasoning can provide a reasoning framework capable of meeting these demands. One important task that a CASE tool based on Case-Based Reasoning can perform adequately, is the retrieval of relevant designs. These designs can be stored in a case library, central to the software development company, thus enabling knowledge sharing through out the company. In this paper we present an approach to case-based retrieval of software designs, and experimental results achieved with this approach.

Conversational case-based reasoning (CCBR) systems engage their users in a series of questions and answers and present them with cases that are most applicable to their decision problem. In previous research, we introduced the Taxonomic CCBR methodology, an extension of standard CCBR that improved performance by organizing features related by abstraction into taxonomies. We recently extended this methodology to include causal relations between taxonomies and claimed that it could yield additional performance gains. In this paper, we formalize the causal extension of Taxonomic CCBR, called Causal CCBR, and empirically assess its benefits using a new methodology for evaluating CCBR performance. Evaluation of Taxonomic and Causal CCBR systems in troubleshooting and customer support domains demonstrates that they significantly outperform the standard CCBR approach. In addition, Causal CCBR outperforms Taxonomic CCBR to the extent causal relations are incorporated in the case bases.

Bayesian Case Reconstruction (BCR) is a case-based technique that broadens the coverage of a case library by sampling and recombining pieces of existing cases to construct a large set of “plausible” cases. It employs a Bayesian Belief Network to evaluate whether implicit dependencies within the original cases have been maintained. The belief network is constructed from the expert’s limited understanding of the domain theory combined with the data available in the case library. The cases are the primary reasoning vehicle. The belief network leverages the available domain model to help evaluate whether the “plausible” cases have maintained the necessary internal context. BCR is applied to the design of screening experiments for Macromolecular Crystallization in the Probabilistic Screen Design program. We describe BCR and provide an empirical comparison of the Probabilistic Screen Design program against the current practice in Macromolecular Crystallization.

The ultimate goal of CBR applications is to satisfy customers using this technology in their daily business. As one of the crucial issues in CBR for practical applications, maintenance is important to cope with demands changing over time. Review and restore are the two steps in CBR that deal with tasks of maintenance. In order to perform these tasks, we suggested case and case base properties, quality criteria, and restore operators in earlier publications. In this paper, we specify concrete performance measures that correspond to general customer requirements, and analyze the relations between these performance criteria, case properties, and restore operators. We present initial results on theoretical analyzes on these relations, and report on examples of experimental studies that indicate that the suggested case properties and the respective restore operators help to identify maintenance strategies in order to optimize performance of CBR systems over time.

Cases are descriptions of situations limited in time and space. The research reported here introduces a method for representation and reasoning with time-dependent situations, or temporal cases, within a knowledge-intensive CBR framework. Most current CBR methods deal with snapshot cases, descriptions of a world state at a single time stamp. In many time-dependent situations, value sets at particular time points are less important than the value changes over some interval of time. Our focus is on prediction problems for avoiding faulty situations. Based on a well-established theory of temporal intervals, we have developed a method for representing temporal cases inside the knowledge-intensive CBR system Creek. The paper presents the theoretical foundation of the method, the representation formalism and basic reasoning algorithms, and an example applied to the prediction of unwanted events in oil well drilling.

This paper presents a novel case-based plan recognition system that interprets observations of plan behavior using a case library of past observations. The system is novel in that it represents a plan as a sequence of action-state pairs rather than a sequence of actions preceded by some initial state and followed by some final goal state. The system utilizes a unique abstraction scheme to represent indices into the case base. The paper examines and evaluates three different methods for prediction. The first method is prediction without adaptation; the second is predication with adaptation, and the third is prediction with heuristics. We show that the first method is better than a baseline random prediction, that the second method is an improvement over the first, and that the second and the third methods combined are the best overall strategy.

Case-based reasoning (CBR) systems solve new problems by retrieving stored prior cases, and adapting their solutions to fit new circumstances. Traditionally, CBR systems draw their cases from a single local case-base tailored to their task. However, when a system’s own set of cases is limited, it may be beneficial to supplement the local case-base with cases drawn from external case-bases for related tasks. Effective use of external case-bases requires strategies for multi-case-base reasoning (MCBR): (1) for deciding when to dispatch problems to an external case-base, and (2) for performing cross-case-base adaptation to compensate for differences in the tasks and environments that each case-base reflects. This paper presents methods for automatically tuning MCBR systems by selecting effective dispatching criteria and cross-case-base adaptation strategies. The methods require no advance knowledge of the task and domain: they perform tests on an initial set of problems and use the results to select strategies reflecting the characteristics of the local and external case-bases. We present experimental illustrations of the performance of the tuning methods for a numerical prediction task, and demonstrate that a small sample set can be sufficient to make high-quality choices of dispatching and cross-case-base adaptation strategies.

There is growing awareness of the need for recommender systems to offer a more diverse choice of alternatives than is possible by simply retrieving the cases that are most similar to a target query. Recent research has shown that major gains in recommendation diversity can often be achieved at the expense of relatively small reductions in similarity. However, there are many domains in which it may not be acceptable to sacrifice similarity in the interest of diversity. To address this problem, we examine the conditions in which similarity can be increased without loss of diversity and present a new approach to retrieval which is designed to deliver such similarity-preserving increases in diversity when possible. We also present a more widely applicable approach to increasing diversity in which the requirement that similarity is fully preserved is relaxed to allow some loss of similarity, provided it is strictly controlled.

Recommendations by salespeople are always based on knowledge about the products and expertise about your tastes, preferences, interests and behavior in the shop. In an attempt to model the behavior of salespeople, AI research has been focussed on the so called recommender agents. Such agents draw on previous results from machine learning and other advances in AI technology to develop user models and to anticipate and predict user preferences. In this paper we introduce a new approach to recommendation, based on Case-Based Reasoning (CBR). CBR is a paradigm for learning and reasoning through experience, as salesmen do. We present a user model based on cases in which we try to capture both explicit interests (the user is asked for information) and implicit interests (captured from user interaction) of a user on a given item. Retrieval is based on a similarity function that is constantly tuned according to the user model. Moreover, in order to cope with the utility problem that current CBR system suffer from, our approach includes a forgetting mechanism (the drift attribute) that can be extended to other applications beyond e-commerce.

Generalized cases are cases that cover a subspace rather than a point in the problem-solution space. Generalized cases can be represented by a set of constraints over the case attributes. For such representations, the similarity assessment between a point query and generalized cases is a difficult problem that is addressed in this paper. The task is to find the distance (or the related similarity) between the point query and the closest point of the area covered by the generalized cases, with respect to some given similarity measure. We formulate this problem as a mathematical optimization problem and we propose a new cutting plane method which enables us to rank generalized cases according to their distance to the query.

In this paper, we propose an approach to acquire cases in the context of project planning, without any extra effort from the end user. Under our definition, a case has a one to one correspondence with the standard elements of a project plan. We exploit this correspondence to capture cases automatically from project planning episodes. We provide an algorithm for extracting cases from project plans. We implemented this algorithm on top of a commercial project-planning tool and perform experiments evaluating our approach.

Data Mining, or Knowledge Discovery as it is also known, is becoming increasingly useful in a wide variety of applications. In the following paper, we look at its use in combating some of the traditional issues faced with recommender systems. We discuss our ongoing work which aims to enhance the performance of PTV, an applied recommender system working in the TV listings domain. This system currently combines the results of separate user-based collaborative and case-based components to recommend programs to users. Our extension to this idea operates on the theory of developing a case-based view of the collaborative component itself. By using data mining techniques to extract relationships between programme items, we can address the sparsity/maintenance problem. We also adopt a unique approach to recommendation ranking which combines user similarities and item similarities to provide more effective recommendation orderings. Experimental results corroborate our ideas, demonstrating the effectiveness of data mining in improving recommender systems by providing similarity knowledge to address sparsity, both at user-based recommendation level and recommendation ranking level.

In this paper, we present three techniques for knowledge discovery in case-based reasoning. The first two techniques D-HS and D-HS+SR are concerned with the discovery of similarity knowledge and operate on an uncompacted case-base while the third technique D-HS+PSR is concerned with the discovery of both similarity and case knowledge and operates on a compacted case-base. All three techniques provide a very efficient and competent means of similarity determination in CBR, which are empirically shown to be up to 25 times faster than k-NN without any loss in competency. D-HS+PSR proposes a novel approach to automatically engineering compact case-bases with a minimal overhead to the system, compared to other approaches such as case deletion/addition. Additionally as the approach provides a means for automatically reducing the number of cases required in the case-base without any loss in problem solving competency it has the greatest implication of the three techniques for reducing the effects of the utility problem in CBR.

Constructive adaptation is a search-based technique for generative reuse in CBR systems for configuration tasks. We discuss the relation of constructive adaptation (CA) with other reuse approaches and we define CA as a search process in the space of solutions where cases are used in two main phases: hypotheses generation and hypotheses ordering. Later, three different CBR systems using CA for reuse are analyzed: configuring gas treatment plants, generating expressive musical phrases, and configuring component-based software applications. After the three analyses, constructive adaptation is discussed in detail and some conclusions are drawn to close the paper.

Providing a flexible and efficient way of consulting a catalog in e-commerce applications is of primary importance in order to guarantee the customer with a set of products actually related to his/her interests. Most electronic catalogs exploit standard database techniques both for storage and retrieval of product information. However, a naive application of ordinary databases may produce unsatisfactory results, since standard query tools are not able to retrieve information (i.e. products) that only partially match the user/customer specification. The use of CBR may alleviate some of the above problems, because of the ability of a CBR system of retrieving products having characteristics similar to the ones specified by the user. While the majority of the approaches is based on k-NN retrieval techniques, in the present paper we propose fuzzy-based retrieval as a natural way for implementing flexible search on electronic catalogs. Since the exploitation of standard DBMS technology is of paramount importance for deploying any E-commerce application, we also propose to use a fuzzy extension to SQL for retrieving a set of products that the customer specifies using precise as well as vague or imprecise features. The proposed implementation is based on a client/server web-based architecture working on top of a relational standard DBMS. A specific example concerning an on-line wine shop is used to demonstrate the capabilities of the approach.

In this paper, we present an approach integrating neurule-based and case-based reasoning. Neurules are a kind of hybrid rules that combine a symbolic (production rules) and a connectionist representation (adaline unit). Each neurule is represented as an adaline unit. One way that the neurules can be produced is from symbolic rules by merging the symbolic rules having the same conclusion. In this way, the number of rules in the rule base is decreased. If the symbolic rules, acting as source knowledge of the neurules, do not cover the full complexities of the domain, accuracy of the produced neurules is affected as well. To improve accuracy, neurules can be integrated with cases representing their exceptions. The integration approach enhances a previous method integrating symbolic rules with cases. The use of neurules instead of symbolic rules improves the efficiency of the inference mechanism and allows for drawing conclusions even if some of the inputs are unknown.

In this paper we propose to represent a case using an object oriented model that enables the description of imprecise knowledge using possibility distributions. The proposed search process is based on this modeling and a fuzzy similarity measure is defined. The adaptation process is achieved with propagation of domain constraints in a neighborhood of the retrieved case. We propose a method to define this neighborhood. We illustrate our proposition by an example in the field of machining operation configuration.

Early work on case based reasoning reported in the literature shows the importance of case base maintenance for successful practical systems. Different criteria to the maintenance task have been used for more than half a century. In this paper we present different sort out techniques for case base maintenance. All the sort out techniques proposed are based on the same principle: a Rough Sets competence model. First of all, we present sort out reduction techniques based on deletion of cases. Next, we present sort out techniques that build new reduced competent case memories based on the original ones. The main purpose of these methods is to maintain the competence and reduce, as much as possible, its size. Experiments using different domains, most of them from the UCI repository, show that the reduction techniques maintain the competence obtained by the original case memory. The results are analysed with those obtained using well-known reduction techniques.

Recent research activities in the field of attribute selection for carrying on dialogs with on-line customers have focused on entropy-based approaches that make use of information gain measures. These measures consider the distribution of attribute values in the case base and are focused on their ability to reduce dialog length. The implicit knowledge contained in the similarity measures is neglected. In previous work, we proposed the similarity-influenced selection method simVar, which selects the attributes that induce the maximum change in similarity distribution amongst the candidate cases, thereby partitioning the case base into similar and dissimilar cases. In this paper we present an evaluation of the selection methods using three domains with distinct characteristics. The comparison of the selection methods is based on the quality of the dialogs generated. Statistical analysis was used to support the evaluation results.

Collaborative filtering is an often successful method for personalized item selection in Recommender systems. However, in domains where items are frequently added, collaborative filtering encounters the latency problem. Characterized by the system’s inability to select recently added items, the latency problem appears because new items in a collaborative filtering system must be reviewed before they can be recommended. Content-based filtering may help to counteract this problem, but runs the risk of only recommending items almost identical to the ones the user has appreciated before. In this paper, a combination of category-based filtering and user stereotype cases is proposed as a novel approach to reduce the latency problem. Category-based filtering puts emphasis on categories as meta-data to enable quicker personalization. User stereotype cases, identified by clustering similar users, are utilized to decrease response times and improve the accuracy of recommendations when user information is incomplete.

Defining similarity measures is a crucial task when developing CBR applications. Particularly, when employing utility-based similarity measures rather than pure distance-based measures one is confronted with a difficult knowledge engineering task. In this paper we point out some problems of the state-of-the-art procedure to defining similarity measures. To overcome these problems we propose an alternative strategy to acquire the necessary domain knowledge based on a Machine Learning approach. To show the feasibility of this strategy several application scenarios are discussed and some results of an experimental evaluation for one of these scenarios are presented.

Design is a complex open-ended task and it is unreasonable to expect a case-base to contain representatives of all possible designs. Therefore, adaptation is a desirable capability for case-based design systems, but acquiring adaptation knowledge can involve significant effort. In this paper adaptation knowledge is induced separately for different criteria associated with the retrieved solution, using knowledge sources implicit in the case-base. This provides a committee of learners and their combined advice is better able to satisfy design constraints and compatibility requirements compared to a single learner. The main emphasis of the paper is to evaluate the impact of specific-to-general and general-to-specific learning on adaptation knowledge acquired by committee members. For this purpose we conduct experiments on a real tablet formulation problem which is tackled as a decomposable design task. Evaluation results suggest that adaptation achieves significant gains compared to a retrieve-only CBR system, but shows that both learning biases can be beneficial for different decomposed sub-tasks.

Ensemble research has shown that the aggregated output of an ensemble of predictors can be more accurate than a single predictor. This is true also for lazy learning systems like Case-Based Reasoning (CBR) and k-Nearest-Neighbour. Aggregation is normally achieved by voting in classification tasks and by averaging in regression tasks. For CBR, this increased accuracy comes at the cost of interpretability however. If we consider the use of retrieved cases for explanation to be one of the advantages of CBR then this is lost in an ensemble. This is because a large number of cases will have been retrieved by the ensemble members. In this paper we present a new technique for aggregation that obtains excellent results and identifies a small number of cases for use in explanation. This new approach might be viewed as a transformation process whereby cases are transformed from their feature based representation to a representation based on the predictions of ensemble members. This new representation produces very accurate predictions and allows a small number of similar neighbours to be identified.

Increasing dversity for case-based reasoning (CBR) is an issue that has recently drawn the attention of researchers in the CBR field. Several diversification techniques have been proposed and discussed in the literature. However, whether and to what extent those techniques can bring about benefits to end-users remains in question. In this paper, we report an experiment in applying a diversification technique to a case-based diagnosis tool in a product maintenance domain. The results of this offer some evidence in support of diversification techniques.

We introduce an application combining CBR and collaborative filtering techniques in the music domain. We describe a scenario in which a new kind of recommendation is required, which is capable of summarizing many recommendations in one suggestion. Our claim is that recommending one set of goods is different from recommending a single good many times. The paper illustrates how a case-based reasoning approach can provide an effective solution to this problem reducing the drawbacks related to the user profiles. CoCoA, a compilation compiler advisor, will be described as a running example of a collaborative case-based recommendation system.

This paper illustrates how the Case Based Reasoning (CBR) paradigm has been applied to the definition of the requirements and the specifications of a Knowledge Management technology supporting the handling of tuning and anomalies in production processes. In particular, the production of truck tyres will be illustrated as the application domain of the proposed approach. This domain is a paradigmatic example where the valorization of the experiential knowledge is a company requirement, so the representation of knowledge concerning production process issues is mandatory. It requires extending CBR techniques to an emerging area of application involving knowledge about the interplay of static objects (i.e. products) with dynamic objects (i.e. processes). The content of production cases, their organization into a Case Base and a first proposal for the Case Base retrieval of similar cases will be presented.

This paper describes an innovative application of Case-Based Reasoning methodologies for the dynamic management of wireless telecommunications systems. In spite of the very dynamic nature of mobile communications, wireless networks only possess limited adaptive management capabilities, which are unable to adequately follow traffic fluctuations through flexible and real-time resource assignment reconfigurations. The study described in this paper is an attempt to improve over these limitations, by allowing wireless networks to alleviate the effects of traffic overloads through an automated reasoning about its performance levels and an on-the-fly reconfiguration of resources assignment. The proposed Case-Based Reasoning approach provides a suitable framework to define a simple and scalable solution that easily incorporates the preferences of the network operators.

This paper describes a case-based approach to user profiling in a Personal Travel Assistant (based on the 1998 FIPA Travel Scenario). The approach is novel in that the user profile is made up of a set of cases capturing previous interactions rather than as a single composite case. This has the advantage that the profile is always up-to-date and also allows for the borrowing of cases from similar users when coverage is poor. Profile data is retrieved from a database in an XML format and loaded into a case-retrieval net in memory. This case-retrieval net is then used to support the two key tasks of requirements elaboration and ranking offers.

A hybrid neuro-symbolic problem solving model is presented in which the aim is to forecast parameters of a complex and dynamic environment in an unsupervised way. In situations in which the rules that determine a system are unknown, the prediction of the parameter values that determine the characteristic behaviour of the system can be a problematic task. The proposed system employs a case-based reasoning model that incorporates a growing cell structures network, a radial basis function network and a set of Sugeno fuzzy models to provide an accurate prediction. Each of these techniques is used in a different stage of the reasoning cycle of the case-based reasoning system to retrieve, to adapt and to review the proposed solution to the problem. This system has been used to predict the red tides that appear in the coastal waters of the north west of the Iberian Peninsula. The results obtained from those experiments are presented.

Software design patterns are used in software engineering as a way to improve and maintain software systems. Patterns are abstract solutions to problem categories, and they describe why, how, and when can a pattern be applied. Their description is based on natural language, which makes the automation of design patterns a difficult task. In this paper we present an approach for automation of design pattern application. We focus on the selection of what pattern to apply, and where to apply it. We follow a Case-Based Reasoning approach, providing a complete framework for pattern application. In our approach cases describe situations for application of patterns.

In this paper, a new approach is proposed for transformational adaptation based on detecting a solution’s incompatibility regarding the new problem situation and trying to overcome the incompatibility in an iterative manner. By incompatibility, we mean a state for which the required objectives are not satisfied due to any change in the status of the constraints. Based upon this approach, we have proposed a framework for redesigning an existing system under new constraints. To show the capability of this framework, a software prototype was developed that it is capable of redesigning an existing digital circuit under presence of new constraints, e.g. type of gates, power dissipation, fan in/out, gate prices and so on.

This paper describes a CBR application in manufacturing industry, a domain where CBR has by and large proved its applicability and success. The paper details a thorough understanding of the field of fused cast manufacturing basically seen from the perspective of glass furnace, where quality of glass produced is straightaway related to the refractory blocks used in furnace linings. The applicability of CBR paradigm is revisited in the present context. The CBR process needed is conceptualized and designed. The paper describes the evolution of the system beginning with tackling hurdles of knowledge acquisition, a number of pitfalls in the prototype phase, to final implementation of InfoFrax, the CBR system specially devised for the project. It gives an overall description of the architecture and usage. The paper also reports the immediate response to the software in form of direct user feedback, expectations from the existing system, and some future work already underway in the project.

Recommender systems combine user profiling and filtering techniques to provide more pro-active and personal information retrieval systems, and have been gaining in popularity as a way of overcoming the ubiquitous information overload problem. Many recommender systems operate as interactive systems that seek feedback from the end-user as part of the recommendation process to revise the user’s query. In this paper we examine different forms of feedback that have been used in the past and focus on a low-cost preference-based feedback model, which to date has been very much under utilised. In particular we describe and evaluate a novel comparison-based recommendation framework which is designed to utilise preference-based feedback. Specifically, we present results that highlight the benefits of a number of new query revision strategies and evidence to suggest that the popular more-like-this strategy may be flawed.

Estuaries are complex natural water systems. Their behaviour depend on many factors, which are possible to analyse only by adopting different study approaches. The physical processes within estuaries, such as floods and pollutant dispersion, are generally investigated through computer modelling. In this paper the application of case-based reasoning technology to support the design of estuarine models is described. The system aims to provide a nonexpert user in modelling with the necessary guidance for selecting a model that matches his goal and the nature of the problem to be solved. The system is based on three components: a case-based reasoning scheme, a genetic algorithm and a library of numerical estuarine models. An example based on the Upper Milford Haven estuary (UK) is used to demonstrate the efficacy of the system’s structure for supporting estuarine model design.

The development of an automatic image classification system is a hard problem since such a system must imitate the visual strategy of a human expert when interpreting the particular image. Usually it is not easy to make this strategy explicit. Rather than describing the visual strategy and the image features human are able to judge the similarity between the objects. This judgement can be the basis for a guideline of the development process. This guideline can help the developer to understand what kind of case description/features are necessary for a sufficient system performance and can give an idea what system performance can be achieved. In the paper we describe a novel strategy which can support a developer in building image classification systems. The development process as well as the elicitation of the case description is similarity-guided. Based on the similarity between the objects the system developer can provide new image features and improve the system performance until a system performance is reached that fits to the experts understanding about the relationship among the different objects.

This paper presents a web based recommender system aimed at supporting a user in information filtering and product bundling. The system enables the selection of travel locations, activities and attractions, and supports the bundling of a personalized travel plan. A travel plan is composed in a mixed initiative way: the user poses queries and the recommender exploits an innovative technology that helps the user, when needed, to reformulate the query. Travel plans are stored in a memory of cases, which is exploited for ranking travel items extracted from catalogues. A new ’collaborative’ approach is introduced, where user past behavior similarity is replaced with session (travel plan) similarity.

The growing complexity of today’s electronic designs requires reusing existing design components, called Intellectual Properties (IPs). Experience management approaches can be used to support design reuse, particularly the process of selecting reusable IPs. For the IP selection, quality criteria concerning the IP code and the documentation must be considered in addition to functional requirements of the IP. We analyse IP quality criteria in detail and show different concepts for their integration into the retrieval process.

Land development control is the process of controlling land development to meet the needs of the society. In this paper, we attempt to advance the use of case-based reasoning in building a case-based decision support system for land development control. Land development control is a complex domain. We first discuss how to deal with the data and knowledge involved in land development control in order to represent the knowledge and define the case in the system. We then propose to use land use function pattern, which is built on geospatial relations and land use functions of the proposed development site with its surrounding environment, to simplify case input, representation and retrieval of the system. Different land use types have different land use function patterns. Land use function pattern can be extracted from the domain knowledge derived from town planning legislation, regulations, and guidelines. Our work mainly aims to support the work of planning officers in seeking similar precedent cases in preparing recommendations to the Town Planning Board. From the preliminary results of the experimental system, we find that cases can be more easily input and represented and similar case(s) can be more efficiently and effectively retrieved using land use function pattern.