Göran Falkman

A Bayesian network approach to threat evaluation with application to an air defense scenario

In this paper, a precise description of the threat evaluation process is presented. This is followed by a review describing which parameters that have been suggested for threat evaluation in an air surveillance context throughout the literature, together with an overview of different algorithms for threat evaluation. Grounded in the findings from the literature review, a threat evaluation system have been developed. The system is based on a Bayesian network approach, making it possible to handle imperfect observations. The structure of the Bayesian network is described in detail. Finally, an analysis of the systempsilas performance as applied to a synthetic scenario is presented.

Online Learning and Sequential Anomaly Detection in Trajectories

Detection of anomalous trajectories is an important problem in the surveillance domain. Various algorithms based on learning of normal trajectory patterns have been proposed for this problem. Yet, these algorithms typically suffer from one or more limitations: They are not designed for sequential analysis of incomplete trajectories or online learning based on an incrementally updated training set. Moreover, they typically involve tuning of many parameters, including ad-hoc anomaly thresholds, and may therefore suffer from overfitting and poorly-calibrated alarm rates. In this article, we propose and investigate the Sequential Hausdorff Nearest-Neighbor Conformal Anomaly Detector (SHNN-CAD) for online learning and sequential anomaly detection in trajectories. This is a parameter-light algorithm that offers a well-founded approach to the calibration of the anomaly threshold. The discords algorithm, originally proposed by Keogh et al. , is another parameter-light anomaly detection algorithm that has previously been shown to have good classification performance on a wide range of time-series datasets, including trajectory data. We implement and investigate the performance of SHNN-CAD and the discords algorithm on four different labeled trajectory datasets. The results show that SHNN-CAD achieves competitive classification performance with minimum parameter tuning during unsupervised online learning and sequential anomaly detection in trajectories.

Detection of vessel anomalies - a Bayesian network approach

In this paper we describe a data mining approach for detection of anomalous vessel behaviour. The suggested approach is based on Bayesian networks which have two important advantages compared to opaque machine learning techniques such as neural networks: (1) possibility to easily include expert knowledge into the model, and (2) possibility for humans to understand and interpret the learned model. Our approach is implemented and tested on synthetic data, where initial results show that it can be used for detection of single-object anomalies such as speeding.

SOMWeb: A Semantic Web-Based System for Supporting Collaboration of Distributed Medical Communities of Practice

Background Information technology (IT) support for remote collaboration of geographically distributed communities of practice (CoP) in health care must deal with a number of sociotechnical aspects of communication within the community. In the mid-1990s, participants of the Swedish Oral Medicine Network (SOMNet) began discussing patient cases in telephone conferences. The cases were distributed prior to the conferences using PowerPoint and email. For the technical support of online CoP, Semantic Web technologies can potentially fulfill needs of knowledge reuse, data exchange, and reasoning based on ontologies. However, more research is needed on the use of Semantic Web technologies in practice. Objectives The objectives of this research were to (1) study the communication of distributed health care professionals in oral medicine; (2) apply Semantic Web technologies to describe community data and oral medicine knowledge; (3) develop an online CoP, Swedish Oral Medicine Web (SOMWeb), centered on user-contributed case descriptions and meetings; and (4) evaluate SOMWeb and study how work practices change with IT support. Methods Based on Java, and using the Web Ontology Language and Resource Description Framework for handling community data and oral medicine knowledge, SOMWeb was developed using a user-centered and iterative approach. For studying the work practices and evaluating the system, a mixed-method approach of interviews, observations, and a questionnaire was used. Results By May 2008, there were 90 registered users of SOMWeb, 93 cases had been added, and 18 meetings had utilized the system. The introduction of SOMWeb has improved the structure of meetings and their discussions, and a tenfold increase in the number of participants has been observed. Users submit cases to seek advice on diagnosis or treatment, to show an unusual case, or to create discussion. Identified barriers to submitting cases are lack of time, concern about whether the case is interesting enough, and showing gaps in one’s own knowledge. Three levels of member participation are discernable: a core group that contributes most cases and most meeting feedback; an active group that participates often but only sometimes contribute cases and feedback; and a large peripheral group that seldom or never contribute cases or feedback. Conclusions SOMWeb is beneficial for individual clinicians as well as for the SOMNet community. The system provides an opportunity for its members to share both high quality clinical practice knowledge and external evidence related to complex oral medicine cases. The foundation in Semantic Web technologies enables formalization and structuring of case data that can be used for further reasoning and research. Main success factors are the long history of collaboration between different disciplines, the user-centered development approach, the existence of a “champion” within the field, and nontechnical community aspects already being in place.

Information visualisation in clinical Odontology: multidimensional analysis and interactive data exploration

In 1995, the MedView project, based on a co-operation between computing science and clinical medicine was initiated. The overall goal of the project was to develop models, methods and tools to support clinicians in their daily diagnostic work. As part of MedView, two information visualisation tools were developed and tested as solutions to the problem of visualising clinical experience derived from large amounts of clinical data. The first tool (The Cube) was based on the idea of dynamic three-dimensional (3D) parallel diagrams, an idea similar to the notion of 3D parallel co-ordinates. The Cube was developed to enhance the clinicians ability to intelligibly analyse existing patient material and to allow for pattern recognition and statistical analysis. The second tool (SimVis) was based on a similarity assessment-based interaction model for exploring data, and was designed to help clinicians to classify and cluster clinical examination data. User interaction was supported by 3D visualisation of clusters and similarity measures. Both tools were tested on a knowledge base containing about 1500 examinations obtained from different clinics. Clinical practice indicated that the basic ideas are conceptually appealing to the involved clinicians as the tools can be used for generating and testing of hypotheses.

Conformal prediction for distribution-independent anomaly detection in streaming vessel data

This paper presents a novel application of the theory of conformal prediction for distribution-independent on-line learning and anomaly detection. We exploit the fact that conformal predictors give valid prediction sets at specified confidence levels under the relatively weak assumption that the (normal) training data together with (normal) observations to be predicted have been generated from the same distribution. If the actual observation is not included in the possibly empty prediction set, it is classified as anomalous at the corresponding significance level. Interpreting the significance level as an upper bound of the probability that a normal observation is mistakenly classified as anomalous, we can conveniently adjust the sensitivity to anomalies while controlling the rate of false alarms without having to find any application specific thresholds. The proposed method has been evaluated in the domain of sea surveillance using recorded data assumed to be normal. The validity of the prediction sets is justified by the empirical error rate which is just below the significance level. In addition, experiments with simulated anomalous data indicate that anomaly detection sensitivity is superior to that of two previously proposed methods.

Visual Analytics for the Detection of Anomalous Maritime Behavior

The surveillance of large sea areas often generates huge amounts of multidimensional data. Exploring, analyzing and finding anomalous behavior within this data is a complex task. Confident decisions upon the abnormality of a particular vessel behavior require a certain level of situation awareness that may be difficult to achieve when the operator is overloaded by the available information. Based on a visual analytics process model, we present a novel system that supports the acquisition of situation awareness and the involvement of the user in the anomaly detection process using two layers of interactive visualizations. The system uses an interactive data mining module that supports the insertion of the users knowledge and experience in the creation, validation and continuous update of the normal model of the environment.

A Comparison between Two Approaches to Threat Evaluation in an Air Defense Scenario

Threat evaluation is a high-level information fusion problem of high importance within the military domain. This task is the foundation for weapons allocation, where assignment of blue force (own) weapon systems to red force (enemy) targets is performed. In this paper, we compare two fundamentally different approaches to threat evaluation: Bayesian networks and fuzzy inference rules. We conclude that there are pros and cons with both types of approaches, and that a hybrid of the two approaches seems both promising and viable for future research.

Enabling an online community for sharing oral medicine cases using semantic web technologies

This paper describes how Semantic Web technologies have been used in an online community for knowledge sharing between clinicians in oral medicine in Sweden. The main purpose of this community is to serve as repository of interesting and difficult cases, and as a support for monthly teleconferences. All information regarding users, meetings, news, and cases is stored in RDF. The community was built using the Struts framework and Jena was used for interacting with RDF.

Implementation and integration of a Bayesian Network for prediction of tactical intention into a ground target simulator

Prediction of the enemys intention is a main issue of threat analysis, and, hence, will be an important part of the C2-systems of tomorrow. A technique that can be useful for this kind of predictions is Bayesian networks (BNs). We have developed a BN for prediction of the enemys tactical intention, and the implemented BN has been integrated into a ground target simulation framework. The general problem of how to find appropriate prior distributions for BNs has been addressed by developing a tool for data collection, which may make it easier to come up with appropriate prior distributions, by learning conditional probability tables from collected cases, i.e. parameter learning