Wireless sensor networks are often distributed, diverse, and large making their monitoring hard. One way to tackle it is to focus on part of the system by creating logical sub-views which can be seen as proxies of the overall system operations. In this manuscript, logical sub-views consist of traffic aggregators and their topology which are monitored for anomaly. The aggregators are selected based on diversity and importance in the system and they are modelled as graphs to capture aggregation topology and data distributions. The aggregators' selection criteria, the method for comparison of partially overlapping sub-views, normal aggregation profiles acquisition, and measures of anomaly are proposed. A simulated wireless sensor network is used to acquire data at the edge and apply the method to demonstrate that focusing on system sub-views and comparing aggregation profiles facilitates anomaly detection also caused elsewhere in the system and the impact the anomaly has on aggregators.

BACKGROUND: Identifying splice site regions is an important step in the genomic DNA sequencing pipelines of biomedical and pharmaceutical research. Within this research purview, efficient and accurate splice site detection is highly desirable, and a variety of computational models have been developed toward this end. Neural network architectures have recently been shown to outperform classical machine learning approaches for the task of splice site prediction. Despite these advances, there is still considerable potential for improvement, especially regarding model prediction accuracy, and error rate. RESULTS: Given these deficits, we propose EnsembleSplice, an ensemble learning architecture made up of four (4) distinct convolutional neural networks (CNN) model architecture combination that outperform existing splice site detection methods in the experimental evaluation metrics considered including the accuracies and error rates. We trained and tested a variety of ensembles made up of CNNs and DNNs using the five-fold cross-validation method to identify the model that performed the best across the evaluation and diversity metrics. As a result, we developed our diverse and highly effective splice site (SS) detection model, which we evaluated using two (2) genomic Homo sapiens datasets and the Arabidopsis thaliana dataset. The results showed that for of the Homo sapiens EnsembleSplice achieved accuracies of 94.16% for one of the acceptor splice sites and 95.97% for donor splice sites, with an error rate for the same Homo sapiens dataset, 4.03% for the donor splice sites and 5.84% for the acceptor splice sites datasets. CONCLUSIONS: Our five-fold cross validation ensured the prediction accuracy of our models are consistent. For reproducibility, all the datasets used, models generated, and results in our work are publicly available in our GitHub repository here: https://github.com/OluwadareLab/EnsembleSplice.

Wireless sensor networks are often distributed which makes detection of cyber-attacks or misconfiguration hard. Topology and data patterns change may result from attacks leading to the compromise of data and service availability or indicate operational problems. Graphs are often used to model topology and data paths to describe and compare state of a system. For anomaly detection, the definition of normal patterns, deviation from normal, and criteria when to declare anomaly are required. In this contribution the process of acquisition of normal patterns (ground truth), and criteria when to declare anomaly based on graph comparison are proposed. The anomaly detection is suitable for deployment at the edge of a network. Finally, the inability to define all security threats is addressed by a custom tree-based classifier which only requires normal patterns for training. A simulated wireless sensor network was used to acquire data and apply the method. Our experiments show that data and topology change can be detected at the edge of a network.

As we move towards a zero-trust environment, collaborative intelligent systems are a vital tool in the cybersecurity workflow. These intelligent systems need to receive accurate and complete information from the multiplicity of low-level network monitoring systems. Typically, these low-level systems use regular expressions which are fast but not robust against minor changes - thus, information is often discarded. This paper outlines a new approach to low-level parsers, able to suggest new regular expression sequences and deal with approximate matching in an efficient way.

Current infrastructures are reaching the point where existing networking methods are unable to cope with the exponential growth of traffic and Quality of Service (QoS) requirements. New techniques are necessary to keep pace. One such technique, Software-Defined Networking (SDN) uses a central controller to program many individual network devices. However, SDN uses heuristic algorithms that do not always select the optimal path. This paper looked at creating three Q-Routing algorithms leveraging SDN and Mesh network topologies. Two algorithms used one network metric each (Latency and Bandwidth) and the third used multiple metrics. Results showed that the single metric Q-Routing algorithms on average performed as well as the K-Shortest Path versions while Q-Routing with multiple network metrics failed to match K-Shortest Path (different combination of metrics means these algorithms are not comparable). Results also showed that Q-Routing was able to calculate paths faster than K-Shortest Path in both static and dynamic networks.

There has been a strong trend towards autonomous and semi-autonomous systems in recent years. Evolving and adaptive systems embody the notion of autonomy, by changing their behavior (and possibly their structure) in response to changes in their environment. A consequence is that a designer may not be able to fully define the functional behavior of a system. Hence, formal verification and testing may not be possible. As a result, the self-adapting aspect of an evolving system is often implemented in an informal, ad hoc, manner and there is potential for causing significant harm if a system malfunctions in some way. A safety case requires more than an assertion that a system will work because it has not failed in testing. A more rigorous approach is essential, in which we can formally show that an evolving system meets its requirements and specifications. This paper outlines initial work in combining the X-mu approach (to model fuzzy uncertainty) with flexible requirements for an evolving system specified in RELAX, a formal framework to capture the uncertainty in evolving system requirements. A simple case study is used to illustrate some of the principles.

Theory of belief functions (Dempster-Shafer theory) is one of the most commonly used mathematical frameworks in the field of uncertain information representation. Two important areas of research in its context are that of evidence combination and decision making. Although they are often considered theoretically separate, the combination process itself drives the final decision. The information contained in each of the sources propagates through the belief aggregation process, and impacts the final assessment to some degree. If the decision made has an impact on the real-world; particularly in scenarios where a wrong decision may bring about significant risks it is prudent to be able to identify the key drivers of this assessment. In this paper we present a novel method of identifying the relative contribution of each source of evidence to the final belief, thus making it possible to track provenance across the information fusion process. This is likely to be useful in intelligent decision-support systems utilising belief functions, where lack of transparency may be hindering adoption. Unlike traditional methods, which focus on the content of the contributing source only, the approach proposed here is based on analysis of dissimilarity between the contributing sources; the result of the fusion process and the decision made. The behaviour of this metric is analysed through simulation. It is shown that the proposed measure performs well with regard to identifying the source having the most significant impact on the decision, often outperforming more traditional metrics.

Collaborative intelligence involves a combination of human and machine-based analysis, in which humans focus on higher-level tasks involving insight and understanding, whilst machines deal with gathering, filtering and processing data into a convenient and understandable form. We have proposed the use of graded concept lattices as a representation for exchanging information between machine and human in a collaborative intelligent system. Graded concepts allow summarization at multiple levels of discernibility (granularity). In this paper, we outline a new interpretation of fuzzy concept lattices as graded sets of crisp lattices. In addition, we prove equivalence between graded (fuzzy) formal concept analysis and the standard crisp framework. Consequently, any software tools developed for crisp data can be extended to the graded case without change.

In the current data-rich / knowledge-poor world, humans require machine assistance to summarize, analyze and understand a situation and the trends in events. The idea of collaborative intelligence enables humans to focus on higher-level tasks involving insight and understanding, whilst machines deal with gathering, filtering and processing data into a convenient and understandable form. In this paper, we propose graded concept lattices as a representation for exchanging information between machine and human in a collaborative intelligent system. Graded concepts allow summarization at multiple levels of discernibility (granularity). We present a novel incremental algorithm to find a graded concept lattice. The lattice can be used to identify associations in data at multiple levels of discernibility.

Unmanned aircraft systems are a specific type of critical embedded system that has been focus of research in recent years. In these systems, when fault of any kind occur, high value assets are placed at risk, and in some cases even human lives. Due to their high criticality, the development of improvements to these systems in order to increase the safety becomes an important topic in the area. However, it is always important to focus on efforts towards energy savings, which naturally pose new challenges to research, since it can significantly impact the performance and safety of these vehicles. This paper aims to study the possibility to disable internal modules of an aircraft (unmanned aerial vehicle) in order to save energy without sacrificing performance, vehicle safety or information security. A study of Fuzzy Logic applied to such scenario is presented in this paper.

Analysing event logs and identifying multiple overlapping sequences of events is an important task in web intelligence and in other applications involving data streams. It is ideally suited to a collaborative intelligence approach, where humans provide insight and machines perform the repetitive processing and data collection. A fuzzy approach allows flexible definition of the relations which link events into a sequence. In this paper we describe a virtual machine which enables a previously published expandable sequence pattern format to be represented as virtual machine instructions, which can filter event streams and identify fuzzily related sequences.

Projected clustering discovers clusters in subsets of locally relevant attributes. There is uncertainty and imprecision about how groups of categorical values are learnt from data for projected clustering and also the data itself. A method is presented for learning discrete possibility distributions of categorical values from data for projected clustering in order to model uncertainty and imprecision. Empirical results show that fewer, more accurate, more compact, and new clusters can be discovered by using possibility distributions of categorical values when compared to an existing method based on Boolean memberships. This potentially allows for new relationships to be identified from data.

Creative knowledge discovery—finding useful, previously unknown links between concepts—is a vital tool in unlocking the economic and social value of the vast range of networked data and services that is now available. We define “standard” knowledge discovery as the search for explanatory and predictive patterns in a specific domain, usually with a large volume of data. In contrast, creative knowledge discovery is concerned with the creation of new (and effective) patterns—either by generalization of existing patterns or by analogy to patterns in other domains. An important precondition for creative knowledge discovery is that we understand the relations within the data. Fuzzy formal concept analysis is a powerful approach that enables us to find embedded structure in data and to extract novel concepts that can be used in subsequent processing such as creative knowledge discovery. This paper outlines a fast algorithm for computing fuzzy formal concepts and provides a brief illustration of the use of fuzzy formal concept analysis in creative knowledge discovery.

It is important for Web search engine providers to study user behaviour in order to have a better understanding of how customers interact with search engines so that they can improve users' overall search experience. However, user behaviour in a search engine is complicated and affected by various factors, e.g. query length, intention/context/time when queries are submitted, etc. It is interesting to find answers to questions such as “whether loyal users are more likely to issue short length queries or medium length queries? If so, is that behaviour linked with high click through rate or is it linked with the user's previous search experience?” In this paper we argue that user behaviour should be better analysed from a subjective angle and introduce a granular analysis algorithm to intelligently extract user behaviour knowledge in a human-centric way to answer above questions. We study six variables relating to user behaviour study and demonstrate how fuzzy association rules mining based on mass assignment theory can intelligently analyse user activity patterns in a large scale Web search log data set.

Digital obesity, or information overload, is a widely recognized yet largely unsolved problem. Lack of metadata—that is, a useful and usable description of what is represented by data—is one of the fundamental obstacles preventing the wider use of computational intelligence techniques in tackling the problem of digital obesity. In this paper, we propose the use of fuzzy formal concept analysis to create simple taxonomies, which can be used to structure data and a novel form of fuzzy association rule to extract simple knowledge from data organized hierarchically according to the discovered taxonomies. The association strength is monitored over time, as data sets are updated. Feasibility of the methods is shown by applying them to a large (tens of thousands of entries) database describing reported incidents of terrorism. © 2010 Wiley Periodicals, Inc.

Recent initiatives in defence related information systems have emphasised the need to bring together information from multiple sources and fuse it into a form suitable for decision makers. This paper outlines a four stage system for fusing unstructured and semi-structured text and numerical data by extraction of entities and relations, identification of duplicate entities, organisation into the most appropriate hierarchical categories and determination of relations between fuzzy categories. The novel contribution of this paper is in the final stage of the process, where we determine associations between fuzzy categories and identify strong and/or unusual levels of association as well as changes over time. A demonstrator application shows how information on terrorist incidents from multiple sources can be integrated and monitored.

It is generally known that most incremental learning systems are order dependent, i.e provide results that depend on the particular order of the data presentation. Our previous work has developed an incremental soft computing algorithm which can be applied to learn text fragment patterns in semi-structured texts. A set of fuzzy grammar fragments is evolved, able to recognize the string set used as examples and any similar strings. Slight modification of the grammar fragments is performed to learn new patterns. This paper investigates the theoretical aspects of order-independence in the algorithm and shows that equivalent grammar fragments are produced irrespective of the order in which illustrative examples are presented.

One of the bottlenecks preventing wider adoption of the semantic Web is the overhead in annotating existing Web content. In cases where we have unstructured text, it is useful to extract fragments of structured data which can then be used as the basis for automatic tagging. A common approach is to use pattern matching (e.g. regular expressions) or more general grammar-based techniques, but these are not robust against small deviations. Fuzzy grammars allow partial matches, and we outline an efficient parsing technique to determine the degree to which a string is parsed by a grammar fragment. A simple application shows the methodpsilas validity.

In many applications, it is useful to extract structured data from sections of unstructured text. A common approach is to use pattern matching (e.g., regular expressions) or more general grammar-based techniques. In cases where exact templates or grammar fragments are not known, it is possible to use machine learning approaches, based on words or n-grams, to identify the structured data. This is generally a two-stage (train/use) process that cannot easily cope with incremental extensions of the training set. In this paper, we combine a fuzzy grammar-based approach with incremental learning. This enables a set of grammar fragments to evolve incrementally, each time a new example is given, while guaranteeing that it can parse previously seen examples. We propose a novel measure of overlap between fuzzy grammar fragments that can also be used to determine the degree to which a string is parsed by a grammar fragment. This measure of overlap allows us to compare the range of two fuzzy grammar fragments (i.e., to estimate and compare the sets of strings that fuzzily conform to each grammar) without explicitly parsing any strings. A simple application shows the method's validity.

The UK initiative in "network enabled capability" (NEC) can be loosely summarised as "right information, right place, right time - and not too much". It is closely related to the US concept of "network-centric operations". A core requirement for NEC is "active intelligence" where information streams from multiple sources can be dynamically blended and fused into the relevant form for a decision maker. We have identified "smart queries" and "adaptive data" (SQuAD) as the key components of active intelligence, and these form the basis of a current research programme. In this paper we outline three existing research streams that can contribute to the SQuAD project by using soft computing and computational intelligence to merge semi-structured data and taxonomic categories

Third generation telecare involves a customised sensor network, able to detect a person's movements and use of furniture and household items, coupled with a sophisticated fuzzy data analysis process able to answer high level queries such as "is the person eating regularly". Soft computing techniques are needed due to the high degree of uncertainty in the inference process. We describe a trial system, which has been installed in two homes. Results indicate that fuzzy analysis enables us to summarise the data in a manner which is useful to the care providers without them needing to be experts in data analysis.

Tele-care systems aim to carry out intelligent analyses of a person's wellbeing using data about their daily activities. This is a very challenging task because the massive dataset is likely to be erroneous, possibly with misleading sections due to noise or missing values. Furthermore, the interpretation of the data is highly sensitive to the lifestyle of the monitored person and the environment in which they interact. In our tele-care project, sensor-network domain knowledge is used to overcome the difficulties of monitoring long-term wellbeing with an imperfect data source. In addition, a fuzzy association analysis is leveraged to implement a dynamic and flexible analysis over individual- and environment-dependent data.

The problem of automating the sensing and classification of odours is one which promises a wide range of industrial applications. During the INTESA project, a prototype electronic nose was developed, using sensors based on novel conducting polymer materials and also more traditional MOS materials. The software component of the prototype processes the transient resistance change signals recorded by the hardware, and classifies the odour sample into one of a number of "odour classes". This paper describes two of the soft computing methods investigated for learning classification rules in this domain. The first method builds on previous work done on the Fril data browser, using clustering, fuzzy matching, Fril rules and evidential logic rules. The second method uses a fuzzy extension of the ID3 decision tree induction method, called "mass assignment tree induction (MATI)". Some of the results of applying these methods to data obtained from the INTESA prototype are presented and discussed.

Object-oriented modelling with words seeks to extend the new field of modelling with words, itself derived from computing with words, with features of object-oriented knowledge representation and programming. We show that object-oriented modelling with words has two key benefits: 1) the uncertain class hierarchy provides a natural knowledge representation framework for real-world problems; and 2) the object-oriented paradigm enforces a strict software engineering design and implementation process. In this paper we present a new approach to object-oriented modelling with words based on a new theory of inheritance where object memberships and property applicabilities are uncertain. We present the uncertain object-oriented logic programming language Fril++ as a tool for object-oriented modelling with words. We present an example application of object-oriented modelling with words: an extendible object-oriented data browser and machine learning environment.

User modelling is fast becoming a major topic for commercial and academic research. A major problem in user modelling is in categorising a new user into known user classes. We consider user classification as a recognition problem where a new user is an uncertain object of undetermined class and we must determine the membership of the object in each of a number of prototypical user classes. The temporal nature of user behaviour suggests a belief updating approach to user recognition. We examine the FILUM approach to user recognition and present a more generalised model. We compare the generalised FILUM model to a new interval support version of Hogarth and Einhorn's (1992) anchor and adjustment belief updating approach. We present an example using the iterated prisoner's dilemma problem implemented in the Fril++ object oriented uncertain logic programming language.

Logic programming, object-oriented programming and soft computing have provided advantageous methodologies and techniques for computer-based problem solving. This paper proposes a framework that combines these three disciplines to exploit their own advantages in dealing with real world problems. The framework is a logic-based one in which class and object properties are represented by clauses. Vague data in properties are represented by fuzzy sets interpreted as possibility distributions. Uncertain applicability of a property to a class or an object is represented either by a support pair defining probability lower and upper bounds, or by a certainty lower bound. Fundamental issues of uncertain membership and inheritance are then discussed and solutions to them are proposed. The result forms a basis for development of soft computing object-oriented programming systems.

Current learning approaches to computer vision have mainly focussed on low-level image processing and object recognition, while tending to ignore higher level processing for understanding. We propose an approach to scene analysis that facilitates the transition from recognition to understanding. It begins by segmenting the image into regions using standard approaches, which are then classified using a discovered fuzzy Cartesian granule feature classifier. Understanding is made possible through the transparent and succinct nature of the discovered models. The recognition of roads in images is taken as an illustrative problem. The discovered fuzzy models while providing high levels of accuracy (97%), also provide understanding of the problem domain through the transparency of the learnt models. The learning step in the proposed approach is compared with other techniques such as decision trees, naive Bayes and neural networks using a variety of performance criteria such as accuracy, understandability, and efficiency.

Cartesian granule features are derived features that are formed over the cross product of words that linguistically partition the universes of the constituent input features. Both classification and prediction problems can be modelled quite naturally in terms of Cartesian granule features incorporated into rule based models. The induction of Cartesian granule feature model involves discovering which input features should be combined to form Cartesian granule features in order to model a domain effectively; an exponential search problem. We present the G-DACG (Genetic Discovery of Additive Cartesian Granule feature models) constructive induction algorithm as a means of automatically identifying additive Cartesian granule feature models from example data. G-DACG combines the powerful optimisation capabilities of genetic programming with a rather novel and cheap fitness function which relies on the semantic separation of learnt concepts expressed in terms of Cartesian granule fuzzy sets. G-DACG is illustrated on a variety of artificial and real world classification problems.

A form of prototypes defined as tuples of marginal probability distributions is introduced. Addition and subtraction operations on such prototypes are then described. A brief introduction to the mass assignment theory of the probability of fuzzy events is given, and it is shown how fuzzy sets can serve as conceptual descriptions of probability distributions. Hence fuzzy descriptions of prototypes can be derived and these can be used for inference as well as enabling rule based representations of a set of prototypes to be formed. A prototype induction algorithm, based on these ideas together with the addition and subtraction operations, is described. The potential of this approach is then illustrated by its application to a number of model and real world machine learning problems. ©1999 John Wiley & Sons, Inc.

Notions of generalised fuzzy conditional and equivalence rules relative to a combination function are introduced and a framework for reasoning with such rules is described. The applicability of this framework to machine learning and knowledge discovery problems is demonstrated. Methods for the automatic generation of two particular types of generalised rule are proposed. The two rule forms considered are rules with weighted AND/OR combination functions, as suggested by Zimmerman and Zysno, and evidential logic equivalence rules as defined by Baldwin. The process of generating rule bases is divided into the problem of generating fuzzy sets from data and that of finding combination functions to optimise the performance of the system given these fuzzy sets. For the former problem a mass assignment based approach is adopted and for the latter semantic discrimination analysis is used in conjunction with customised optimisation algorithms. The potential of rule bases of both forms is illustrated by their application to a number of model and real world machine learning problems.

A mass assignment based ID3 algorithm for learning probabilistic fuzzy decision trees is introduced. Fuzzy partitions are used to discretize continuous feature universes and to reduce complexity when universes are discrete but with large cardinalities. Furthermore, the fuzzy partitioning of classification universes facilitates the use of these decision trees in function approximation problems. Generally the incorporation of fuzzy sets into this paradigm overcomes many of the problems associated with the application of decision trees to real-world problems. The probabilities required for the trees are calculated according to mass assignment theory applied to fuzzy labels. The latter concept is introduced to overcome computational complexity problems associated with higher dimensional mass assignment evaluations on databases. ©1997 John Wiley & Sons, Inc.