Nick J. Pizzi

A fuzzy classifier approach to estimating software quality

With the increasing sophistication of todays software systems, it is often difficult to estimate the overall quality of underlying software components with respect to attributes such as complexity, utility, and extensibility. Many metrics exist in the software engineering literature that attempt to quantify, with varying levels of accuracy, a large swath of qualitative attributes. However, the overall quality of a software object may manifest itself in ways that the simple interpretation of metrics fails to identify. A better strategy is to determine the best, possibly non-linear, subset of many software metrics for accurately estimating software quality. This strategy may be couched in terms of a problem of classification, that is, determine a mapping from a set of software metrics to a set of class labels representing software quality. We implement this strategy using a fuzzy classification approach. The software metrics are automatically computed and presented as features (input) to a classifier, while the class labels (output) are assigned via an experts (software architect) thorough assessment of the quality of individual software objects. A large collection of classifiers is presented with subsets of the software metric features. Subsets are selected stochastically using a fuzzy logic based sampling method. The classifiers then predict the quality, specifically the class label, of each software object. Fuzzy integration is applied to the results from the most accurate individual classifiers. We empirically evaluate this approach using software objects from a sophisticated algorithm development framework used to develop biomedical data analysis systems. We demonstrate that the sampling method attenuates the effects of confounding features, and the aggregated classification results using fuzzy integration are superior to the predictions from the respective best individual classifiers.

Indigenous populations health protection: A Canadian perspective

The disproportionate effects of the 2009 H1N1 pandemic on many Canadian Aboriginal communities have drawn attention to the vulnerability of these communities in terms of health outcomes in the face of emerging and reemerging infectious diseases. Exploring the particular challenges facing these communities is essential to improving public health planning. In alignment with the objectives of the Pandemic Influenza Outbreak Research Modelling (Pan-InfORM) team, a Canadian public health workshop was held at the Centre for Disease Modelling (CDM) to: (i) evaluate post-pandemic research findings; (ii) identify existing gaps in knowledge that have yet to be addressed through ongoing research and collaborative activities; and (iii) build upon existing partnerships within the research community to forge new collaborative links with Aboriginal health organizations. The workshop achieved its objectives in identifying main research findings and emerging information post pandemic, and highlighting key challenges that pose significant impediments to the health protection and promotion of Canadian Aboriginal populations. The health challenges faced by Canadian indigenous populations are unique and complex, and can only be addressed through active engagement with affected communities. The academic research community will need to develop a new interdisciplinary framework, building upon concepts from ‘Communities of Practice’, to ensure that the research priorities are identified and targeted, and the outcomes are translated into the context of community health to improve policy and practice.

Classification of volumetric storm cell patterns

Meteorological volumetric radar data are used to detect thunderstorms, storm events responsible for nearly all severe summer weather. Discriminating between different types of thunderstorms is challenge due to the high dimensionality of the data, the paucity of labeled data, and the imprecision of the labels. Several classification strategies and preprocessing techniques are tested to facilitate the discrimination between four types of storm events: wind, heavy rain, tornado and hail.

A Software Development Framework for Agent-Based Infectious Disease Modelling

From the Black Death of 1347–1350 (Murray, 2007) and the Spanish influenza pandemic of 1918–1919 (Taubenberger & Morens, 2006), to the more recent 2003 SARS outbreak (Lingappa et al., 2004) and the 2009 influenza pandemic (Moghadas et al., 2009), as well as countless outbreaks of childhood infections, infectious diseases have been the bane of humanity throughout its existence causing significant morbidity, mortality, and socioeconomic upheaval. Advanced modelling technologies, which incorporate the most current knowledge of virology, immunology, epidemiology, vaccines, antiviral drugs, and public health, have recently come to the fore in identifying effective disease mitigation strategies, and are being increasingly used by public health experts in the study of both epidemiology and pathogenesis. Tracing its historical roots from the pioneering work of Daniel Bernoulli on smallpox (Bernoulli, 1760) to the classical compartmental approach of Kermack and McKendrick (Kermack & McKendrick, 1927), modelling has evolved to deal with data that is more heterogeneous, less coarse (based at a community or individual level), and more complex (joint spatial, temporal and behavioural interactions). This evolution is typified by the agent-based model (ABM) paradigm, lattice-distributed collections of autonomous decision-making entities (agents), the interactions of which unveil the dynamics and emergent properties of the infectious disease outbreak under investigation. The flexibility of ABMs permits an effective representation of the complementary interactions between individuals characterized by localized properties and populations at a global level. However, with flexibility comes complexity; hence, the software implementation of an ABM demands more stringent software design requirements than conventional (and simpler) models of the spread and control of infectious diseases, especially with respect to outcome reproducibility, error detection and system management. Outcome reproducibility is a challenge because emergent properties are not analytically tractable, which is further exacerbated by subtle and difficult to detect errors in algorithm logic and software design. System management of software simulating populations/individuals and biological /physical interactions is a serious challenge, as the implementation will involve distributed (parallelized), non-linear, complex, and multiple processes operating in concert. Given these

Fuzzy quartile encoding as a preprocessing method for biomedical pattern classification

A fuzzy set based preprocessing method is described that may be used in the classification of patterns. This method, dispersion-adjusted fuzzy quartile encoding, determines the respective degrees to which a feature (attribute) belongs to a collection of fuzzy sets that overlap at the respective quartile boundaries of the feature. The fuzzy sets are adjusted to take into account the overall dispersion of values for a feature. The membership values are subsequently used in place of the original feature value. This transformation has a normalizing effect on the feature space and is robust to feature outliers. This preprocessing method, empirically evaluated using five biomedical datasets, is shown to improve the discriminatory power of the underlying classifiers.

Tactical Clinical Text Mining for Improved Patient Characterization

Clinical sources of information are markedly increasing in both volume and variety. A significant portion of the valuable data resides in the unstructured or semi-structured clinical text of documents stored in disparate repositories or embedded in HL7 messages. Clinical documents such as discharge summaries, prescriptions, lab reports, and free-form physician notes are filled with abbreviations, acronyms, misspellings, and ungrammatical phrases. However, synoptic reporting methods are restrictive for health care practitioners who wish to express critical and comprehensive patient information in electronic medical records. Furthermore, they have been superseded by systems that use natural language processing (NLP) to extract clinical concepts from free-form text. To address the growing need for efficient NLP solutions that can handle the volume and variety of clinical text, we have developed an optimized rules-based clinical concept extractor called TRACE (Tactical Rules-based AQL Clinical Extractor) using the Annotation Query Language (AQL). We present the experience we have gained applying text mining tools to this challenging domain, as well as a comparison of our solution to cTAKES (clinical Text Analysis and Knowledge Extraction System), an open-source clinical text miner, on a set of prescription documents. We also describe how efficient and scalable clinical text mining techniques will improve several of our companys offerings.

Mapping Software Metrics to Module Complexity: A Pattern Classification Approach

A desirable software engineering goal is the prediction of software module complexity (a qualitative concept) using automatically generated software metrics (quantitative measurements). This goal may be couched in the language of pattern classification; namely, given a set of metrics (a pattern) for a software module, predict the class (level of complexity) to which the module belongs. To find this mapping from metrics to complexity, we present a classification strategy, stochastic metric selection, to determine the subset of software metrics that yields the greatest predictive power with respect to module complexity. We demonstrate the effectiveness of this strategy by empirically evaluating it using a publicly available dataset of metrics compiled from a medical imaging system and comparing the prediction results against several classification system benchmarks.

Variance analysis and biomedical pattern classification

Component analysis is a common method used for the interpretation of data; however, in the case of pattern classification, the transformation of possibly correlated features into a new set of uncorrelated variables, must be used with caution since a principal component, which may account for significant variance in the data, is not necessarily discriminatory. To compensate for this deficiency, we present a classification method using an adaptive network of fuzzy logic connectives to select the most discriminatory principal components. We empirically evaluate the effectiveness of this classification method using a suite of biomedical datasets and comparing its performance against a set of benchmark classifiers.

Fuzzy adaptive logic networks

In this study, we elaborate on an important synergy between geometry and fuzzy logic in pattern recognition and show it translates into a coherent architecture of a classifier. The crux of the proposed topology lies in a collection of simple linear classifiers (perceptrons) being combined into a logically coherent topology. In a nutshell: perceptrons come with a simple geometrical interpretation while processing based on fuzzy operators (AND and OR logic units-fuzzy neurons) results in highly transparent and interpretable results. When combined together, forming a fuzzy adaptive logic network they give rise to the computing construct that retains the advantages of these two paradigms of information processing. We discuss a comprehensive development environment of adaptive logic networks and show their application to several classification problems.

Classification using an adaptive fuzzy network

The analysis of feature variance is a common approach used for data interpretation. In the case of pattern classification, however, the transformation of correlated features into a new set of uncorrelated variables must be used with caution, as there is no necessary causal connection between discriminatory power and variance. To compensate for this potential shortcoming, we present a classification method that blends variance analysis with an adaptive fuzzy logic network that identifies the most discriminatory set of uncorrelated variables. We empirically evaluate the effectiveness of this method using a suite of biomedical datasets and comparing its performance against two benchmark classifiers.