Glenn I. Hawe

Agent-based simulation for large-scale emergency response: A survey of usage and implementation

When attempting to determine how to respond optimally to a large-scale emergency, the ability to predict the consequences of certain courses of action in silico is of great utility. Agent-based simulations (ABSs) have become the de facto tool for this purpose; however, they may be used and implemented in a variety of ways. This article reviews existing implementations of ABSs for large-scale emergency response, and presents a taxonomy classifying them by usage. Opportunities for improving ABS for large-scale emergency response are identified.

A multi-objective combinatorial model of casualty processing in major incident response

During the emergency response to mass casualty incidents decisions relating to the extrication, treatment and transporting of casualties are made in a real-time, sequential manner. In this paper we describe a novel combinatorial optimization model of this problem which acknowledges its temporal nature by employing a scheduling approach. The model is of a multi-objective nature, utilizing a lexicographic view to combine objectives in a manner which capitalizes on their natural ordering of priority. The model includes pertinent details regarding the stochastic nature of casualty health, the spatial nature of multi-site emergencies and the dynamic capacity of hospitals. A Variable Neighborhood Descent metaheuristic is employed in order to solve the model. The model is evaluated over a range of potential problems, with results confirming its effective and robust nature.

Agent-based simulation of emergency response to plan the allocation of resources for a hypothetical two-site major incident

During a major incident, the emergency services work together to ensure that those casualties who are critically injured are identified and transported to an appropriate hospital as fast as possible. If the incident is multi-site and resources are limited, the efficiency of this process is compromised as the finite resources must be shared among the multiple sites. In this paper, agent-based simulation is used to determine the allocation of resources for a two-site incident which minimizes the latest hospital arrival times for critically injured casualties. Further, how the optimal resource allocation depends on the distribution of casualties across the two sites is investigated. Such application supports the use of agent-based simulation as a tool to aid emergency response.

Investigating the effect of overtriage on hospital arrival times of critically injured casualties during a major incident using agent-based simulation

This paper uses agent-based simulation to simulate the prehospital response to a hypothetical major incident in the UK. The rate of overtriage by the operational-level secondary triage officer is varied, and its effect on the latest arrival time of a critically injured casualty to hospital is modelled as a function of the number of ambulances involved in the response. Modelling such relationships could aid strategic-level planning for emergencies, by providing insight into how to compensate for the effect of overtriage.

Online optimization of casualty processing in major incident response: An experimental analysis

When designing an optimization model for use in mass casualty incident (MCI) response, the dynamic and uncertain nature of the problem environment poses a significant challenge. Many key problem parameters, such as the number of casualties to be processed, will typically change as the response operation progresses. Other parameters, such as the time required to complete key response tasks, must be estimated and are therefore prone to errors. In this work we extend a multi-objective combinatorial optimization model for MCI response to improve performance in dynamic and uncertain environments. The model is developed to allow for use in real time, with continuous communication between the optimization model and problem environment. A simulation of this problem environment is described, allowing for a series of computational experiments evaluating how model utility is influenced by a range of key dynamic or uncertain problem and model characteristics. It is demonstrated that the move to an online system mitigates against poor communication speed, while errors in the estimation of task duration parameters are shown to significantly reduce model utility.

The STORMI Scenario Designer: A program to facilitate setting up agent-based simulations of major incident emergency response

Compared to live exercises, agent-based simulation is an inexpensive method to perform ‘what-if’ style experiments of emergency response. However, the lack of a user-friendly interface often prevents such software being widely adopted by practitioners. This paper describes the STORMI Scenario Designer, a program designed specifically to facilitate practitioners in setting up agent-based simulations of emergency scenarios.

Ensemble learning utilising feature pairings for intrusion detection

Network intrusions may illicitly retrieve data/information, or prevent legitimate access. Reliable detection of network intrusions is an important problem, misclassification of an intrusion is an issue in and of itself reducing overall accuracy of detection. A variety of potential methods exist to develop an improved system to perform classification more accurately. Feature selection is one potential area that may be utilized to successfully improve performance by initially identifying sets and subsets of features that are relevant and nonredundant. Within this paper explicit pairings of features have been investigated in order to determine if the presence of pairings has a positive effect on classification, potentially increasing the accuracy of detecting intrusions correctly. In particular, classification using the ensemble algorithm, StackingC, with F-Measure performance and derived Information Gain Ratio, as well as their subsequent correlation as a combined measure, is presented.

Simulating the spatial organization of the UK Ambulance Service at major incident sites

Documentation such as generic major incident plans and action cards describe how the UK Ambulance Service is to respond to major incidents, including how paramedics should organize themselves at the incident site. Towards the aim of testing existing procedures to hypothetical events in-silico, this paper proposes a way to simulate the on-site spatial organization of the UK Ambulance Service as part of an agent-based simulation of emergency response. The 2001 Selby train crash is used as a case study to demonstrate the approach.

Multi-objective optimization of base classifiers in StackingC by NSGA-II for intrusion detection

Multiple Classifier Systems are often found to improve results of intrusion detection by combining a set of classifier decisions where single classifiers may not achieve the same level of detection. However not every set of classifiers is more able, therefore selection of more capable sets is required. A misclassification is a false positive or negative instance; a set of classifiers may produce one more than the other. An optimal set of classifiers is required to reduce both, thus treating them as individual objectives allows a balance to be found. The aim of this work is the selection of optimal sets of base level classifies using an evolutionary computation approach. A comparative analysis is made of the performance of the generated ensembles against the individual base level classifiers, it is shown that optimal ensembles can be found to perform better than a majority of individuals.

Centrality-Based Approach for Supervised Term Weighting

The huge amount of text documents has made the manual organization of text data a tedious task. Automatic text classification helps to easily handle the large number of documents by organising them automatically into predefined classes. The effectiveness and efficiency of automatic text classification largely depends on the way text documents are represented. A text document is usually viewed as a bag of terms (or words) and represented as a vector using the vector space model where terms are assumed unordered and independent and term frequencies (or weights) are used in the representation. Graphs are another text representation scheme that considers the structure of terms in the text document which is important for natural language. Terms weighted on the basis of graph representation increase the performance of text classification. In this paper, we present a novel approach for graph-based supervised term weighting which considers information relevant for the classification task using node centrality in the co-occurrence graphs built from the labelled training documents. Our experimental evaluation of the proposed term weighting scheme on four benchmark datasets shows the scheme has consistently superior performance over the state-of-the-art term weighting methods for text classification.