San-Yih Hwang

A probabilistic approach to modeling and estimating the QoS of web-services-based workflows

Web services promise to become a key enabling technology for B2B e-commerce. One of the most-touted features of Web services is their capability to recursively construct a Web service as a workflow of other existing Web services. The quality of service (QoS) of Web-services-based workflows may be an essential determinant when selecting constituent Web services and determining the service-level agreement with users. To make such a selection possible, it is essential to estimate the QoS of a WS workflow based on the QoSs of its constituent WSs. In the context of WS workflow, this estimation can be made by a method called QoS aggregation. While most of the existing work on QoS aggregation treats the QoS as a deterministic value, we argue that due to some uncertainty related to a WS, it is more realistic to model its QoS as a random variable, and estimate the QoS of a WS workflow probabilistically. In this paper, we identify a set of QoS metrics in the context of WS workflows, and propose a unified probabilistic model for describing QoS values of a broader spectrum of atomic and composite Web services. Emulation data are used to demonstrate the efficiency and accuracy of the proposed approach.

Dynamic Web Service Selection for Reliable Web Service Composition

This paper studies the dynamic web service selection problem in a failure-prone environment, which aims to determine a subset of Web services to be invoked at run-time so as to successfully orchestrate a composite web service. We observe that both the composite and constituent web services often constrain the sequences of invoking their operations and therefore propose to use finite state machine to model the permitted invocation sequences of Web service operations. We assign each state of execution an aggregated reliability to measure the probability that the given state will lead to successful execution in the context where each web service may fail with some probability. We show that the computation of aggregated reliabilities is equivalent to eigenvector computation and adopt the power method to efficiently derive aggregated reliabilities. In orchestrating a composite Web service, we propose two strategies to select Web services that are likely to successfully complete the execution of a given sequence of operations. A prototype that implements the proposed approach using BPEL for specifying the invocation order of a web service is developed and served as a testbed for comparing our proposed strategies and other baseline Web service selection strategies.

Efficient mining of group patterns from user movement data

In this paper, we present a new approach to derive groupings of mobile users based on their movement data. We assume that the user movement data are collected by logging location data emitted from mobile devices tracking users. We formally define group pattern as a group of users that are within a distance threshold from one another for at least a minimum duration. To mine group patterns, we first propose two algorithms, namely AGP and VG-growth. In our first set of experiments, it is shown when both the number of users and logging duration are large, AGP and VG-growth are inefficient for the mining group patterns of size two. We therefore propose a framework that summarizes user movement data before group pattern mining. In the second series of experiments, we show that the methods using location summarization reduce the mining overheads for group patterns of size two significantly. We conclude that the cuboid based summarization methods give better performance when the summarized database size is small compared to the original movement database. In addition, we also evaluate the impact of parameters on the mining overhead.

Mining mobile group patterns: a trajectory-based approach

In this paper, we present a group pattern mining approach to derive the grouping information of mobile device users based on a trajectory model. Group patterns of users are determined by distance threshold and minimum time duration. A trajectory model of user movement is adopted to save storage space and to cope with untracked or disconnected location data. To discover group patterns, we propose ATGP algorithm and TVG-growth that are derived from the Apriori and VG-growth algorithms respectively.

On the discovery of process models from their instances

A thorough understanding of the way in which existing business processes currently practice is essential from the perspectives of both process reengineering and workflow management. In this paper, we present a framework and algorithms that derive the underlying process model from past executions. The process model employs a directed graph for representing the control dependencies among activities and associates a Boolean function on each edge to indicate the condition under which the edge is to be enabled. By modeling the execution of an activity as an interval, we have developed an algorithm that derives the directed graph in a faster, more accurate manner. This algorithm is further enhanced with a noise handling mechanism to tolerate noise, which frequently occur in the real world. Experimental results show that the proposed algorithm outperforms the existing ones in terms of efficiency and quality.

Mining exception instances to facilitate workflow exception handling

The importance of exception handling within the context of workflow management has been widely recognized. While some exceptions are expected at design time and thus can be incorporated into the workflow design via some flexible mechanism, others are totally unexpected. Previous work in handling unexpected workflow exceptions focuses on the run-time support to, for example, allow the rollback of some already completed activities, validate the correctness of dynamic workflow change, and deploy a solution to handle exceptions. Authorized persons are responsible for deriving solutions to handle exceptions. We propose a novel approach to facilitating users in proposing solutions for resolving a given exception. Specifically, our approach scans through the previous records in handling exceptions, looking for those that are close to the current exception. The ways in which those exceptions were handled serve as useful information in determining how to handle the current one. Several algorithms are proposed and evaluated through both theoretical analysis and a synthetic data set.

On Mining Group Patterns of Mobile Users

In this paper, we present a group pattern mining approach to derive the grouping information of mobile device users based on the spatio-temporal distances among them. Group patterns of users are determined by a distance threshold and a minimum duration. To discover group patterns, we propose the AGP and VG-growth algorithms that are derived from the Apriori and FP-growth algorithms respectively. We further evaluate the efficiencies of these two algorithms using synthetically generated user movement data.

Discovery of temporal patterns from process instances

Existing work in process mining focuses on the discovery of the underlying process model from their instances. In this paper, we do not assume the existence of a single process model to which all process instances comply, and the goal is to discover a set of frequently occurring temporal patterns. Discovery of temporal patterns can be applied to various application domains to support crucial business decision-making. In this study, we formally defined the temporal pattern discovery problem, and developed and evaluated three different temporal pattern discovery algorithms, namely TP-Graph, TP-Itemset and TP-Sequence. Their relative performances are reported.

Consulting past exceptions to facilitate workflow exception handling

In this paper, we propose an architecture model that deals with both expected and unexpected exceptions in the context of workflow management. Expected exceptions and their handling approaches are specified by ECA rules, while cases of unexpected exceptions are characterized by their features and resolution approaches. The handling of unexpected exceptions is then assisted by the system providing information about how recent similar cases were resolved. The ways in which the previous exception cases were handled provides useful information in determining how to handle the current one. Quantifying the similarity of exception cases is described, and three algorithms for efficiently searching for similar exception cases are proposed and evaluated both theoretically and by experimenting with synthetic data sets. D 2002 Elsevier B.V. All rights reserved.

On Composing a Reliable Composite Web Service: A Study of Dynamic Web Service Selection

Dynamic Web service selection refers to determining a subset of component Web services to be invoked so as to orchestrate a composite Web service. Previous work in Web service selection usually assumes the invocations of Web service operations to be independent of one another. This assumption however does not hold in practice as both the composite and component Web services often impose some orderings on the invocation of their operations. Such orderings constrain the selection of component Web services to orchestrate the composite Web service. We therefore propose to use finite state machine (FSM) to model the invocation order of Web service operations. We define a measure, called aggregated reliability, to measure the probability that a given state in the composite Web service will lead to successful execution in the context where each component Web service may fail with some probability. We show that the computation of aggregated reliabilities is equivalent to eigenvector computation. The power method is hence adopted to efficiently derive aggregated reliabilities. In orchestrating a composite Web service, we propose two strategies to select component Web services that are likely to successfully complete the execution of a given sequence of operations. Our experiments on a synthetically generated set of Web service operation execution sequences show that our proposed strategies perform better than the baseline random selection strategy.