Geetika T. Lakshmanan

Placement Strategies for Internet-Scale Data Stream Systems

Optimally assigning streaming tasks to network machines is a key factor that influences a large data-stream-processing systems performance. Although researchers have prototyped and investigated various algorithms for task placement in data stream management systems, taxonomies and surveys of such algorithms are currently unavailable. To tackle this knowledge gap, the authors identify a set of core placement design characteristics and use them to compare eight placement algorithms. They also present a heuristic decision tree that can help designers judge how suitable a given placement solution might be to specific problems.

A stratified approach for supporting high throughput event processing applications

The quantity of events that a single application needs to process is constantly increasing. RFID related events have doubled within the past year and reached 4 trillion events per day, financial applications in large banks are processing 400 million events per day, and Massively Multiplayer Online (MMO) games are monitoring millions of events per second during peak periods. It is evident that scalability in event throughput is a major requirement for such applications. While the first generation of event processing systems is centralized, we see various solutions that attempt to use both scale-up and scale-out techniques. Alas, partitioning of the processing manually is difficult due to the semantic dependencies among event processing agents. It is also difficult to manually tune up the partition dynamically. This paper proposes a horizontal partition that is automatically created by analyzing the semantic dependencies among agents using a stratification principle. Each stratum contains a collection of independent agents, and events are routed to subsequent strata. We also implement a profiling-based technique for assigning agents to nodes in each stratum with the goal of maximizing throughput. A complementary step is to distribute load among different execution nodes dynamically based on their performance characteristics and the event traffic model. Experimental results show significant improvement in the ability to process high throughput of events relative to both centralized solutions as well as vertical partitions. We find this to be a promising approach to achieve high scalability particularly when the traffic model and network topology change frequently.

A markov prediction model for data-driven semi-structured business processes

In semi-structured case-oriented business processes, the sequence of process steps is determined by case workers based on available document content associated with a case. Transitions between process execution steps are therefore case specific and depend on independent judgment of case workers. In this paper, we propose an instance-specific probabilistic process model (PPM) whose transition probabilities are customized to the semi-structured business process instance it represents. An instance-specific PPM serves as a powerful representation to predict the likelihood of different outcomes. We also show that certain instance-specific PPMs can be transformed into a Markov chain under some non-restrictive assumptions. For instance-specific PPMs that contain parallel execution of tasks, we provide an algorithm to map them to an extended space Markov chain. This way existing Markov techniques can be leveraged to make predictions about the likelihood of executing future tasks. Predictions provided by our technique could generate early alerts for case workers about the likelihood of important or undesired outcomes in an executing case instance. We have implemented and validated our approach on a simulated automobile insurance claims handling semi-structured business process. Results indicate that an instance-specific PPM provides more accurate predictions than other methods such as conditional probability. We also show that as more document data become available, the prediction accuracy of an instance-specific PPM increases.

Investigating clinical care pathways correlated with outcomes

Clinical care pathway analysis is the process of discovering how clinical activities impact patients in their care journeys, and uses the discovered knowledge for various applications including the redesign and optimization of clinical pathways. We present an approach for mining clinical care pathways correlated with patient outcomes that involves a combination of clustering, process mining and frequent pattern mining. Our approach is implemented as a set of interactive tools in the business process insight (BPI) platform, a a collaborative software as a service platform, that provides an event-driven process-aware analytics toolset. After interactively utilizing the individual clustering, process mining, and frequent pattern mining capabilities in BPI, users can overlay frequent patterns, ranked according to their correlation with a particular patient outcome, on a mined model of the patient population with that outcome. We have tested our approach for mining care pathways correlated with outcomes on electronic medical record data obtained from a US based healthcare provider on congestive heart failure (CHF) patients. Experimental results show that the tools we have developed and implemented can provide new insights to facilitate the improvement of existing clinical care pathways.

Discovering event correlation rules for semi-structured business processes

In this paper we describe an algorithm to discover event correlation rules from arbitrary data sources. Correlation rules can be useful for determining relationships between events in order to isolate instances of a running business process for the purposes of monitoring, discovery and other applications. We have implemented our algorithm and validate our approach on events generated by a simulator that implements a real-world inspired export compliance regulations scenario consisting of 24 activities and corresponding event types. This simulated scenario involves a wide range of heterogeneous systems (e.g. Order Management, Document Management, E-Mail, and Export Violation Detection Services) as well as workflow-supported human-driven interactions (Process Management System). Experimental results demonstrate that our algorithm achieves a high level of accuracy in the detection of correlation rules. This paper confirms that our algorithm is a step towards semi-automating the task of detecting correlations. We also demonstrate how correlation rules discovered by our algorithm can be used to create aggregation nodes that allow more efficient querying, filtering and analytics. The results in this paper encourage future directions such as distributed statistics calculation, and scalability in terms of handling massive data sets.

A Business Centric End-to-End Monitoring Approach for Service Composites

Enterprise applications today are composed of multiple independently executing services and processes that collectively provide a solution to a business problem. These composite applications contain a heterogeneous collection of services that execute in a variety of runtimes making them difficult to manage while maintaining a business centric point of view, as opposed to a service point of view. This paper introduces a business centric monitoring framework to bridge the gap between the business and service levels in complex business applications. Our technical approach focuses on using business information invariants to define one or more monitor sets in order to relate service activity to business composite execution. We apply this framework to enable end-to-end monitoring of composite business applications. In this paper we present an initial prototype of our business centric monitoring approach using monitor sets for monitoring a simple loan application composite implemented on IBM’s WebSphere Business Modeler, Process Server and Business Monitor. Our prototype implementation demonstrates the convenience, effectiveness and ease of design and deployment of our monitoring solution to attain a single end-to-end business centric view of a collection of heterogeneous services executing together. Our work also exposes potential challenges as we extend this work to support more powerful end-to-end monitoring.

Business Process Insight: An Approach and Platform for the Discovery and Analysis of End-to-End Business Processes

Tracking and analyzing the execution of semi-structured processes is essential for understanding process behavior and its evolution, increasing the effectiveness of business operations, and managing operational risk. A semi structured process is a single process from the perspective of the business but is executed across loosely coupled, heterogeneous, distributed systems and may be cross-organizational and contain human interactions and human decision making. Our goal is to enable the management of semi-structured processes by providing improved semi-automated visibility into their behavior and improved runtime management of their execution by leveraging process intelligence and process-aware analytics. In this paper we present a methodology, system architecture and implementation of a platform for end to end process insight and analytics for the full life-cycle of semi-structured processes: from discovery through execution and evolution. We describe the algorithms employed and how they interact together to create a collaborative platform for business users to analyze and extract insight from such processes.

Guest Editors' Introduction: Provenance in Web Applications

The Web has completely changed the way in which we share data, rapidly shifting us from a world of paper documents to a world of digital objects that include online documents, videos, photos, artwork, and databases. This shift has also made data management an increasingly complex problem as applications take advantage of loosely coupled resources brought together by distributed computing systems and abundant storage capacity. Its now easier than ever to modify documents, particularly with the help of general-purpose specifications such as XML, and extract data from documents or databases through the use of technologies such as query languages, REST interfaces, and Web service interconnectivity.

Leveraging path information to generate predictions for parallel business processes

In semi-structured processes, the set of activities that need to be performed, their order and whether additional steps are required are determined by human judgment. There is a growing demand for operational support of such processes during runtime particularly in the form of predictions about the likelihood of future tasks. We address the problem of making predictions for a running instance of a semi-structured process that contains parallel execution paths where the execution path taken by a process instance influences its outcome. In particular, we consider five different models for how to represent an execution trace as a path attribute for training a prediction model. We provide a methodology to determine whether parallel paths are independent, and whether it is worthwhile to model execution paths as independent based on a comparison of the information gain obtained by dependent and independent path representations. We tested our methodology by simulating a marketing campaign as a business process model and selected decision trees as the prediction model. In the evaluation, we compare the complexity and prediction accuracy of a prediction model trained with five different models.

Predictive Analytics for Semi-structured Case Oriented Business Processes

The goal of our work is to examine the utility of predictive analytics for case-oriented semi-structured business processes. As a first step towards this goal, this paper describes an approach to leverage case history to predict outcomes at decision points in case-oriented semi-structured processes, and examine how the contents of documents at these decision points influence their outcomes. We apply an ant-colony optimization (ACO) based algorithm to create a probabilistic activity graph from traces, and use it to identify key decision points in a given process. For each activity node that represents a decision point in the mined probabilistic graph, the likelihood of different outcomes from the node can be correlated with the contents of documents accessed by the activity node. This is achieved by using a standard decision tree learning algorithm. We validate our approach on correlated case instance traces generated by a simulator that we constructed to implement non-deterministic executions of an automobile insurance claims scenario. In practice we find that our approach can lead to useful predictions at different stages of execution in a semi-structured case oriented process.