Scenario planning is a commonly used method by companies to develop their long-term plans. Scenario planning for risk management puts an added emphasis on identifying and managing emerging risk. While a variety of methods have been proposed for this purpose, we show that applying AI planning techniques to devise possible scenarios provides a unique advantage for scenario planning. Our system, the Scenario Planning Advisor (SPA), takes as input the relevant information from news and social media, representing key risk drivers, as well as the domain knowledge and generates scenarios that explain the key risk drivers and describe the alternative futures. To this end, we provide a characterization of the problem, knowledge engineering methodology, and transformation to planning. Furthermore, we describe the computation of the scenarios, lessons learned, and the feedback received from the pilot deployment of the SPA system in IBM.

Imagining the future helps anticipate and prepare for what is coming. This has great importance to many, if not all, human endeavors. In this paper, we develop the Planning Projector system prototype, which applies plan-recognition-as-planning technique to both explain the observations derived from analyzing relevant news and social media, and project a range of possible future state trajectories for human review. Unlike the plan recognition problem, where a set of goals, and often a plan library must be given as part of the input, the Planning Projector system takes as input the domain knowledge, a sequence of observations derived from the news, a time horizon, and the number of trajectories to produce. It then computes the set of trajectories by applying a planner capable of finding a set of high-quality plans on a transformed planning problem. The Planning Projector prototype integrates several components including: (1) knowledge engineering: the process of encoding the domain knowledge from domain experts; (2) data transformation: the problem of analyzing and transforming the raw data into a sequence of observations; (3) trajectory computation: characterizing the future state projection problem and computing a set of trajectories; (4) user interface: clustering and visualizing the trajectories. We evaluate our approach qualitatively and conclude that the Planning Projector helps users understand future possibilities so that they can make more informed decisions.

In this paper, we address the problem of finding diverse high-quality plans motivated by the hypothesis generation problem. To this end, we present a planner called TK*that first efficiently solves the “top-k” cost-optimal planning problem to find k best plans, followed by clustering to produce diverse plans as cluster representatives.

A Data Scientist typically performs a number of tedious and time-consuming steps to derive insight from a raw data set. The process usually starts with data ingestion, cleaning, and transformation (e.g. outlier removal, missing value imputation), then proceeds to model building, and finally a presentation of predictions that align with the end-users objectives and preferences. It is a long, complex, and sometimes artful process requiring substantial time and effort, especially because of the combinatorial explosion in choices of algorithms (and platforms), their parameters, and their compositions. Tools that can help automate steps in this process have the potential to accelerate the time-to-delivery of useful results, expand the reach of data science to non-experts, and offer a more systematic exploration of the available options. This work presents a step towards this goal.

In this paper we apply AI planning to address the hypothesis exploration problem and provide assistance to network administrators in detecting malware based on unreliable observations derived from network traffic.Building on the already established characterization and use of AI planning for similar problems, we propose a formulation of the hypothesis generation problem for malware detection as an AI planning problem with temporally extended goals and actions costs. Furthermore, we propose a notion of hypothesis ``plausibility'' under unreliable observations, which we model as plan quality. We then show that in the presence of unreliable observations, simply finding one most ``plausible'' hypothesis, although challenging, is not sufficient for effective malware detection. To that end, we propose a method for applying a state-of-the-art planner within a principled exploration process, to generate multiple distinct high-quality plans. We experimentally evaluate this approach by generating random problems of varying hardness both with respect to the number of observations, as well as the degree of unreliability. Based on these experiments, we argue that our approach presents a significant improvement over prior work that are focused on finding a single optimal plan, and that our hypothesis exploration application can motivate the development of new planners capable of generating the top high-quality plans.

With the widespread adoption of location tracking technologies like GPS, the domain of intelligent transportation services has seen growing interest in the last few years. Services in this domain make use of real-time location-based data from a variety of sources, combine this data with static location-based data such as maps and points of interest databases, and provide useful information to end-users. Some of the major challenges in this domain include i) scalability, in terms of processing large volumes of real-time and static data; ii) extensibility, in terms of being able to add new kinds of analyses on the data rapidly, and iii) user interaction, in terms of being able to support different kinds of one-time and continuous queries from the end-user. In this paper, we demonstrate the use of IBM InfoSphere Streams, a scalable stream processing platform, for tackling these challenges. We describe a prototype system that generates dynamic, multi-faceted views of transportation information for the city of Stockholm, using real vehicle GPS and road-network data. The system also continuously derives current traffic statistics, and provides useful value-added information such as shortest-time routes from real-time observed and inferred traffic conditions. Our performance experiments illustrate the scalability of the system. For instance, our system can process over 120000 incoming GPS points per second, combine it with a map containing over 600,000 links, continuously generate different kinds of traffic statistics and answer user queries.

In this paper, we tackle the problem of helping domain experts to construct, parameterize and deploy mashups of data and code. We view a mashup as a data processing flow, that describes how data is obtained from one or more sources, processed by one or more components, and finally sent to one or more sinks. Our approach allows specifying patterns of flows, in a language called Cascade. The patterns cover different possible variations of the flows, including variations in the structure of the flow, the components in the flow and the possible parameterizations of these components. We present a tool that makes use of this knowledge of flow patterns and associated metadata to allow domain experts to explore the space of possible flows described in the pattern. The tool uses an AI planning approach to automatically build a flow, belonging to the flow pattern, from a high-level goal, specified as a set of tags. We describe examples from the financial services domain to show the use of flow patterns in allowing domain experts to construct a large variety of mashups rapidly.

The Web services research community has proposed a number of approaches for service composition, ranging from manual to semi-automatic to completely automatic. However, it is often difficult to take independently developed services and compose them, since they may not work together correctly. For service composition to occur, the services in question must be designed and developed in a manner that facilitates their composition. In this paper, we propose a novel approach for service design and composition that combines top-down and bottom-up elements. Our approach is driven by faceted, tag-based functional requirements provided by end-users. These requirements describe, at a high-level, the families of compositions that end-users desire. The requirements kick off a top-down service development lifecycle, where enterprise architects and service developers design, develop and test workflows and services, possibly reusing existing flows and services in the process. At runtime, end-users can specify goals, which are satisfied through a bottom-up composition of flows from the available services. The composed flows include those explicitly designed by the architects as well as new ones that are assembled in a serendipitous manner from the available services. With examples from a case study in the financial services domain, we demonstrate our approach for designing and developing services that can be composed into myriad workflows based on end-user goals.

In the realm of component-based software systems, pursuers of the holy grail of automated application composition face many significant challenges. In this paper we argue that, while the general problem of automated composition in response to high-level goal statements is indeed very difficult to solve, we can realize composition in a restricted context, supporting varying degrees of manual to automated assembly for specific types of applications. We propose a novel paradigm for composition in flow-based information processing systems, where application design and component development are facilitated by the pervasive use of faceted, tag-based descriptions of processing goals, of component capabilities, and of structural patterns of families of application. The facets and tags represent different dimensions of both data and processing, where each facet is modeled as a finite set of tags that are defined in a controlled folksonomy. All data flowing through the system, as well as the functional capabilities of components are described using tags. A customized AI planner is used to automatically build an application, in the form of a flow of components, given a high-level goal specification in the form of a set of tags. End-users use an automatically populated faceted search and navigation mechanism to construct these high-level goals. We also propose a novel software engineering methodology to design and develop a set of reusable, well-described components that can be assembled into a variety of applications. With examples from a case study in the Financial Services domain, we demonstrate that composition using a faceted, tag-based application design is not only possible, but also extremely useful in helping end-users create situational applications from a wide variety of available components.

With the emergence of Yahoo Pipes and several similar services, data mashup tools have started to gain interest of business users. Making these tools simple and accessible ton users with no or little programming experience has become a pressing issue. In this paper we introduce MARIO (Mashup Automation with Runtime Orchestration and Invocation), a new tool that radically simplifies data mashup composition. We have developed an intelligent automatic composition engine in MARIO together with a simple user interface using an intuitive "wishful search" abstraction. It thus allows users to explore the space of potentially composable data mashups and preview composition results as they iteratively refine their "wishes", i.e. mashup composition goals. It also lets users discover and make use of system capabilities without having to understand the capabilities of individual components, and instantly reflects changes made to the components by presenting an aggregate view of changed capabilities of the entire system. We describe our experience with using MARIO to compose flows of Yahoo Pipes components.

Disclosing any information contained within an information system that stores personal data can be associated with risk. Nevertheless, the risk of privacy violation is often considered acceptable, since otherwise the most routine business operations can become impossible. Traditional privacy protection methods limit this risk indirectly by using access control policies for the protection of private information, authorizing the release of information only when the purpose of access justifies doing so. While simple and robust, these policies are binary, and therefore they can be too rigid in practice. A data access operation that is only slightly more risky than usual will be denied, and treated no differently than disclosing all possible data contained in the system. If the risk was justified, the access control policy will be modified later to allow it, but the original declined operation will not be performed in time. In this paper we build upon existing research in disclosure risk assessment, and propose a new flexible privacy protection approach based on soft constraints, as opposed to the hard constraints of traditional systems. The proposed model uses identifiability risk computation to estimate the risk of data access, and allows those requesting data access to decide whether the risk is justified. To prevent abuse of the system, each granted access will be recorded, and those taking high risks will need to justify their decisions later. However, the system will not decline access at the time when the request is made, unless, of course, the risk is unjustifiably high. We believe that this novel approach will help achieve the perfect balance between privacy protection and business efficiency. We illustrate our approach using data published by the U. S. Census Bureau.

In this paper, we propose a novel way of modeling Web services using semantic graph transformations. Each operation supported by a Web service is associated with a semantic annotation that describes the input and output messages using RDF graph patterns. The terms used in these patterns are defined in OWL ontologies that describe the application domain. A key difference between our model and existing semantic Web service models like OWLS is that it describes the inputs and outputs in terms of instance-based graph patterns, rather than in terms of concepts. This allows associating a rich set of constraints on the input and output data in terms of relations between instances. We also propose a composition model for Web service operations, that describes the conditions for composing services into workflows. The composition model includes the notion of semantic propagation, i.e. the semantic description of the output message of an operation depends on the semantics of the input message. We have developed a planner that uses this model to compose services, automatically. The planner uses DLP reasoning to aid plan search. We present performance results for the planner.

Intelligence Transportation Systems are critical to improve the efficiency of modern transportation. A system that is flexible and powerful enough to handle diverse demands from a large user base, is still elusive. Studies have shown that developing and integrating the various components constitute a significant portion of the capital cost and complexity of such systems. In this paper, we present a stream processing infrastructure we call System S. System S enables the deployment of large scale applications. It supports a mechanism for sharing data sources, software components, and even intermediate results allowing a reduction in the cost of software integration, and ownership. We experiment the stream processing infrastructure with a Fleet Management Center, and demonstrate how the infrastructure can be used to address unique issues in traffic management.

We consider reliable multicast in overlay networks where nodes have finite-size buffers and are subject to failures. We address issues of end-to-end reliability and throughput scalability in this framework. We propose a simple architecture which consists of using distinct point-to-point TCP connections between adjacent pairs of end-systems, together with a back-pressure control mechanism regulating the transfers of adjacent TCP connections, as well as a back-up buffering system handling node failures. This architecture, that we call the one-to-many TCP overlay, is a natural extension of TCP to the one-to-many case, in that it adapts the rate of the group communication to local congestion in a decentralized way via the window back-pressure mechanism. Using theoretical investigations, experimentations in the Internet, and large network simulations, we show that this architecture provides end-to-end reliability and can tolerate multiple simultaneous node failures, provided the backup buffers are sized appropriately. We also show that under random perturbations caused by cross traffic described in the paper, the throughput of this reliable group communication is always larger than a positive constant, that does not depend on the group size. This scalability result contrasts with known results about the non-scalability of IP-supported multicast for reliable group communication.

Overlay multicast (or application-level multicast) has become an increasingly popular alternative to IP-supported multicast. End nodes participating in overlay multicast can form a directed tree rooted at the source using existing unicast links. For each receiving node there is always only one incoming link. Very often, nodes can support no more than a fixed number of outgoing links due to bandwidth constraints. Here, we describe an algorithm for constructing a multicast tree with the objective of minimizing the maximum communication delay (i.e. the longest path in the tree), while satisfying degree constraints at nodes. We show that the algorithm is a constant-factor approximation algorithm. We further prove that the algorithm is asymptotically optimal if the communicating nodes can be mapped into Euclidean space such that the nodes are uniformly distributed in a convex region. We evaluate the performance of the algorithm using randomly generated configurations of up to 5,000,000 nodes.

This study provides some new insights into the scalability of reliable group communication mechanisms using overlays. These mechanisms use individual TCP connections for packet transfers between end-systems. End-systems store incoming packets and forward them to downstream nodes using different unicast TCP connections. In this paper we assume that buffers in end-systems are large enough for the transfers. It is shown that the throughput of the reliable overlay group communication scales in the sense that for all multicast tree sizes and topologies, the group throughput is strictly positive under natural conditions. This is in contrast with the IP supported multicast paradigm where reliable protocols have vanishing throughput when the group size tends to infinity. The scalability of packet delay and buffer occupancy is then investigated. In the absence of additional control, the occupancy of the buffer and the latency in the end-systems explodes with time. It is then shown that proactive rate throttle mechanism implemented at the source leads to finite packet latency and buffer occupancy in any end-system of the network provided certain moment conditions are satisfied by cross traffic in the routers.

This paper introduces new algorithms specifically designed for content-based publication-subscription systems. These algorithms can be used to determine multicast groups with as much commonality as possible, based on the totality of subscribers' interests. The algorithms are based oil concepts borrowed from the literature on spatial databases and clustering. These algorithms perform well in the context of highly heterogeneous subscriptions, and they also scale well. Based on concepts borrowed from the spatial database literature, we develop an algorithm to match publications to subscribers in real-time. We also investigate the benefits of dynamically determining whether to unicast, multicast or broadcast information about the events over the network to the matched subscribers. We call this the distribution method problem. Some of these same concepts can be applied to match publications to subscribers in real-time, and also to determine dynamically whether to unicast, multicast or broadcast information about the events over the network to the matched subscribers. We demonstrate the quality of our algorithms via a number of realistic simulation experiments.

We consider efficient communication schemes based on both network-supported and application-level multicast techniques for content-based publication-subscription systems. We show that the communication costs depend heavily on the network configurations, distribution of publications and subscriptions. We devise new algorithms and adapt existing partitional data clustering algorithms. These algorithms can be used to determine multicast groups with as much commonality as possible, based on the totality of subscribers' interests. They perform well in the context of highly heterogeneous subscriptions, and they also scale well. An efficiency of 60% to 80% with respect to the ideal solution can be achieved with a small number of multicast groups (less than 100 in our experiments). Some of these same concepts can be applied to match publications to subscribers in real-time, and also to determine dynamically whether to unicast, multicast or broadcast information about the events over the network to the matched subscribers. We demonstrate the quality of our algorithms via simulation experiments.