Izack Cohen

A core component of any AI-Augmented Business Process Management System (ABPMS) is the process frame, which gives the system process-awareness and defines its maximal behavioral boundaries. Compared to traditional process models, the process frame should, in principle, provide a somewhat more permissive representation of the managed processes, such that the (semi) autonomous behavior of an ABPMS, referred to as framed autonomy, could emerge. In addition, the process frame is not limited to a single linguistic or symbolic formalism and may incorporate heterogeneous knowledge ranging from predefined procedures to common sense rules and best practices. In this paper, we first conceptualize the ABPMS process frame as a hybrid business process representation, consisting of semi-concurrently executed procedural and declarative process models, extending the open-world assumption of the declarative paradigm also to procedural models. The latter allows any set of (non-conflicting) models of either type to be combined for execution, but complicates the automated discovery of these models from event data. Existing approaches for procedural models are particularly affected due to their reliance on observing directly-follows relations between pairs of activities. In search of an alternative, we present an in-depth analysis of how different procedural behaviors manifest as sets of discovered Declare constraints, each corresponding to a specific type of eventually-follows relation. This reveals behavioral overlaps between declarative and procedural models, while also laying the foundation for developing corresponding process (frame) discovery techniques.

Our focus is on projects, i.e., business processes, which are emerging as the economic drivers of our times. Differently from day-to-day operational processes that do not require detailed planning, a project requires planning and resource-constrained scheduling for coordinating resources across sub- or related projects and organizations. A planner in charge of project planning has to select a set of activities to perform, determine their precedence constraints, and schedule them according to temporal project constraints. We suggest a data-driven project planning approach for classes of projects such as infrastructure building and information systems development projects. In such projects, a significant portion of activities recurs within other organizational projects, which may be similar, while each project is unique in its realization. The first steps of the suggested approach include learning a project network from historical records of similar projects. The discovered network relaxes temporal constraints embedded in individual projects, thus uncovering where planning and scheduling flexibility can be exploited for greater benefit. Then, the network, which contains multiple project plan variations, is enriched by identifying decision rules and frequent paths in favor of selecting a specific variation as the chosen project plan. The planner can rely on the suggested approach for: 1) Decoding a project variation such that it forms a new project plan and 2) applying resource-constrained project scheduling procedures to determine the project's schedule and resource allocation. Using two real-world project datasets, we show that the suggested approach may provide the planner with significant flexibility (up to a 26% reduction of the critical path of a real project) to adjust the project plan and schedule. We believe that the proposed approach can play an important part in supporting decision making toward automated data-driven project planning.

Developments in machine learning together with the increasing usage of sensor data challenge the reliance on deterministic logs, requiring new process mining solutions for uncertain, and in particular stochastically known, logs. In this work we formulate {trace recovery}, the task of generating a deterministic log from stochastically known logs that is as faithful to reality as possible. An effective trace recovery algorithm would be a powerful aid for maintaining credible process mining tools for uncertain settings. We propose an algorithmic framework for this task that recovers the best alignment between a stochastically known log and a process model, with three innovative features. Our algorithm, SKTR, 1) handles both Markovian and non-Markovian processes; 2) offers a quality-based balance between a process model and a log, depending on the available process information, sensor quality, and machine learning predictiveness power; and 3) offers a novel use of a synchronous product multigraph to create the log. An empirical analysis using five publicly available datasets, three of which use predictive models over standard video capturing benchmarks, shows an average relative accuracy improvement of more than 10 over a common baseline.

Customer waiting time prediction is key to managing service systems. Predicting how long a customer will wait for service at the time of their arrival can provide important information to the customer and serve as a tool for the operations manager. Recent studies that suggested machine learning algorithms for waiting time prediction as an alternative to the standard queueing theory approaches investigated specific systems with mixed results regarding the superiority of a particular approach. We provide a systematic investigation of common violations of queueing theory assumptions on waiting time prediction in the context of single-queue many-server systems. These violations include nonstationarity, nonexponential service times, state-dependent service times, abandonments, and customers with different priorities. Using different machine learning models as well as queueing-theory-based methods, we seek to determine under what regimes machine learning prediction is to be preferred to queueing-theory-based predictors. Our results suggest that queueing theory models produce comparable and frequently better predictions versus machine learning algorithms at a much lower computational cost. Under other assumptions, such as high priority for a specific type of customer, machine learning predictions may outperform queueing theory predictions. Our results may guide the selection of a delay prediction approach for service systems.