event detection and segmentation

Existing evaluation approaches for streaming process mining algorithms predominantly rely on static logs or synthetic event streams, which fail to capture the complexity of real-world event streams in IoT environments—such as out-of-order events, concurrency, incomplete cases, and concept drift. This work addresses this gap by introducing, for the first time, a feature framework from data stream research into streaming process mining. It proposes an intent-oriented event stream generation methodology, extends the conceptual model of event streams, and implements a prototype tool, Stream of Intent. This tool enables customizable configuration of key stream characteristics reflective of real-world scenarios, facilitating the generation of controlled, reproducible, and realistically complex event streams. Consequently, it significantly enhances the relevance and adaptability of algorithm evaluation and development in streaming process mining.

This study addresses sequential multi-stream detection under the constraint that only one data stream can be observed at each time step, with the goal of simultaneously controlling global false alarm and missed detection probabilities while minimizing detection delay. To this end, the work introduces a novel optimality criterion based on the expected order statistics of detection times and proposes an active sampling strategy—dubbed “follow-the-leader”—that integrates exploration and exploitation mechanisms. Theoretical analysis demonstrates that the proposed strategy achieves asymptotic optimality for all such criteria as error probabilities vanish. Numerical experiments further confirm its superior finite-sample performance compared to existing methods and show that it closely approaches the performance of an ideal oracle policy that has full knowledge of the anomalous streams.

In streaming process mining, fixed-size sliding windows struggle to accommodate dynamic process evolution and concept drift, leading to model bias. To address this, we propose a dynamic window optimization method grounded in species estimation theory—introducing, for the first time, sample representativeness quantification into streaming process mining. Our approach establishes a real-time representativeness assessment model under sliding windows and adaptively adjusts window size to balance timeliness and statistical sufficiency. It requires no prior knowledge and enables online detection and response to concept drift. Experiments on multiple real-world event streams demonstrate that our method significantly improves process model accuracy (average +12.7% F1-score) and robustness to concept drift (38.5% reduction in false positive rate) compared to static-window baselines. This work establishes a novel paradigm for real-time, adaptive process analysis.

This paper addresses the limitation of generic event boundary detection (GEBD)—its reliance on complete videos and inability to operate in real time—by proposing online GEBD (On-GEBD): a new task that detects category-agnostic, fine-grained event boundaries in streaming video using only historical frames. To tackle robust boundary discrimination without future-frame context, the authors introduce the Estimator framework, which integrates a Consistent Event Anticipation (CEA) module for frame-level prediction and an Online Boundary Discriminator (OBD) that dynamically identifies boundaries via error-statistical hypothesis testing and adaptive thresholding. This work is the first to adapt event segmentation theory to an online setting, significantly enhancing immediate perception of diverse, weak-signal event transitions in long videos. On Kinetics-GEBD and TAPOS benchmarks, On-GEBD outperforms all existing online baselines and approaches the performance of optimal offline methods.

Traditional Business Process Management (BPM) struggles to unify discrete events with continuous sensor signals from Cyber-Physical Systems (CPS). Existing Signal Temporal Logic (STL)-based hybrid declarative approaches support only retrospective monitoring and lack real-time execution capabilities. To address this, we propose the first three-layer architecture enabling real-time execution of hybrid declarative processes, deeply integrating STL into a Complex Event Processing (CEP) engine. This integration supports joint temporal constraints over discrete events and real-valued signals, proactive activity triggering, and dynamic enforcement of process boundaries. Our approach achieves, for the first time in BPM, STL-driven online execution and closed-loop control—bridging the semantic and execution gap between declarative modeling and real-time physical-world operation. We validate its effectiveness and scalability in智能制造 and Industrial IoT scenarios.

This work addresses the challenge that continuous, high-channel data streams from distributed acoustic sensing (DAS) are poorly supported by conventional batch-processing frameworks, hindering interactive exploration, scalable annotation, and real-time algorithm integration. To overcome this, we propose FiLark—the first DAS analysis framework built entirely around a “stream-first” abstraction. FiLark unifies multi-file or continuous recordings into a constant-memory data stream, enabling low-memory interactive browsing of arbitrarily long records, in-stream generation of machine learning–ready labels, and seamless transition from development to production environments. Implemented in Python, FiLark integrates OpenGL-based circular-buffer rendering, CPU/GPU-accelerated time–space–frequency signal operators, stateful chunked execution, and standardized monitoring interfaces, substantially enhancing the efficiency, scalability, and end-to-end reproducibility of DAS analytics.

This study addresses the challenge of rapid change-point detection in high-dimensional multisensor systems under structural constraints and limited sensing resources. By integrating sparse modeling, heterogeneous data fusion, and a resource-adaptive sequential sampling strategy, the work extends classical change-point detection theory to large-scale, resource-constrained sensing scenarios and incorporates machine learning to handle cases with unknown system models. The proposed approach unifies sparse signal processing, multi-stream statistical decision-making, and resource-constrained optimization to enable simultaneous detection of multiple change points. This framework significantly enhances both applicability and scalability in high-dimensional, heterogeneous, and resource-limited environments while maintaining high detection efficiency.

Existing research on video event detection lacks a unified large-scale dataset and standardized evaluation protocols, hindering fair method comparison and reproducibility. To address this gap, this work proposes the first integrated, three-pronged development framework encompassing dataset construction, performance evaluation, and deployment scenarios. By introducing structured data design, a standardized metric system, and diverse application-oriented modeling, the framework establishes a generalizable paradigm for the field. This approach substantially enhances the fairness of algorithmic comparisons, improves research reproducibility, and supports systematic methodological advancement in video event detection.

This work addresses the lack of deterministic, invertible, and efficient preprocessing mechanisms for high-frequency time-series signals at the edge by proposing a novel edge signal processing engine grounded in number-theoretic covering systems. The engine introduces rational Beatty sequences and Fraenkel’s partition theorem into signal processing algebra, enabling deterministic resampling with bit-level invertible interleaving and deinterleaving. It further features a declarative RQL query language for precise resampling and filtering of constant-rational-interval differential time-series streams. The system employs dependency DAG compilation, slot-based scheduling, and a metadata-augmented, inspectable artifact format, transmitting only deterministic results upstream. The framework fully reproduces the Pan-Tompkins QRS detection pipeline on MIT-BIH ECG data, with all operations expressed within the proposed algebra and formally verified for semantic correctness.

This study addresses the challenge of generating structured event logs from multimodal data such as videos to support business process mining. The authors propose an end-to-end approach that first maps video frames into feature vectors using image embeddings, then performs temporal segmentation via an inter-frame similarity matrix. Subsequently, a generalized few-shot classification method automatically assigns semantic labels to the resulting segments, yielding a timestamped, structured event sequence. This work represents the first integration of image embeddings with few-shot learning for the automatic transformation of raw video into process-mining-ready event logs, thereby overcoming the traditional reliance on pre-structured input data. The method’s effectiveness and practicality are validated through experiments in real-world scenarios.