PERCEPTUS: Predictive complex event processing and reasoning for IoT-enabled supply chain

Document Type

Journal Article

Publication Title

Knowledge-Based Systems


Elsevier BV


School of Science




This research was partially supported by Edith Cowan University (ECU), Australia Early Career Researcher Grant — 2018.


Nawaz, F., Janjua, N. K., & Hussain, O. K. (2019). PERCEPTUS: Predictive complex event processing and reasoning for IoT-enabled supply chain. Knowledge-Based Systems, 180, 133-146. Available here


Internet of Things (IoT) is an emerging paradigm that connects various physical sensor devices spread across different locations. IoT-enabled supply chain provides a natural combination to achieve supply chain visibility (SCV) which refers to ability of supply chain partners to collect and analyse distributed supply chain data for the planning and decision support. This data is normally collected and analysed in real-time by a specialized software known as Complex Event Processing (CEP) engines. However, current CEP engines have two well-known limitations. Firstly, current CEP engines are job specific and fail to combine multiple related sensor data streams coming from distributed sources, thereby, supply chain partners are exposed to manage the underlying information heterogeneity. Secondly, these CEP systems do not provide decision support to the supply chain planners when the information about the potential disruptive event is incomplete and/or uncertain. In this paper, a PERCEPTUS framework is proposed to address the above mentioned issues. It, firstly, utilizes semantic annotation process to integrate and annotate events coming from heterogeneous data streams. Secondly, it performs complex event processing to process and correctly interpret annotated complex events. Thirdly, it provides complex event reasoning (by combining logical and probabilistic reasoning) to predict disruption events (such as process failure) under incomplete and/or uncertain information. Finally, the proposed framework is validated using the dataset of a semi-conductor manufacturing process to demonstrate its superiority in terms of accuracy in predicting disruptive events as compared to the baseline approach.



Access Rights

subscription content