Dynamic Context-Aware Event Recognition Based on Markov Logic Networks.

Area of Science:

• Artificial Intelligence
• Computer Science
• Ubiquitous Computing

Background:

• Event recognition in smart spaces is challenging due to data uncertainty and the need for structured information representation.
• Existing methods often struggle with dynamic uncertainty or lack robust structured information handling.
• Dynamic changes in sensing data uncertainty over time pose significant challenges for accurate event recognition.

Purpose of the Study:

• To propose a multi-level information fusion model for sensing data and contextual information for event recognition.
• To develop a method for handling uncertainty in event recognition using Markov logic networks (MLNs).
• To address the challenge of dynamically changing data uncertainty in smart spaces.

Main Methods:

• Developed a multi-level information fusion model integrating sensing data and contextual information.
• Utilized Markov logic networks (MLNs), combining first-order logic (FOL) and probabilistic graphical models (PGMs), to manage uncertainty.
• Introduced an algorithm for updating formula weights in MLNs to adapt to data dynamics.

Main Results:

• The proposed approach effectively recognizes events by fusing uncertain data and contextual information using MLNs.
• The method demonstrates superior performance compared to traditional MLNs in handling dynamic data.
• Experiments on two diverse datasets validated the approach's effectiveness and robustness.

Conclusions:

• The MLN-based multi-level fusion model offers an effective solution for event recognition in dynamic smart spaces.
• The approach successfully addresses the limitations of existing methods in handling uncertainty and structured information.
• The developed algorithm for updating formula weights enhances the model's adaptability to changing data environments.