Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

GNN-XAR: A Graph Neural Network for Explainable Activity Recognition in Smart Homes

Michele Fiori, Davide Mor, Gabriele Civitarese, Claudio Bettini

TL;DR

GNN-XAR addresses the need for explainable activity recognition in smart homes by introducing a dynamic graph-based GNN framework. It constructs time-window graphs from environmental sensor events, processes them with a Graph Convolutional Network, and uses an adapted GNNExplainer to identify influential nodes and arcs, translating these into natural language explanations. The approach demonstrates that explainable GNNs can outperform state-of-the-art explainable HAR methods in both classification accuracy (slightly improved F1) and explanation quality, as evidenced by evaluations on CASAS Milan and Aruba. This work has practical implications for trusted, transparent HAR in ambient assisted living, enabling clinicians and caregivers to understand model decisions and justify healthcare interventions.

Abstract

Sensor-based Human Activity Recognition (HAR) in smart home environments is crucial for several applications, especially in the healthcare domain. The majority of the existing approaches leverage deep learning models. While these approaches are effective, the rationale behind their outputs is opaque. Recently, eXplainable Artificial Intelligence (XAI) approaches emerged to provide intuitive explanations to the output of HAR models. To the best of our knowledge, these approaches leverage classic deep models like CNNs or RNNs. Recently, Graph Neural Networks (GNNs) proved to be effective for sensor-based HAR. However, existing approaches are not designed with explainability in mind. In this work, we propose the first explainable Graph Neural Network explicitly designed for smart home HAR. Our results on two public datasets show that this approach provides better explanations than state-of-the-art methods while also slightly improving the recognition rate.

GNN-XAR: A Graph Neural Network for Explainable Activity Recognition in Smart Homes

TL;DR

GNN-XAR addresses the need for explainable activity recognition in smart homes by introducing a dynamic graph-based GNN framework. It constructs time-window graphs from environmental sensor events, processes them with a Graph Convolutional Network, and uses an adapted GNNExplainer to identify influential nodes and arcs, translating these into natural language explanations. The approach demonstrates that explainable GNNs can outperform state-of-the-art explainable HAR methods in both classification accuracy (slightly improved F1) and explanation quality, as evidenced by evaluations on CASAS Milan and Aruba. This work has practical implications for trusted, transparent HAR in ambient assisted living, enabling clinicians and caregivers to understand model decisions and justify healthcare interventions.

Abstract

Sensor-based Human Activity Recognition (HAR) in smart home environments is crucial for several applications, especially in the healthcare domain. The majority of the existing approaches leverage deep learning models. While these approaches are effective, the rationale behind their outputs is opaque. Recently, eXplainable Artificial Intelligence (XAI) approaches emerged to provide intuitive explanations to the output of HAR models. To the best of our knowledge, these approaches leverage classic deep models like CNNs or RNNs. Recently, Graph Neural Networks (GNNs) proved to be effective for sensor-based HAR. However, existing approaches are not designed with explainability in mind. In this work, we propose the first explainable Graph Neural Network explicitly designed for smart home HAR. Our results on two public datasets show that this approach provides better explanations than state-of-the-art methods while also slightly improving the recognition rate.

Paper Structure

This paper contains 27 sections, 11 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. GNN-based methods for HAR
  4. XAI methods for Human Activity Recognition
  5. GNN-XAR under the hood
  6. Overall Architecture
  7. Graph Construction
  8. Graph Neural Network
  9. Message Passing
  10. Graph Pooling and Classification
  11. Explainer
  12. Generating explanations in natural language
  13. Experimental Evaluation
  14. Datasets
  15. CASAS Milan
  16. ...and 12 more sections

Figures (11)

  • Figure 1: Overall architecture of GNN-XAR.
  • Figure 4: Example of a time window. For the magnetic sensor, the black lines represent the events. For the movement sensors, the black regions represent the active states
  • Figure 5: The directed graph computed from the sensor window in Figure \ref{['fig:EX1']}.
  • Figure 6: The model architecture of GNN-XAR.
  • Figure 7: Output of the original GNNExplainer limited to arcs importance. The thickness of each arrow represents the importance values on the arcs.
  • ...and 6 more figures