Optimisation of Environmental Risk Assessment Architecture using Artificial Intelligence Techniques
The integration of artificial intelligence techniques is becoming necessary for environmental risk assessment systems and decision-making, particularly under the limitations of individual intelligence techniques. A comprehensive architecture, called Environmental Risk Assessment Architecture (ERAA),...
Saved in:
| Summary: | The integration of artificial intelligence techniques is becoming necessary for environmental risk assessment systems and decision-making, particularly under the limitations of individual intelligence techniques. A comprehensive architecture, called Environmental Risk Assessment Architecture (ERAA), was developed to capitalise the strengths of intelligent computing techniques and compensate for the limitations of individual intelligence techniques. This architecture is based on the combination of three well-known techniques, namely, artificial neural networks, fuzzy logic and genetic algorithm. The proposed architecture is implemented in the form of two models, namely, the neuro-fuzzy risk assessment model and the safe path selection model. Fuzzy arithmetic operations on fuzzy numbers and artificial neural networks with a back-propagation learning algorithm were used to represent the structure of the neuro-fuzzy risk assessment model, whereas genetic algorithms were used to develop the safe path selection model. Two methods were used to validate the proposed architecture, that is, the analytical method was used to validate the neuro-fuzzy risk assessment model and the safe path selection model, whereas the experimental method was used to evaluate the prototype. Results of the neuro-fuzzy risk assessment model were compared with the results obtained using individual intelligence techniques, such as the Mamdani and Sugeno models. By contrast, the results of the safe path selection model were compared with the results obtained using Dijkstra's algorithm and the Floyd-Warshall algorithm. The results obtained using the neuro-fuzzy risk assessment model show that the model exhibits a satisfactory performance in environmental risk assessment and an improvement in results with a difference rate of up to 10.8% compared with the Mamdani and Sugeno models. By contrast, the running time of the safe path selection model (96 µs) is shorter than the running times of Dijkstra's algorithm and the Floyd- Warshall algorithm (150 µs and 184 µs, respectively). The architecture proposed in this research provides the opportunity for combining intelligent computing techniques in a comprehensive architecture. Thus, the proposed architecture can be applied by developers of environmental risk assessment as a tool for developing applications on tracking and environmental risk assessment systems. |
|---|