Water Resource Security Evaluation Based on Improved SVM Model and Indicator System Construction

Artificial intelligence, a pivotal technology spearheading a fresh wave of technological advancements and industrial evolution, is integrating into every corner of society at an unprecedented speed. To evaluate the security of water resources, a water resources security evaluation system grounded on the water poverty index theory was developed, and a support vector machine model improved by particle swarm optimization algorithm was also constructed. In order to better utilize the effectiveness of particle swarm optimization algorithm, improvements have been made from three aspects: inertia weight, learning factor, and flight factor, in order to better optimize the parameters of support vector machine models. The results showed that the maximum evaluation accuracy of the improved support vector machine model was $99.04 \%$, which was $11.39 \%, 6.89 \%, 5.93 \%$, and $4.52 \%$ higher than the comparison methods, respectively. Meanwhile, the mean square error and maximum time consumption of the improved support vector machine model are 1.02 and 78 ms, respectively, with an average $F 1$ value of 0.980, both of which are superior to the comparison model. In the selected experimental area, the water resource evaluation values are all below 0.7, indicating that the water resources in this area mainly belong to the less safe and critical safety levels. The improved support vector machine model designed has good performance and can evaluate the safety of water resources, providing corresponding scientific basis for the policy-making of relevant government departments.