水质预测是众多水务相关问题的重要内容之一，通过水质预测，可以发现水质恶化的预兆，方便决策者提前采取措施。依据常见的水质数据，使用基于遗传算法与支持向量机的水质预测模型在实际应用环境下自行适配污染物权重，提高预测准确率。本模型首先使用遗传算法，训练当前数据的特征权重向量，使得权重适配当前预测问题，然后使用该特征权重向量应用于SVM模型训练。在以重庆某污水处理厂数据为对象进行实验后，验证了该模型在实际应用中的可行性，为水质预测提供了一种新思路。

Water quality prediction is one of the important aspects of many water-related issues. Through water quality prediction, we can find signs of water quality deterioration, which facilitates decision-makers to take measures in advance. In this paper, a water quality prediction model based on genetic algorithm and SVM is used to adapt the weight of pollutants in current application to improve the accuracy of prediction on the basis of common water quality data. The model first uses the genetic algorithm to train the feature weight vector of the current data to adapt the weight to the current prediction, and then apply the feature weight vector in the SVM model training. After conducting experiments with a sewage treatment plant in Chongqing, the feasibility of the model in practical application was verified. Our study provides a new idea for water quality prediction.

TP181

重庆市人工智能技术创新重大主题专项（CSTC2017-rgznzdyf-0140）；重庆市技术创新与应用示范重大主题专项项目（CSTC2018JSZX-CYZTZX0178，CSTC2018JSZX-CYZTZX0185）。