Analysis of pollutant levels in central Hong Kong applying neural network method with particle swarm optimization

Air pollution has emerged as an imminent issue in modern society. Prediction of pollutant levels is an important research topic in atmospheric environment today. For fulfilling such prediction, the use of neural network (NN), and in particular the multi-layer perceptrons, has presented to be a cost-effective technique superior to traditional statistical methods. But their training, usually with back-propagation (BP) algorithm or other gradient algorithms, is often with certain drawbacks, such as: 1) very slow convergence, and 2) easily getting stuck in a local minimum. In this paper, a newly developed method, particle swarm optimization (PSO) model, is adopted to train perceptrons, to predict pollutant levels, and as a result, a PSO-based neural network approach is presented. The approach is demonstrated to be feasible and effective by predicting some real air-quality problems.