Reviewer #1: (Previous Version)
This paper presents a hybrid forecasting approach to predict ozone concentrations in the Semi-arid regions of China. It is based on association rule technique in order to find correlations between meteorological data and ozone concentrations, chaotic particle swarm optimization algorithm and multilayer perceptron.  The proposed hybrid model shows the promising results in comparison with simple multilayer perceptron and regression approach.

COMMENTS:
 The data mining technique (section 2.1) used by authors is described quite blurry. The BP neural network is described with some mistakes. So, for instance, the authors wrote the neural network consist of 4 layer (input, two hidden and output layer). However the figure 1 contains only 3 layers. The equation (5) is not true (the bracket is passed). The comparison is performed only with simple models, namely BP neural network and regression approach. I think it is not enough. The idea of using hybrid approach for forecasting is not innovative. Though, from practical point of view the results sound good.