The correlation methods are developed for real time natural gas quality analysis. Various statistical models are often used in correlation methods due to the impossibility or high complexity of solving the problems with traditional methods. The choice of a model for the problem under discussion is mostly made by heuristic methods due to the lack of a general algorithm for choosing a model and a variety of both statistical models and architectures of specific models. This paper provides a comparative analysis of the main models that are used to solve the task of natural gas composition analysis. Based on this analysis, the conclusions are drawn about a specific model that is most appropriate to apply to existing data.