Research On Clustering Method To Analyze The Functional Principal Components Based On The Energy Consumption In The Region

Cluster analysis is often used as a practical analysis technique. However, traditional clustering methods are sometimes affected by collinearity between the sample indicators. Principal component analysis is an effective classical dimension reduction method and can effectively solve the problem of collinearity and significantly reduce the effective indicators. Samples can be clustered basing on the principal component scores.Since the amount of actual data in time series data is very large, also the noise, missing data and the abnormal data typically present in the data. To solve this problem, we used functional data analysis techniques to deal with data. Combing principal component clustering method above, we can get the clustering algorithm based on functional principal component scores. Firstly, smoothing coefficients are selected by GCV criterion, and multiple sets of time series data (panel data) are transformed into a number of functional data; Secondly, do the functional principal components analysis(FPCA) and detennine the number of functional principal components based on the cumulative contribution rate;Thirdly, collect the discrete principal component scores of each functional data after dimension reduction;Finally, cluster samples by conventional clustering methods on principal component scores.Clustering method proposed in this paper is applied to energy consumption and energy consumption per capita in the Asia-Pacific region. The results show that the method is valid, and the results in line with the actual situation. The analysis of the energy situation in countries or regions disputes some guidance to facilitate the country in the future to make the appropriate response. In addition, the functional data analysis has many advantages compared with traditional statistical method, and will be applied and developed in more and more areas.