The feature-preserving map of high-dimensional complex chemical objects using non-linear map integrated with correlative component analysis

Abstract

A novel procedure integrating non-linear map (NLM) with correlative component analysis (CCA) was proposed to obtain the two-dimensional feature map of high-dimensional complex chemical objects that can more concisely and efficiently represent the object classification. Firstly, CCA was employed to identify the most important correlative components (CCs) from the original high-dimensional complex chemical object. Then, NLM maps the first several CCs, which include the most useful classification information of the object, onto a two-dimensional plane, on which the object classification feature is concisely represented. To overcome the deficiencies of NLM, modified NLM (M-NLM) combining a new adaptive mapping error and a novel optimization approach named as select-best and prepotency evolution algorithm (SPEA) was proposed. The value of the weight factor of the adaptive mapping error is determined adaptively according to the relative deviation square between the two mapping points distance and the corresponding original objects distance. The larger the relative deviation square between two distances is, the larger the value of the corresponding weight factor is. The main operators of SPEA are the proposed select-best and prepotency operator (SPO) and the mutation operator. The comparison results demonstrated that the whole performance of SPEA is better than that of genetic algorithm (GA). Finally, a typical example of mapping two classes' natural spearmint essence was employed to verify the effectiveness of the new approach. The feature-preserving map obtained by NLM integrated with CCA can well represent the classification of the original objects and is much better than the map obtained by NLM alone. In addition, the feature-preserving map obtained by M-NLM integrated with CCA is better than that obtained by NLM integrated with CCA.