Recent advances of multiway data modeling for classification issues

Abstract

Classification has been a challenging analytical problem over time, addressed to solve a myriad of issues in several research fields [1]. Moreover, modern laboratories are multianalytical platforms, where there is normally the possibility of making different analyses on the same sample offering very rich chemical information. This situation is able to generate several thousand variables per sample, and then, the collected data may have two, three, and more modes (dimensional structure) and/or may come from different sources comprising multiblocks of data. If by one way the instrumentation evolves tremendously, it makes to increase new trends in advanced classification procedures to use more sophisticated chemometrics approaches. This progressive growth from both sides in data generation, storage, and treatment has been introduced to the classification in the multiway context [2,3]. In multivariate classification settings, data handling is producing a significant impact on the development of analytical strategies, especially for determining characteristic patterns of analytes of interest in complex matrices, such as those found in environmental, biological, and food samples, among others. In this context, the mixing of analytical signals has arisen to achieve a better knowledge and characterization of a sample revealing certain synergic answers [4,5]. Irrespective of the specific samples under consideration or the classification problem at hand, multiway classification methods can be employed out of necessity to manage highly complex data structures. Moreover, these methods play a crucial role in elevating the data’s organization, thereby enhancing the informational quality of the sample set. This, in turn, contributes to the resolution of classification problems.