Chemometrics: From Data Preprocessing to Fog Computing

Introduction

The rise of several hyphenated analytical techniques and their applications have led to the development of various chemometric methods in order to come up with meaningful information from the data generated by these instruments (Kumar, Bansal, Sarma & Rawal, 2014). The applications of chemometrics are extensive, ranging from multicomponent analysis in spectroscopy to the areas of bioinformatics, molecular genetics, and genetic epidemiology in recent years (Dumancas, 2012; Dumancas et. al., 2014; Dumancas et. al., 2015).

One of the areas of chemometrics is in Process Analytical Technology (PAT). PAT is an initative designed to improve the efficiencies of both the manufacturing and regulatory processes by utilizing an integrated approach to quality analysis. One of the central cores of PAT is data analysis (Willis, 2004), which encompasses various chemometric tools. Thus, the advances that are now visible in PAT using chemometrics involve both the use of analytical instrumentation and mathematical methods for multivariate data analysis (Bogomolov, 2011; Dubrovkin, 2014; Kessler, 2013; Pomerantsev & Rodionova, 2012). The primary driving forces that led to the success of PAT would be the development of novel analytical methods and the continuous expansion of their applications (Dubrovkin, 2014).

As mentioned earlier, there has been rapid growth of data due to the rise of various analytical instruments. However, the main challenge comes from processing these data in a facile manner. In certain cases, multiple sensors are studying the same variables or compounds of interest. As such, the process of Data Fusion, a subclass of Chemometrics, is now considered an important topic (Esteban et. al., 2005; Ovalles & Rechsteiner, Jr., 2015). Multi-sensor Data Fusion is a tool used to combine the data from multiple sensors with the overall goal of providing a more reliable and accurate output (Castanedo, 2013; Rashinka & Krushnasamy, 2017). The Joint Directors of Laboratories (JDL) defines data fusion as a “multi-level and multifaceted process handling the automatic detection, association, correlation, estimation, and combination of data and information from various sources” (Steinberg et. al., 1999). The corresponding informational models emanating from data fusion should simulate extremely complex problems by fitting to the massive amount of empirical semi-structured and unstructured data (Isaeva et. al., 2012). Consequently, the algorithmic support and the interface of a computerized analytical system (often with limited computer resources) should be adjustable to systems with features of new types. Such challenge arising from analytical information management led to several new perspectives and solutions, such as the concept of Cloud Computing, all of which are part of the development of “Big Data Approach” (BDA) (Dubrovkin, 2014). Cloud Computing can simply be defined as the operation of computer power or storage on remote servers by means of a network. Using the Cloud, very high-level services with high computational power is now possible. Fog computing, on the other hand, constitutes the layer below Cloud computing in connected Things (Paret and Huon, 2017). In other words, Fog Computing is an extension of the Cloud Computing paradigm to the edge of the network, thus enabling a new breed of applications and services (Bonomi et al, 2012).

In this manuscript, the major aspects of Big Data utilization and processing in Analytical Chemistry (Chemometrics), specifically some commonly used algorithmic and instrumental techniques and aspects of computerized analytical systems, will be discussed. An interesting discussion will also be the role of fog computing in chemometrics.