Applied Stochastic Models for Business and Industry has recently published a special issue. The issue, Data Science in Process Industries (issue 38:5) includes nine papers that present contributions on the application of  various data science methodologies to effectively address relevant problems faced by modern Process Industries. It was launched in the scope of the ENBIS 2021 Online Spring Meeting dedicated to the same theme. ENBIS, the European Network of Business and Industrial Statistics, has been an important international platform for the exchange of knowledge, scientific results, and experiences in the field of applied statistics and data analytics for over 20 years. A portion of the foreword by guest editors Marco S. Reis and Nikolaus Haselgruber is included below: 

In “Self-supervised cross validation using data generation structure“, Kenett et al. find a gap in the current “automated” cross-validation methods and propose a new one that considers the data structure and goals of the analysis. The new approach is called befitting cross-validation. In contrast to the conventional automated approaches, it requires an assessment with domain experts’ inputs to secure the quality of the analysis.

In “Spatial correction of low-cost sensors observations for fusion of air quality measurement“, Bobbia et al. address the problem of fusing measurements collected from sensors of different quality, to construct the air quality map. Predictions are corrected through kriging of the residuals instead of using the pollutant concentration level. 

In “A Bayesian data modelling framework for chemical processes using adaptive sequential design with Gaussian process regression” (OPEN ACCESS), Fleming et al. use Gaussian processes to efficiently capture the nonlinearities of chemical processes, using data collected through a Bayesian adaptive sequential design.

In “Separable spatio-temporal kriging for fast virtual sensing” (OPEN ACCESS), Lambardi di San Miniato et al. investigate separable spatio-temporal kriging for environmental monitoring. Their approach involves a spatial as well as a temporal model, which can be handled separately under a composite likelihood approach. The model was applied to develop a spatio-temporal prediction rule for the temperature in an office room.

In “Designing acceptance single sampling plans: An optimization-based approach under generalized beta distribution” (OPEN ACCESS), Facchinetti et al. design Bayesian acceptance sampling plans under the use of a generalized beta prior distribution. The optimal sampling plan was obtained by minimizing the expected total cost of quality.

In “Quick-switch inspection scheme based on the overall process capability index for modern industrial web-based processing environment“, Wang et al. present a quick-switch inspection scheme for the consideration of multiple quality characteristics based on the overall process capability. They provide a web application to support the automation process of lot deposition correspondingly and applied the inspection scheme to ultra-mini chip resistors. 

In “Neural network based control charting for multiple stream processes with an application to HVAC systems in passenger railway vehicles” (OPEN ACCESS), Lepore et al. analyze a multi-stream process and develop a neural network to combine the outcomes of the different streams into a single quantity that reflects the status of the process. 

In “Non-parametric local capability indices for industrial planar manufacts: An application to the etching phase in the microelectronic industry“(OPEN ACCESS), Borgoni et al. address process capability and propose a methodology based on additive quantile regression with inference in an adopted Bayesian framework in order to estimate the local process capability of planar surfaces. The authors exemplify the approach with an application to the dry etching phase in semiconductor fabrication.

Finally, in “Defective products management in a furniture production company: A data mining approach“, Ersöz et al. compare different data mining approaches in order to identify the root causes of defective products. They investigate artificial neural networks, classification and regression trees based on data from furniture production. 

Future IoT and Industry 4.0 need these data science methodologies presented in this special issue to develop further and be successful. How to take advantage of them, without relegating to a secondary role or even to obliviation the classical methods that were the workforce for addressing many problems in the industry over the years, is a challenge and concern that most of us have on a daily basis. This special issue can spike the readership’s curiosity on different corners of the spectra of methodologies, processes, and problems; to bring their thoughts, concerns and solutions to forthcoming ENBIS events.

ASMBI – Applied Stochastic Models in Business and Industry (formerly Applied Stochastic Models and Data Analysis) was first published in 1985, publishing contributions in the interface between stochastic modelling, data analysis and their applications in business, finance, insurance, management and production. In 2007 ASMBI became the official journal of the International Society for Business and Industrial Statistics.