Multiple linear regression modeling: Prediction of cheese curd dry matter during curd treatment. (July 2019)

• Record Type: Journal Article
• Authors: Kern, Christian; Stefan, Thorsten; Hinrichs, Jörg
• Is Part Of: Food research international. Volume 121(2019)
• Page Start: 471
• Page End: 478
• Publication Date: 2019-07
• Subjects: Multiple linear regression -- Cheese curd -- Dry matter -- Syneresis -- Modeling -- Monte Carlo approach; Food -- Analysis -- Periodicals; Food industry and trade -- Periodicals; Food industry and trade -- Canada -- Periodicals; Food Technology -- Periodicals; Food -- Periodicals; Food-Processing Industry -- Periodicals; Aliments -- Industrie et commerce -- Périodiques; Aliments -- Industrie et commerce -- Canada -- Périodiques; Aliments -- Recherche -- Périodiques; Food industry and trade; Canada; Periodicals; Electronic journals; 664.005
• DOI: 10.1016/j.foodres.2018.11.061
• Languages: English
• ISSNs: 0963-9969

Abstract

Abstract: Cheese curd dry matter determines functional properties and process parameters during cheese manufacture. Dry matter is affected by many internal (milk composition and pre-treatment) and external (cheese process parameters) factors that are not considered in the most common models. The purpose of this study was to consider a large number of multiple linear regression models that use these internal and external factors as predictor variables, and select the most suitable of these models in order to predict the cheese curd dry matter during curd treatment. Dry matter ( DM exp, nat ) was experimentally determined to create a native data set ( n = 1013) for fitting the regression model. Dry matter was affected by curd treatment time ( CTT ), curd treatment temperature ( ϑ ), pH-value ( pH ), curd grain size ( CGS ), fat level ( f ) and degree of microfiltration( i ). A large number of empirical regression models, organized into three different groups, depending on the predictors used, were developed on basis of DM exp, nat . A Monte Carlo approach was used to select the optimal model, taking into account the value of Akaike's information criterion (AICc) and the coefficient of determination (R 2 ) of each model. The best models were further analyzed to check for potential bias and to verify that the model assumptions were met. We considered one model of group G2 with 11 terms to most closely fit the aforementioned criteria (native data set; R 2 = 95.55). This model was successfully validated by an independent validation data set ( n = 120; R 2 = 91.95). Graphical abstract: Unlabelled Image Highlights: Examination of dry matter during curd treatment by modified Dynamic System Model Development of MLR models based on experimental native data set ( n = 1013) MLR models based on milk composition, pretreatment and cheese processing parameters Validation of selected models (n = 120)