Composition-based statistical model for predicting CO2 solubility in modified atmosphere packaging application

Carbon dioxide (CO2) is an important gas used in modified atmosphere packaging of non-respiring foods where it solubilizes into the aqueous and lipid phases food exerts antimicrobial effect. Prediction CO2 solubility within thus paramount importance to anticipate its benefit on preservation. In present study, machine learning algorithms were applied a set 362 values solubilities collected from scientific literature tentatively predict as function composition (water, protein, fat salt content) temperature. The best option kept was random forest algorithm that four case studies (ham, salmon, cheese pâté) further input parameters MAP’ OPT tool, predicting evolution headspace composition. Predicted succeeded representing dynamic time studies.