Differentiation of Rubber Cup Coagulum Through Machine Learning

Abstract

Abstract A support vector machine classification algorithm was formulated to differentiate rubber cup coagulum according to the type of acid coagulant used. Two classification models were established, a binary classification algorithm and a model that can identify if formic, acetic, sulfuric acid, or no acid was used to induce coagulation. The models were based on the properties of the rubber cup coagulum that are easy to measure, such as tensile strength, water contact angle, and density. The binary classification model, which differentiates the industry-accepted formic acid-coagulated rubber cup coagulum from those which are not, exhibited satisfactory reliability, as evidenced by a 92% overall prediction accuracy and 71.4% cross-validation accuracy. Moreover, it was also determined that the rubber properties density, and water contact angle were important contributors for the classification. Acid-induced rubber coagulation is an important post-harvest process that influences the resulting rubber quality. Thus, the accurate differentiation of the rubber samples is useful for quality assurance purposes, as well as in policy enforcement.