Gated recurrent unit model for assessment of food quality based on E-nose sensors supported with one-way analysis of variance

Abstract

Ensuring the quality of global food supplies has emerged as a significant challenge in recent times. Overseeing perishable items' excellence, freshness, and longevity poses considerable intricacy. A special kind of system established on electronic scent detection systems has been engaged for quality assessment. Recent advancements have concentrated on integrating electronic scent detection systems with machine learning (ML) and deep learning (DL), which comprise encouraging remedies to meet these hurdles. Mainly, this investigation aims to present a pioneering strategy for addressing this issue by binding DL with electronic olfaction technology. Gated recurrent units (GRU) were used for classification actions. The research entails examining from the literature a benchmark dataset acquired from electronic noses (E-noses) across beef cuts. These cuts are allocated into four classes: i) outstanding, ii) satisfactory, iii) passable, and iv) spoiled, depending on their quality. The proposed model, exploiting a GRU for classification tasks, was developed with active dataset attributes identified over the analysis of variance (ANOVA) feature selection method. As a consequence, three key features were selected and employed for the classification process, such as MQ5, MQ137, and total volatile content (TVC). Experimental outcomes demonstrate an impressive classification accuracy of 99.77%, accomplished by the proposed model across further literature models.