Machine learning predictive model for evaluating the cooking characteristics of moisture conditioned and infrared heated cowpea

Abstract

AbstractCowpea is widely grown and consumed in sub-Saharan Africa because of its low cost and high mineral, protein, and other nutritional content. Nonetheless, cooking it takes considerable time, and there have been attempts on techniques for speeding up the cooking process without compromising its nutritious value. Infrared heating has recently been proposed as a viable way of preparing instantized cowpea grains that take a short amount of time to cook while maintaining desired sensory characteristics. Despite this, only a few studies have shown the impact of moisture, temperature, and cooking time on cooking characteristics such as bulk density, water absorption (WABS), and the pectin solubility of infrared heated cowpea precooked using this technology. Artificial neural network was used as a machine learning tool to study the effect of a prediction model on the infrared heating performance and cooking characteristics of precooked cowpea seeds. With R values of 0.987, 0.991, and 0.938 for the bulk density, WABS, and pectin solubility, respectively, the prediction model created in this study utilizing an artificial neural network (a type of machine learning) outperformed the traditional linear, 2-factor interaction, and quadratic models.