An intelligent mushroom strain selection model based on their quality characteristics

Abstract

The great versatility of mushroom production and the significant nutritional and medicinal properties of the crop make it a highly attractive product that is in continuous expansion around the world. However, its quality can be affected the combination of a large number of evaluable variables that are essential to take into account. Thus, the aim of this work was to build an intelligent model for the prediction of mushroom strains through the development of neural networks (ANNs) that takes into account the control of data processing times, with the use of the minimum possible number of parameters that define their quality control and subsequent selection. In addition, a user-friendly and intuitive graphical interface has been generated that shows the selection of the appropriate mushroom strain and may be useful for potential end-users in this field. For this purpose, 7 mushroom strains (Agaricus bisporus) defined by a total of 27 quality parameters were used (texture, colour, etc.). The results showed that, in the analysis of individual parameter combinations (Rt), the best overall accuracy achieved (OAA) was 52.43%, reaching 81.30% with the combination of four parameters (dry matter (%), crude protein (Nx4. 38. %), Fb and Wt) and 94.32% with 9 parameters (Cap diameter (mm), dry matter (%), crude protein (Nx4.38. %), ΔE, browning index (BI), Fb, Wb = Fb x Db/2, Wr and Rt). The development of this model allows for the identification of some of the most important commercial white hybrid strains of high-yielding mushrooms, while also being a useful tool for the selection of the most important parameters of interest as regards the quality and benefits of this product for the consumer.