Identification of Suitable Complex Machine Learning Algorithms for Amylose Content Prediction in Rice with an IoT-based Colorimetric Sensor

AMYLOSE CONTENT PREDICTION IN RICE WITH AN IOT-BASED SENSOR

Authors

  • Shrinivas Deshpande ICAR-Krishi Vigyan Kendra, Kandali, Hassan 573 217, UAS, GKVK, Bangalore, Karnataka, India
  • Udaykumar Nidoni Dept. of Processing and Food Engineering, CAE, UAS, Raichur 584 104, Karnataka, India
  • Rahul Patil Dept. of Soil & Water Conservation Engineering, CAE, UAS, Raichur 584 104, Karnataka, India
  • Sharanagouda Hiregoudar Dept. of Processing and Food Engineering, CAE, UAS, Raichur 584 104, Karnataka, India
  • Ramappa K T Dept. of Processing and Food Engineering, CAE, UAS, Raichur 584 104, Karnataka, India
  • Devanand Maski Dept. of Renewable Energy Engineering, CAE, UAS, Raichur 584 104, Karnataka, India
  • Nagaraj Naik Pesticide Residue and Food Quality Analysis Laboratory, CAE, UAS, Raichur 584 104, Karnataka, India

DOI:

https://doi.org/10.56042/jsir.v83i1.2458

Keywords:

Ageing of rice, Amylose sensor, IoT device, Mathematical modeling, Rice quality

Abstract

Rice ageing is a complex phenomenon that is hard to investigate thoroughly. Many physicochemical qualities change gradually because of moisture content and storage temperature. Among these characteristics, amylose quantity is particularly essential, and most indexes rely on it. To address these challenges, various gadgets, IoT, ICT, AI and predictive technologies are frequently applied in diagnostic procedures. This study evaluated AdaBoost, Artificial neural network (ANN), k-Nearest Neighbour classifier (KNN), Decision tree, Logistic regression, Support Vector Machine (SVM), and Random forest classifiers to categorize distinct quantities of amylose using slope data gathered from the novel colorimetric amylose sensor. The random forest approach had greater coefficients and precision ratings of 0.85 for the slope dataset, followed by the decision tree, ANN, KNN, AdaBoost, logistic regression, and support vector algorithms, which had precision scores of 0.83, 0.81, 0.80, 0.29, 0.18, and 0.18, respectively, based on the efficiency of the tested learning models. The random forest model was shown to be promising in forecasting the various classes of amylose based on the data.

Downloads

Published

2024-01-11