BIG DATA APPLICATION IN THE NEYMAN-PEARSON REGRESSION AND DEEP BERNOULLI AND BOLTZMANN FOR IOT BASED SOIL QUALITY PREDICTION

Abstract

Purpose. The aim of the work is to enhance the soil quality prediction accuracy and time by using feature selection and classification-based model. Background. Soil quality analysis was handled based on the farmer’s first-hand data mining competence and with the world population expected to increase exponentially (i.e., big data), erratic changes in climate have started influencing soil capitulates incorrectly. Big data is employed used to examine the large amount of dataset for soil quality analysis. It is helps to address a lot of new and significant farming decisions and issue. Soil quality analysis depends on fertility of the soil. By soil quality analysis accuracy prediction is very crucial for practical utilization of resources. But, the existing data mining techniques failed to select the correct crop based on the soil and environmental features. Material and methods. The study presents the data mining techniques based on smart and efficient soil quality prediction for agriculture development. In our work, first, linear regression and the Neyman-Pearson correlation-based feature selection model is employed to obtain the computationally efficient and relevant features. Next, an enhanced deep learning model called deep Bernoulli and Boltzmann IoT-based soil quality prediction is designed to classify the complex soil features with better sensitivity and specificity. Results. Experimental results obtained confirmed the performance and reliability of the proposed method. The result evaluations are carried out on the basis of the prediction accuracy, prediction time, sensitivity and specificity. Conclusion. The result shows that the NPR-DBB method achieves better results than the state-of-the-art methods.

Keywords

• big data
• internet of things
• linear regression
• neyman-pearson correlation
• feature selection
• deep bernoulli
• boltzmann soil quality prediction