Employing machine learning and Internet of things based real time NPK fertilizer prediction for Cassava crop in Rwanda

Abstract

In the Agriculture sector the pressure is increased immensely due to the rise in population. In this present year we mainly witness a move from traditional methods were used in the agriculture to the advanced technology. Machine learning and IoT technologies have transformed the quality and the yield production. In fact IoT helps to collect real time data from the field via sensing technique and machine learning analyzes those data from sensors for generating information that will help in business growth. Knowing a suitable crop and soil fertilizer option is one of the things that can make a farmer more productive and help him to avoid losses. Soil fertilization activities contribute a lot in crops production volume. However, if the quantity of soil composition (fertilizer) is not controlled and maintained consistent, this may lead to less crop production volume. Choosing the appropriate crop type and the corresponding quantity of soil fertilizer is one of the measures to be taken prior for preventing the inferior quality and less quantity of the crop production. Thus, the measurement of soil nutrients is greatly required for better plant growth and fertilization. Therefore, temperature, soil moisture, Nitrogen, phosphorus pentoxide, potassium oxide, pH level are among the parameters commonly measured to monitor the cassava crop as they are the ones mostly important and informative soil parameters to determine the soil fertility. This research will focus on “Employing Machine Learning and Internet of Things Based Real Time Fertilizer Prediction for Cassava crop in Rwanda” by using a machine learning (ML) algorithm to build a model which may help the farmers to predict the cassava fertilizer components. Through this research, different parameters are respectively controlled by a network of sensors such as; temperature sensors, soil moisture sensors, soil nutrients sensors, and PH sensors then the data corresponding to these parameters will be feed to the different machine learning algorithms such as Linear Regression, Random Forest, Gradient Boosting, Random Forest, K-Nearest Neighbors and Decision Tree will be tested for optimizing the prediction accuracy using python programming packages. These algorithms have been selected because they are mostly used in classification and regression problems.