A Novel Determination of Soluble NPK Nutrient Measurement System Using K-means and SOM Methods Compared with UV Visible Spectrophotometer

Technologies are currently being used in smart farms to improve productivity. Agricultural innovation is also widespread. Therefore, the development of Internet of Things (IoTs) technologies and devices to collect data is important. Finding essential nutrients for growth such as Nitrogen (N), Phosphorus (P) and Potassium (K) are important factors. Competitiveness is the important key for growing vertically or indoors. However, soil testing can be expensive, complicated and time consuming. The development of devices that can effectively measure nutrient NPK levels could be used over time. This paper presents a suitable clustering using the K-means method and a forecasting method using the Self-Organizing Map (SOM) method to enhance the device's ability to analyze nutrients in real time. A total of 50,688 samples were tested using the K-Means clustering algorithm, and the sample data were also analyzed. A SOM was prepared to predict the accuracy of the analysis. A total of 11,486 samples were tested and analyzed, then the results of the soluble NPK nutrient monitoring system were taken. It has an efficiency of 84.15% compared to using a UV Visible spectrophotometer, allowing data to be predicted during nutrient solubility changes and to operate according to the proposed system. Therefore, the unsupervised machine learning will improve the measurement system of future cultivation systems.