<![CDATA[Modern agriculture relies heavily on technology, especially in irrigation management and crop watering. Several previous studies have applied field data-based predictive techniques to improve crop yields. This research aims to develop a prediction system for optimal watering time in plantations and agriculture using a machine learning approach. The rigorous methodology includes data capture, pre-processing, model evaluation and testing, validation, and visualization. High accuracy demonstrates the system's reliability in determining optimal watering needs to improve resource efficiency and crop yields in agriculture. The data obtained from the automatic weather station (AWS) via thingsboard is processed sequentially, starting from data retrieval in json format using postman to transformation into csv files with proper timestamp adjustment. The pre-processing stage includes data cleaning, variable selection, data integration, and generating a clean dataset. In the evaluation stage, the dataset is divided into training data and test data, with the application and comparison of logistic regression, random forest and decision tree models applied as classifiers. Furthermore, the validation and results stage includes prediction, performance testing using the confusion matrix, and visualization of prediction results in the form of text and icons that aim to increase interpetability for users through Google Collaboratory. The results of this research provide an overview of the optimal watering time based on the dataset from the automatic weather station. Further analysis shows that the implementation of machine learning models significantly improves the prediction accuracy, proving the effectiveness of the system in providing more precise watering time recommendations to increase agricultural productivity. The main objective is to develop a machine learning-based watering time prediction system using data from the automatic weather station and evaluate various classifier algorithms to select the best model.]]>
