Implementing an Intelligent Irrigation System in Oman Using Machine Learning Technique

In this study, machine-learning technique is used to collect environmental data in real time and measures other aspects like soil moisture, temperature, water level, humidity. Water use efficiency has recently increased because to new agricultural techniques, sophisticated sensors as Micro Controller, and Wireless Sensor Network (WSN). Irrigation Fault Detection method based on XGBoost (IFD-XGBoost) Algorithm used as proposed method as robust to missing values imputing them naturally during tree splitting. The Suggested method is optimized for speed, using parallel and distributed computing to make predictions faster compare to other Regression Algorithm and avoid over fitting. Using Min-Max Normalization, features with different sizes and units can affect the efficacy of several machine learning algorithms. By warping the statistical relationship between features, removal of noisy data and error detection can reduce the machine learning models' efficiency. Lasso Regression is a linear model that combines regularization and feature selection. Additionally, less data will be kept in the centralized cloud as a result. Using the IFD-XGBoost Algorithm's ability to handle missing data, learn from past errors, and adjust parameters, our proposed approach would be the best model to implement. It outperform from other algorithm in Predictive accuracy, making it reliable for calculating water needs and forecasting weather. Investigating the implementation of several machine learning models that can offer the best real-world irrigation decision management is the main goal of this article. The comparative analysis to evaluate the IFD-XGBoost method yielded a minimum RMSE of 0.3524, MAE of 0.1832, MAPE of 4.24 with the highest accuracy of 99.2%, precision of 96.2% and AUC-ROC of 95%, indicates when combined, these metrics provide insightful data about each model's ability to anticipate future developments in irrigation systems.