Makine öğrenmesi yöntemlerini kullanarak tarım için Nesnelerin İnterneti (IoT) tabanlı sistem geliştirilmesi

Internet of Things (IoT) has become one of the common technologies used in the agriculture field. The prediction of soil temperature and soil moisture in agricultural field has a significant role in the growth and development of plants. Considering this motivation, in this thesis, two case studies were performed for predicting soil temperature and soil moisture on real-world datasets collected by IoT sensors. In the first study, a novel method, named Soil Temperature Ordinal Classification (STOC), which considers the relationships between the class labels (i.e., low, medium, high) during soil temperature prediction was proposed. To prove the effectiveness of the proposed approach, we applied the STOC method using five different machine learning algorithms (decision tree, Naive Bayes, k-nearest neighbors, support vector machines, and random forest) to a real-world data obtained by IoT sensors from 16 stations in three states (Utah, Alabama, and New Mexico) of United States at five soil depths (2, 4, 8, 20, and 40 inches) between the years of 2011 and 2020. In the second study, an intelligent Multi-Output Regression for Soil Moisture Prediction (MOR-SMP) method was implemented for estimating soil moisture at three soil depths (15, 30, and 45 cm). This approach was tested by applying nine different machine learning algorithms on daily values of meteorological and soil data obtained by IoT sensors from Kemalpaşa-Örnekköy station in Izmir, Turkey. The experimental results showed that the proposed STOC and MOR-SMP methods achieved good performance in predicting soil temperature and soil moisture, respectively. Hence, they can be effectively used in the agriculture sector.