Machine Learning-Based Smart Agriculture System

Global climate change and increasing food demand necessitate resource efficiency in agricultural production. Therefore, the integration of IoT and AI technologies into modern greenhouse cultivation becomes a critical activity. However, the raw data obtained from sensors within greenhouse environments may be inconsistent due to environmental noise, hardware drift, and transmission errors. These may lead to erroneous decisions by automation systems and result in resource wastage. In this project, a hybrid decision support system that combines the Kalman Filter with machine learning algorithms for improving the accuracy of greenhouse microclimate parameters like temperature, humidity, light, and CO2 to optimize resource usage was proposed and developed. For this project, instead of directly sending the gathered environmental data with Arduino-based sensor nodes via LoRa communication infrastructure to the control mechanism, this data was passed through a data refinement layer. The instantaneous noise in measurements was minimized, and data stability was ensured using the Kalman Filter, as has been suggested in sensor data fusion approaches. Refined data were fed as input to an AI model trained for generating control commands for irrigation, ventilation, and lighting in optimizing the plant growth pattern. This approach is to reduce plant stress and increase productivity by offering greater precision than the traditional water-energy management technique of operating either above or below given thresholds. The results of this study indicate that the proposed system improves sustainability and automation reliability in agricultural production, even in the event of data loss or sensor error.