MACHINE LEARNING MODEL FOR MULTIPARAMETRIC CROP TYPE AND YIELD FORECASTING

The multidisciplinary process of "smart farming" entails forecasting the kind and quality of plant that will produce the highest yield in a specific geographic area. This entails accounting for elements like soil, climate, and other environmental elements. Deep learning models are used by the great majority of intelligent farming systems to achieve this, however their accuracy, scalability, and real-time deployment capabilities are all lacking. This is due to the fact that most deep learning models need a lot of training data, which lengthens the time needed before they can be put into use. Furthermore, a model that has been trained on one kind of crop may not be as applicable to other types of crops. To improve the usability and practicality of deep learning models, this paper suggests an improved machine learning model for multiparametric crop type and yield prediction. The proposed model fuses the 'you look only once' (YoLo) model with the VGGNet-19 model, the Inception Net model, the Xception Net model, and the GoogleNet model.This makes processing at a high speed possible. The model uses multiparametric data to predict the maximum yield that can be produced from a crop and the kind of crop that can be grown under a specific set of environmental conditions. This includes the use of plant imagery, soil parameters, weather data, temporal geographical data, and nutrient information. The suggested model has a high degree of accuracy when tested against a variety of crop and soil types; it received scores of 98.7% and 97.6%, respectively, for crop-type prediction and yield prediction. The suggested model performs, on average, 8% better in terms of accuracy, 6% better in terms of precision, 3% better in terms of recall, and 6.5% better in terms of area under the curve (AUC) when compared to other cutting-edge models. The suggested model also demonstrates a 9% decrease in latency, which qualifies it for high-speed real-time deployments. Additionally, this text provides a number of case studies that were conducted in order to validate the model's performance and makes some recommendations for future research in order to achieve even higher application-specific performance levels.