Leveraging Bi-LSTM for Data-Driven Crop Selection and Yield Optimization in Precision Farming

Agriculture is essential for feeding the global population, and in light of climate change and the limited resources available, it is becoming more crucial to choose the appropriate crops and estimate the amount of food that they will provide. This study presents a framework for deep learning that makes use of Bidirectional Long Short-Term Memory (Bi-LSTM) networks. The purpose of this framework is to improve decision-making in crop selection and to properly estimate crop yields. Historical information on farming is included into the model. This includes information on how crops have performed in the past, the weather (including temperature, rainfall, and humidity), and the qualities of the soil. Due to the fact that it is able to discover temporal correlations in both directions, Bi-LSTM is superior to both standard machine learning models and unidirectional LSTM approaches. The system that is being suggested is trained and assessed using agricultural information that is available to the general public. It has the ability to accurately forecast how much of a crop will grow and determine which kind of crops are most suitable for certain locations. Both the training and testing of the Bi-LSTM model were carried out with the help of the Agricultural Crop Yield dataset. The performance is compared against a number of other approaches, such as Linear Regression, Random Forest, and a simple LSTM, in order to determine the Mean Absolute Error, Root Mean Squared Error, R2 score, and Mean Absolute Percentage Error. An MAE of 0.32, an RMSE of 0.47, and an R2 score of 0.91 are all characteristics of the Bi-LSTM model, which is the most effective. For the purpose of maximizing resource use and increasing agricultural output via the utilization of data-driven insights, the results of this research may be of assistance to farmers, agronomists, and policymakers.