Crop Yields Prediction in Bangladesh: A Hybrid Machine Learning and DNN Approach

In todays world, machine learning and artificial intelligence are gaining popularity. In today’s digital age, many tasks are being transformed through artificial intelligence and machine learning. Agriculture, as a vital component of human sustenance, remains The principal source of livelihood for many. Bangladesh is an agrarian economy due to its geographical location. Bangladeshs economy is heavily dependent on agriculture as it contributes a lot to the countrys GDP. Even though artificial intelligence and machine learning are being used to digitize every industry in this day and age, the agricultural sector is falling behind. In Bangladesh agricultural work is mostly done in old ways and manually. From crop selection to fertilizer selection, all work is done by farmers manually based on their experience. In addition, agricultural institutions in Bangladesh use manual data, which is not a proper solution to accurately select crops and yield prediction. To address these problems, there is a need for modern and innovative approaches to yield prediction and crop selection. This paper demonstrates the most efficient and cost-efficient approach to predicting crop yield utilizing deep neural networks (DNN) and machine learning techniques. Artificial Neural Network and Machine Learning are considered powerful tools for modeling and prediction. The effectiveness of the algorithms that were assessed in this study are Deep Neural Network- Gradient Boosting Regressor(DNN-GBR),Support vector machine(SVR), K-Nearest Neighbors Algorithm. The goal of this algorithm is to improve output and prediction. The performance of the proposed hybrid DNN-GBR model was evaluated using MSE and R-squared (coefficient of determination) metrics. In this paper, we have suggested using the hybrid DNN-GBR model for agricultural yield prediction and crop selection.