Accurately predicting crop yields is a crucial concern in agricultural management, and it is an area where precision farming techniques can offer significant advantages. The primary goals of agricultural productivity evaluation are to optimize crop yield while minimizing expenses, while also maintaining a sustainable ecosystem through the utilization of various technology. The utilization of environmental elements and satellite imagery in the analysis of productivity is currently acknowledged as a cutting-edge advancement in technology. A challenging problem in the agricultural industry is the application of machine learning algorithms to enhance crop productivity. Our analysis indicates that Convolutional Neural Networks (CNN) is the most common deep learning algorithm in research, while another frequently used deep learning algorithm is Long-Short Term Memory (LSTM). The primary deep learning algorithms include of Region-Based Convolutional Neural Network, Long-Short Term Memory (LSTM), recurrent neural network (RNN), and Deep Neural Network (DNN) methods. This paper presents a thorough examination of the advantages and difficulties associated with using machine learning to estimate crop yields. It also provides a comprehensive analysis of the existing and future challenges faced by the agricultural industry in this regard. This technology will achieve its intended purpose by addressing new research challenges in yield prediction analysis and constructing a highly efficient model for yield prediction with little computational complexity.