Performance analysis of deep learning techniques for crop yield prediction

Abstract

India is primarily an agricultural nation, and the success of its farmers is the primary driver of its economy. The choice of crops will be influenced by a variety of factors, considering things like the rate of production, the current market price, and the various policies of the government. In addition to all of the advancements that have been made in the tools and technologies that are used in farming, another factor that plays a key part in it is the availability of valuable and reliable information on a variety of subjects. Due to plant diseases like rust and millet blast, farmers in the agricultural sector are worried about a decline in crop yield and a decline in the quality of the product. As a result, there is a need for the identification and categorization of plant diseases to be carried out using automated means. Even though these techniques have the potential to address the issue of yield prediction, there are a few problems with them: The researchers are unable to construct a linear connection between the original data and the crop production values, either linear or non-linear, and the achievement of these models is primarily reliant on the accuracy of the features extracted. Because of deep reinforcement learning, it is possible to remedy the shortcomings that were previously discussed by providing additional guidance and incentive. This system can convert raw data into numbers that represent crop predictions. In the view of above problems, this thesis attempts the novelty described in stages 1 and 2. newlineStage 1 proposes a Visual Geometry Group (VGG) net classification technique to predict agricultural productivity instead of the modified chick swarm optimization strategy. Stacks of data were supplied into the network one by one. The network architecture is used to build a crop production forecast environment based on the input parameters. The best qualities of the input data are preprocessed using the tweak chick swarm optimization approach, and the optimum output is utilized as input for the clas