Indonesian Journal of Electrical Engineering and Computer Science Vol. 30, No. 1, April 2023, pp. 518~527 ISSN: 2502-4752, DOI: 10.11591/ijeecs.v30.i1.pp518-527  518 Journal homepage: http://ijeecs.iaescore.com Prediction on field crops yield based on analysis of deep learning model Iniyan Shanmugam, Jebakumar Rethnaraj, Srinivasan Rajendran, Senthilraja Manickam Department of Computing Technologies, SRM Institute of Science and Technology, Chennai, India Article Info ABSTRACT Article history: Received Aug 20, 2022 Revised Dec 2, 2022 Accepted Dec 10, 2022 Agriculture has a key role in the overall economic development of the country. Climate change, irregular rainfall, changes in the nutrient content of the soil, and other environmental changes are seen as a severe problem in crop yield prediction. Using deep learning (DL) models that incorporate multiple factors can be viewed as an essential strategy for attaining accurate and effective solutions to this issue. The crop yield can be predicted using yield data obtained from a historical source that includes information about the weather, soil nutrient content, soil type, the season in which the crop was grown, and its yield. In order to train the model and achieve high accuracy, a large set of data including multiple factors would be required. This research aims to forecast the yield of a certain crop using long short-term memory (LSTM) time series analysis and the information currently available. The data used to construct the models was obtained from a reputable source and contains correct numbers. Before growing a crop that has been sown on a piece of agricultural land, the yield prediction utilizing advanced methodologies can assist farmers predict the yield of a specific crop. Keywords: Crop yield prediction Dataset Linear regression Long short-term memory Time series analysis This is an open access article under the CC BY-SA license. Corresponding Author: Iniyan Shanmugam Department of Computing Technologies, SRM Institute of Science and Technology Kattankuathur, Chennai, Tamil Nadu, India Email: iniyans@srmist.edu.in 1. INTRODUCTION Crop yield is considered as a difficult and complex trait that can be determined by various factors which are the soil type, physical environment, and the changes occurring in it. To predict crop yield continuously, it needs a lot of information that can be used to investigate the relation between obtained crop yield and other parameters. In order to understand these dependencies, it is essential to know about the extensive datasets as well as the algorithms that might be incorporated [1]. Time can be considered as an essential parameter that might be taken into consideration if the model has to forecast anything, be it the expected stock price, the amount of crop yield or the amount of rainfall that might fall at a particular location. For instance, a condition can be quite fascinating wherein the model can predict the time which would be having the most consumption in electricity. It may allow us control the consumption expenses so that we’ll be able to produce more electricity during the peak time and could even save resources when not needed [2]. To understand time series, we can consider it as a simple continuous data that are arranged based on the time. While implementing this method, the role of time is generally considered as a non-dependent entity whose main objective generally emphasizes on forecasting. By using time series analysis one can predict the future outcomes on the basis of past data [3]. Seasonality is one of the aspects of time series analysis which can be referred as a periodic fluctuation. In order to understand it we can take following example into consideration, different types of crops grow in different seasons. It could be studied and understood by a relation and find if it is in a sinusoidal shape. It can be observed from the complete duration of a season [4].