Prediction of Agricultural Emissions in Malaysia Using the Arima, LSTM, and Regression Models

Abstract

Agriculture has always been an important economical factor for any country which is causing emissions every day without realizing how much it is leading towards increasing number of Greenhouse Gas (GHG). Agricultural emissions has been forecasted for Malaysia to have a better understanding and to take measures right away. This can be done through a machine learning model including collecting data, pre-processing, training, building a model and testing the model for accuracy. This project aims to develop a model to forecast agricultural emissions using three most accurate forecasting models. The time series analysis consists of two models, auto regressive integrated moving average(ARIMA) and long short-term memory(LSTM) and simple linear regression model. These models illustrate the forecasted upward trend values until 2040 in Malaysia. The ARIMA model provides good prediction curves which is close to the actual values taken since 1960 and the LSTM model provides a decreasing curve for every value loss epochs which concludes to be good model for forecasting. It was concluded that, agricultural emissions is causing soaring of temperature in Malaysia and immense amount of emissions causing from agriculture. The techniques used in this paper can be enhanced more in the future and the visualizations can help the Malaysian agricultural sectors to take proper measurements to prevent this uprising agricultural emissions.