Proceedings of the National Conference on Emerging Computer Applications (NCECA)-2022 Vol.4, Issue.1 581 DOI: 10.5281/zenodo.6935207 ISBN: 978-93-5607-317-3@2022 MCA, Amal Jyothi College of Engineering Kanjirappally, Kottayam A Case Study on Natural Gas Price Prediction Using Machine Learning Algorithms Julia George Department of Computer Applications Amal Jyothi College of Engineering, Kanjirappally, India juliageorge2022b@mca.ajce.in Lisha Varghese Department of Computer Applications Amal Jyothi College of Engineering, Kanjirappally, India lishavarghese@amaljyothi.ac.in Abstract — Natural gas (also known as fossil gas) is a hydrocarbon gas mixture that occurs naturally and is mostly composed of methane. It is a clean and efficient fuel that is frequently recommended. Its price fluctuates on a daily basis, based on a variety of factors such as location, demand, and manufacturing, to name a few. In this model, we are using the Machine learning algorithms to help predict the values. We chose Decision-Tree Regression Algorithm in this model. Algorithm based on decision trees belongs to the supervised learning algorithms category. It is suitable for both continuous and intermittent use in addition to category output variables. Keywords— natural gas, machine learning, multilinear regression, decision tree regression algorithm. I. INTRODUCTION Natural gas prediction is crucial in the business world. The goal of this model is to forecast the spot price of natural gas when the user provides the date as an input. It then forecasts the price of gas on that particular day. Natural gas has been offered as a strategy for increasing energy supply security and lowering pollution levels all around the world. Estimating natural gas prices has turned into a truly valuable device for all market participants in competitive natural gas markets, benefiting a variety of stakeholders. Machine Learning Regression algorithms have rapidly gained popularity as a method for predicting natural gas prices. The main goal here is to forecast Natural Gas Prices based on demand and supply. With a high percent accuracy, Decision Tree Regression is the best and most efficient approach for training the machine out of all the techniques. Because of its great environmental benefits, natural gas, one of the most essential energy supplies, will play a larger role in the future of global energy. Forecasting natural gas costs could aid businesses in expanding their operations. This research looks at how machine learning may be used to anticipate natural gas prices. Best accuracy was achieved using the Decision Tree Regression. The dataset is trained for various methods, and the efficiency of each is tested. Finally, the algorithm with the highest accuracy is completed and considered. The ML Algorithm for Linear Regression predicts the worth of the reliant variable y in view of the worth of the autonomous variable x. As a result, it only evaluates one input, whereas we have three: day, month, and year. As a result, the ML Algorithm for Linear Regression cannot be applied. Instead, you might utilise the Multiple Linear Regression Algorithm. It predicts the result of a response variable using many explanatory variables. A reliable natural gas price projection model is incredibly beneficial. This data is used by natural gas operators to assess the economics. II. LITERATURE REVIEW Dimitrios Mouchtaris, Emmanouil Sofianos , Periklis Gogas, Theophilos Papadimitriou[1] developed an attempt using machine learning approaches, to forecast the natural gas spot price for the next 1, 3, 5, and 10 days: using SVMs, regression trees, linear regression, and Gaussian process regression. Moting Su, Zongyi Zhang, Ye Zhu[2] Using conventional machine learning tools, constructed data-driven predictive models for natural gas price forecasting. such as neural networks created artificially, support vector machines (SVM), gradient boosting machines (GBM), and Gaussian process regression in this research. (GPR) For model training, used cross-validation and monthly Henry Hub natural language processing. Jiarong Mu , Xianfeng Liu, Peiyan Zang[3] developed a natural gas futures price's non-linear and long-term dependence, this paper makes a prediction using the long- term and short-term memory (LSTM) model, which can be extended in time and has a long-term memory function, and effectively solves the gradient explosion problem of traditional neural networks. Ervin Ceperic, Sasa Zikovic, Vladimir Ceperic[4] developed a short-term Henry Hub spot natural gas price predictions based on the performance of conventional time series models and machine learning methods, notably neural networks (NN) and strategic seasonality-adjusted support vector regression machines, are presented (SSA-SVR). The use of feature selection (FS) methods to generate model inputs and choose model inputs is suggested. Yuanyuan Tang, Qingmei Wang, Wei Xu[5] examines whether internet search data and news sentiment may help with predicting and compares which can produce better outcomes. The findings of the experiments suggest that both internet search and news sentiment contain additional information that can increase the prediction effect, and that internet search data can produce better predictions. III. MOTIVATION Natural gas has been offered as a strategy for increasing energy supply security and lowering pollution levels around the world. Forecasting natural gas prices has become a very