Time Series Modelling Approach for Predictive Analytics

• Published in: 2024 Third International Conference on Distributed Computing and Electrical Circuits and Electronics (ICDCECE) pp. 1 - 6
• Main Authors: Pande, Shilpa Mangesh, Kumar Ramesh, Prem
• Format: Conference Proceeding
• Language: English
• Published: IEEE 26-04-2024
• Subjects: Computational modeling; Crops; demand prediction; holt winter's seasonal method; machine learning; Predictive models; Pricing; SARIMA; Smoothing methods; Supply and demand; Time series analysis; time series modelling
• DOI: 10.1109/ICDCECE60827.2024.10548032

India has been predominately an agrarian nation since the dawn of mankind. Agriculture continues to be the primary source of income in many parts of the country. Due to the population's rapid growth, there has been a surge in the requirement for crops and food in recent years. This gives the agricultural industry considerable room to create novel initiatives to satisfy the demand. Despite of rising demand, Indian farming sector didn't prove very successful. Poor demand and supply management typically leads to both high pricing for consumers and losses for farmers in developing nations like India. Our analysis demonstrates a significant mismatch between the demand for and supply of rice crops. Under various market conditions, this mismatch causes problems for both consumers and farmers in the Karnataka region. Taking into account the rice crop's demand, supply, and price fluctuations, we propose a model to balance demand and supply by predicting the rice price. In the proposed model, we discuss two methodologies for forecasting demand of rice crop. The proposed model uses time series models to predict demand. The rice crop price is predicted using SARIMA and Holt-Winter's seasonal approach, and their results are contrasted using RSME values. In order to reduce errors in prediction and to get accurate predictions from the best fit model, SARIMA and Holt- Winter's seasonal approach techniques were examined on the datasets by selecting initial values long with smoothing factors. This evaluation demonstrated trends and seasonal patterns. SARIMA model revealed promising results on the selected data sets.