This seminar explores Bayesian time series modeling by integrating ARIMA (Auto-Regressive Integrated Moving Average) with MCMC (Markov Chain Monte Carlo) to enhance parameter estimation accuracy and predict uncertainty. The ARIMA model, known for its efficacy in analyzing time series data, is significantly improved when combined with Bayesian methods like MCMC. MCMC facilitates efficient sampling from complex posterior distributions, leading to more reliable parameter estimation and, consequently, more accurate forecasts. This integration is particularly beneficial in dealing with non-linear and complex data structures, where traditional methods may struggle. The combined ARIMA-MCMC approach provides a robust framework for time series forecasting, capable of handling uncertainty and improving decision-making in uncertain environments. The seminar also discusses the practical implementation of this integrated model, including the steps involved in defining likelihood functions, applying the Metropolis-Hastings algorithm, eva‎luating model performance using MAPE and RMSE metrics, and determining forecast intervals to quantify predictive uncertainty. The study concludes by emphasizing the importance of Bayesian methods in time series analysis and recommending further research to explore alternative sampling algorithms and applications in other complex time series scenarios.

10/2/2023 12:00:00 AM

83743

1403/07/03

Bayesian Time Series Modeling: Merging MCMC and ARIMA for Robust Parameter Estimation and Predictive Uncertainty

سري زماني بيزي , ARIMA , MCMC , تخمين پارامتر , عدم قطعيت پيش بيني , الگوريتم متروپليس-هيستينگز , پيش بيني , تجزيه و تحليل سري زماني , استنتاج بيزي

Bayesian Time Series , ARIMA , MCMC , Parameter Estimation , Predictive Uncertainty , Metropolis-Hastings Algorithm , Forecasting , Time Series Analysis , Bayesian Inference