The College of Administration and Economics at the University of Baghdad discussed, a PhD dissertation in field of Statistics by the student (Ahmed Alwan Salih) and tagged with (Using Fuzzy Hybrid Models for Time Series Forecasting with Application ) , Under supervision of (Prof. Dr. Munaf Yousif Hmood )
Forecasting is an important topic in time series analysis due to its prominent role in making future decisions. Forecasting plays a prominent role in the economic field, achieving economic growth and price stability. In particular, accurate forecasting of time series enables the best decision-making and developing the necessary plans to avoid the loss that results from price fluctuations.
This dissertation aims to employ hybrid fuzzy models to forecast global gold prices due to their importance, as they support investors in understanding market movement and making good investment decisions.
Hybrid models combine the advantages of time series in dealing with linear components, the benefits of neural networks in dealing with non-linear components, and the advantages of fuzzy logic in dealing with uncertainty. It refers to combining the linear component that represents time series models and the nonlinear component that represents the fuzzy neural networks, which are the Adaptive Fuzzy Neural Inference System (ANFIS) model, the Recurrent Fuzzy Neural Network (RFNN), and Fuzzy inference‑based LSTM for time series prediction (FLSTM). In addition to these models are trained by the Gray Wolf Optimizer (GWO) and Particle Swarm Optimization (PSO) training algorithms are used. Finally, comparing these hybrid models used in the prediction is made to choose the best one. The results showed that (ARIMA-ANFIS_PSO) and (ARIMA-ANFIS_GWO) were close in terms of performance, with (ARIMA-ANFIS_GWO) being superior. Thus, the superiority of the hybrid model with the GWO optimizer is evident.
