TY - JOUR
TI - Trade value prediction using hybrid graph convolutional recurrent neural network with lion optimization model
AU - Binsuwadan Jawaher
AU - Younis Halima A
AU - Abdulrahim Hiyam
AU - Yagoub Rahmtalla Y
AU - Elhag Azhari A
JN - Thermal Science
PY - 2024
VL - 28
IS - 6
SP - 5019
EP - 5028
PT - Article
AB - Trade value prediction (TVP) is major for understanding financial dynamics  and directing policy decisions in the perspective of complex systems  science. The study emphases on an analytical model intended to predict  future trade values by evaluating financial indicators, past trade data,  and geopolitical powers. By using advanced statistical models and machine  learning techniques, the model explores relationships and patterns in trade  flows among countries. The perceptions increased from this technique offer  beneficial support for policymakers and businesses, guiding them to  forecast the effects of financial and policy changes on global trade. Also,  the study emphasizes the importance of a complicated method to enhance the  accuracy of trade predictions and aid tactical decision-making in a  worldwide interconnected economy. This study proposes trade value  prediction using hybrid graph convolutional recurrent neural network with  Lion optimizer algorithm (TVP-HGCRNNLOA) methodology. The objective function  of the TVP-HGCRNNLOA methodology is to develop an accurate predictive model  for trade values between countries. Primarily, the TVP-HGCRNNLOA approach  undergoes the data normalization by employing linear scaling normalization  technique. Then, the hybrid graph convolutional recurrent neural network  (HGCRNN) method is used for forecasting process. At last, the TVP-HGCRNNLOA  model performs the hyperparameter tuning by utilizing the Lion optimization  algorithm model. The experimental analysis of the TVP-HGCRNNLOA methodology  is investigated in terms of various measures under mean squared error, mean  absolute error, and mean absolute percentage error. The performance  validation portrayed the superior performance of the TVP-HGCRNNLOA  methodology over other existing approaches.