Prediction and Analysis of International Trade Data Based on Deep Learning

Abstract

There are several interrelated factors at play behind the scenes that contribute to unpredictable changes in stock prices. Global economic statistics, shifts in unemployment rate, monetary policies of influential nations, immigration laws, natural catastrophes, public health issues, number of other factors might all be contributing factors. Everybody involved in stock market wants to increase earnings, lower risks due to careful analysis of market. The main difficulty is compiling the diverse data, placing it in a single basket, and building a trustworthy model to produce precise forecasts. In this manuscript Prediction and Analysis of International Trade Data Based on Deep Learning (PA-ITD-DTG-MG-CRN) is proposed for Prediction and Analysis of ITD. The data for proposed PA-ITD-DTG-MG-CRN method collected from Import Genius Trade Dataset and the input data is pre-processed using a Multi-Window Savitzky-Golay filter (MWSGF) for Tokenization, removing stop words and text vectorization for prediction and analysis. The Dual Temporal Gated Multi-Graph Convolution Recurrent Network (DTG-MGCN) is expected and analyzes data related to International trade, classifying it into groups such as financial data, KYC data, global trade data, bank data. Therefore, that the DTG-MGCN does not include explicit optimization procedures to guarantee precise forecasting and analysis of data related to International trade data and the Giraffe Kicking Optimization Algorithm (GKOA) is used for enhance the performance of the DTG-MGCN by optimizing its parameters, thereby increasing its accuracy in prediction, analyzes of the International trade data. The effectiveness of the proposed PA-ITD-DTG-MG-CRN method is analyzed with performance metrics likes accuracy, precision, sensitivity, specificity, FI-score, computational time, error rate. Proposed PA-ITD-DTG-MG-CRN method achieves 31.89%, 25.45% and 19.32% higher accuracy, 32.12%, 23.49%, 30.94% higher precision and 26.87%, 34.65%, 23.94% lower error rate when analyzed with existing methods such as Prediction Method of International Trade Risk Depend on Stochastic Time-Series Neural Network (PM-ITR-STSN), novel ensemble deep learning method for stock prediction depend on stock prices and news (NEDL-SP-SPN) and predicting stock market index utilizing LSTM (PDT-STI-LSTM) respectively.