WITHDRAWN: Multi currency exchange rate prediction using convolutional neural network

Abstract

Abstract Financial market is non-linear and chaotic in nature. So, Accurate prediction of foreign exchange rate is very difficult & a challenging task. Hence, many neural network techniques are used for forecasting various country’s currency exchange rates with different parameters. This article proposed convolutional neural networks for foreign currency exchange rate prediction. In this article we would like to propose a model which could develop a multivariate exchange rates information and put these features for better use. The performance of the proposed system has been tested with Australian Dollar against Us Dollar (AUD/USD), European Euro against Us Dollar (EUR/USD), European Euro against Canadian Dollar (EUR/CAD), British pound against Us Dollar (GBP/USD), Us Dollar against Canadian Dollar (USD/CAD), Us Dollar against Japanese Yen (USD/JPY) and used to predict one day and five days exchange rate in advance. Adaptive learning rate method (ADAM) optimization technique is used here which generates optimal weight for the proposed Model. The proposed model has been found with the best prediction result after measuring the performance of R2, Mean Absolute Error, Root Mean Square Error, Mean Absolute Percentage Error and CEV in comparison with ARIMA, Linear Regression and Multi-layer perceptron techniques.