A Novel Approach for Prediction Challenges by Statistical Performance Using Artificial Neural Network

International Journal of Research and Scientific Innovation (IJRSI) | Volume VI, Issue V, May 2019 | ISSN 2321–2705

A Novel Approach for Prediction Challenges by Statistical Performance Using Artificial Neural Network

Arpitha K Shetty¹, Pratheksha Rai N², Bhargavi K³

^{1, 2}Assistant Professor, A.J Institute of Engineering and Technology, Kottara Chowki, Mangalore, Karnataka, India
³Assistant Professor, Krupanidhi Degree College, Bengaluru, Karnataka, India

Abstract — The various models are proposed to handle the dollar rate predictions by using mathematical operations like conventional methods, support vector machines, artificial neural networks, association rule mining, markov model etc. We proposed this work to forecast the dollar rate series data for varies applications by using neural network. This is done by benchmarking the statistical performance of an Autoregressive Integrated Moving Average (ARIMA) model, and two NNs, namely a Multi-Layer Perceptron (MLP), Recurrent Neural Network (RNN) and Simple Average method(SM). The statistical performance is measured based on Mean absolute error (MEA) and Root Mean Square Error (RMSE). This work proposed that MEA and RMS errors of MLP and SM is less as compare to the RNN, ARIMA methods. Hence the MLP neural network and SM will provide accurate results compare to RNN and ARIMA model based on the statistical performance.

Keywords — statistical performance, Markov model, Data mining, vector machines, Artificial neural network, Mean absolute error (MEA), Root Mean Square Error(RMSE)

I. INTRODUCTION

In dollar rate prediction, the available source of training data is the closing rates of dollar generated in the previous days by the European Central Bank (ECB) at 11:55 pm. Based on these predicted rates from input data the trader can decide whether to buy or sell the shares and can know the loss and gain from the marketing. Dollar rate prediction overcomes the prediction challenges trading and extensive training/prediction time. These challenges are addressed by implementing this work in R language. It is a software environment and programming language for statistical computing and graphics widely used among statisticians and data miners for data analysis and developing statistical software . In this process, we use Neural Networks for prediction.

Read More