To Build A Forecasting Model To Predict Steel Billet Cash Prices Traded in the London Metal Exchange

Chelladurai, Vinoth babu (2010) To Build A Forecasting Model To Predict Steel Billet Cash Prices Traded in the London Metal Exchange. [Dissertation (University of Nottingham only)] (Unpublished)

[img] PDF - Registered users only - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader
Download (1MB)


The forecasting of asset prices using time series analysis techniques has focussed a great deal on the accuracy of the forecasting models. Among the traditional techniques of time series forecasting, the Box Jenkins Autoregressive Integrated Moving Average (ARIMA) models have been one of the most widely used linear time series models. There had been growing number of research indicating that novel techniques like Artificial Neural Networks (ANN) can be a promising model for forecasting and could prove to be a better alternative to the traditional models.

The aim of this dissertation is to build a forecasting model for predicting the steel billet cash (spot) prices traded in the London Metal Exchange using the time series data of the steel billet cash prices from the period of September 2008 to June 2010. In order to achieve the aim, the R software is used to identify the appropriate ARIMA model specification, model validation and forecasting. In order to predict the time series in ANN model the Matlab software is used. The forecasting performance of the selected ARIMA and ANN models are compared in order to justify a better model for predicting steel billet cash prices.

The Box-Jenkins ARIMA modelling methodology adopted indicated that the first order differencing model was weak form stationary and ARIMA (2,1,2) model was comparatively better fit than other models for both Mediterranean and Far-East steel prices. However the lack of model stability for different sub-period data limited the forecasting confidence of the selected model. The forecasting performance of ANN was limited with the measure of directional accuracy being lower for the best fit ANN architecture which could be attributed to the non-stationary nature of the steel prices and the random behaviour of steel prices.

Item Type: Dissertation (University of Nottingham only)
Depositing User: EP, Services
Date Deposited: 09 Nov 2010 11:46
Last Modified: 24 Jan 2018 19:22

Actions (Archive Staff Only)

Edit View Edit View