Forecasting volatility evidence from the futures market in India

Title

Forecasting volatility evidence from the futures market in India

Creator

A S, Veena.

Contributor

Mathew, Jain.

Description

This thesis focuses on modelling and forecasting of select products in the Indian futures market using econometric time series models and artificial neural network based models. These models have been compared for their forecasting accuracy to determine the best forecasting model for a particular futures series. This study applies GARCH, EGARCH, PARCH, TARCH, and Artificial Neural Networks (ANN) to assess the best predicting model for exchange rate futures, commodity index futures and stock index futures. After testing for stationarity of data series, GARCH, EGARCH, PARCH and TARCH models are developed. In addition to in-sample forecasts, 1-day, 5-day, 10-day, 15-day and 30-day out-of-sample forecasts have been carried out. For ANN, data is scaled using the minmax scaling methodology to ensure that newlinethe data series is normalised and in the range of 0 to 1. ANN is developed using the feedforward methodology. While the basic neural network architecture has one input layer, one hidden layer and one output layer, the number of neurons in the input and hidden layers vary from 1 to 20. The optimum number of input and hidden neurons in their respective layers are then selected based on the combination which gives the least error. These network combinations are used for out-of-sample forecasting and errors are compared with the forecast output of the GARCH models. RMSE, MAE, MAPE, Theil s-U statistic and Correlation coefficient is computed for error newlinecomparison. Results indicate that for currency futures and commodity index futures, ANN provides better forecast accuracy. For stock index futures, GARCH family models work better in some cases.

Source

Author's Submission

Date

2018-01-01

Publisher

Christ(Deemed to be University)

Subject

Management Studies

Rights

Open Access

Relation

No Thesis

Format

PDF

Language

English

Type

PhD

Identifier

http://hdl.handle.net/10603/426707