Vadlamani, Ravi and Sharma, Anurag (2014) Support Vector–Quantile Regression Random Forest Hybrid for Regression Problems. In: Multi-disciplinary Trends in Artificial Intelligence. , ed. by Murty, M. Narasimha. Lecture Notes in Computer Science, 8875 . Springer, Cham, pp. 149-160. ISBN 978-3-319-13364-5 DOI 10.1007/978-3-319-13365-2_14.

Text Vadlamani.pdf - Reprinted Version Restricted to Registered users only Download (410kB)

Abstract

In this paper we propose a novel support vector based soft computing technique which can be applied to solve regression problems. Proposed hybrid outperforms previously known techniques in literature in terms of accuracy of prediction and time taken for training. We also present a comparative study of quantile regression, differential evolution trained wavelet neural networks (DEWNN) and quantile regression random forest ensemble models in prediction in regression problems. Intervals of the parameter values of random forest for which the performance figures of the Quantile Regression Random Forest (QRFF) are statistically stable are also identified. The effectiveness of the QRFF over Quantile Regression and DWENN is evaluated on Auto MPG dataset, Body fat dataset, Boston Housing dataset, Forest Fires dataset, Pollution dataset, by using 10-fold cross validation.

Actions (login required)

View Item