Using Support Vector Machines for Direct Marketing Models
A. Nachev1, T. Teodosiev2
1A. Nachev, BIS, Cairnes Business School, NUI Galway, Galway, Ireland.
2T. Teodosiev, Department of Computer Science, Shumen University, Shumen, Bulgaria
Manuscript received on 15 April 2015 | Revised Manuscript received on 25 April 2015 | Manuscript Published on 30 April 2015 | PP: 183-190 | Volume-4 Issue-4, April 2015 | Retrieval Number: D3963044415/15©BEIESP
Open Access | Editorial and Publishing Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: This paper presents a case study of data mining modeling for direct marketing, based on support vector machines. We address some gaps in previous studies, namely: dealing with randomness and ‘lucky’ set composition; role of variable selection, data saturation, and controlling the problem of under-fitting and over-fitting; and selection of kernel function and model hyper-parameters for optimal performance. In order to avoid overestimation of the model performance, we applied a double-testing procedure, which combines cross-validation, and multiple runs. To illustrate the points discussed, we built predictive models, which outperform those discussed in previous studies.
Keywords: Classification, Data Mining, Direct Marketing, Support Vector Machines
Scope of the Article: Classification