UNCORRECTED PROOF 1 2 Predicting service request in support centers based on 3 nonlinear dynamics, ARMA modeling and neural networks q 4 Emili Balaguer a , Alberto Palomares a , Emilio Soria b, * , Jose David Martı ´n b 5 a Tissat S.A., R&D Department, Av. Leonardo Da Vinci, 5. 46980 Paterna, Valencia, Spain 6 b G.P.D.S, Digital Signal Processing Group, Electronic Engineering Department, Escuela Te ´ cnica Superior de Ingenierı ´a, 7 Universitat de Vale `ncia, C/ Dr. Moliner 50, 46100 Burjassot, Valencia, Spain 8 9 Abstract 10 In this paper, we present the use of different mathematical models to forecast service requests in support centers (SCs). A successful 11 prediction of service request can help in the efficient management of both human and technological resources that are used to solve these 12 eventualities. A nonlinear analysis of the time series indicates the convenience of nonlinear modeling. Neural models based on the time 13 delay neural network (TDNN) are benchmarked with classical models, such as auto-regressive moving average (ARMA) models. Models 14 achieved high values for the correlation coefficient between the desired signal and that predicted by the models (values between 0.88 and 15 0.97 were obtained in the out-of-sample set). Results show the suitability of these approaches for the management of SCs1. 16 Ó 2006 Elsevier Ltd. All rights reserved. 17 Keywords: Time series analysis; Neural networks; Call centers 18 19 1. Introduction 20 Support centers (SCs) usually deal with all the requests 21 reported by either external customers/citizens or internal 22 users of an organization/company. Although telephone 23 has been the traditional way to provide support in the call 24 centers (CC’s), nowadays the Internet provides new mech- 25 anisms of communication that overcome some of the limi- 26 tations associated with telephone support, and it is widely 27 used by SCs. The formal contract between the service pro- 28 vider and the service recipient is known as the service level 29 agreement (SLA). SLA must define the quality measures or 30 service level measurements (SLMs) that are used to evalu- 31 ate the quality of the support service. SLA tends to contain 32 clauses that economically incentive/penalize the SC 33 depending on its degree of fulfillment according to SLMs. 34 However, SLMs is one of the most crucial and complex 35 tasks in the management of SCs and is out of the scope 36 of this study. 37 Different system performance and business parameters 38 can be considered for SLMs: cumulative time-based 39 parameters, service availability, number of affected users, 40 metrics based on a particular business process, etc. These 41 measurements must be automatically collected, maintained 42 and analyzed in order to manage the service support pro- 43 cess. HelpDesk consists of several software applications 44 that allow to record, track and report all the information 45 involved in a SC, including the SLAs and SLMs manage- 46 ment2. In general, the information than can be obtained 47 from a HelpDesk is considerably wide: general informa- 48 tion, purchases, complaints, etc. Operators usually answer 49 many calls of different nature: sales, relational marketing, 50 customer services, citizen services, technical support, and 51 in general, any specialized activity related to business or 52 public administration. Moreover, some contracts require 53 a minimum fulfillment degree, and apply penalizations if 54 this degree is not achieved. 55 There are several aspects that should be covered by a 56 methodology which tries to help in the management of a 0957-4174/$ - see front matter Ó 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.eswa.2006.10.003 q This work has been partially supported by the Spanish Ministry of Industry, Tourism and Commerce, with the project FIT-340001-2004-11. * Corresponding author. E-mail address: Emilio.soria@uv.es (E. Soria). www.elsevier.com/locate/eswa Expert Systems with Applications xxx (2006) xxx–xxx Expert Systems with Applications ESWA 1859 No. of Pages 9, Model 5+ 8 November 2006 Disk Used ARTICLE IN PRESS Please cite this article in press as: Balaguer, E. et al., Predicting service request in support centers based on ..., Expert Systems with Applications (2006), doi:10.1016/j.eswa.2006.10.003