Portal de Periódicos da CAPES

Capítulo de livro Produção Nacional Revisado por pares

BIBTEX RIS

Improving the AHT in Telecommunication Companies by Automatic Modeling of Call Center Service

2019; Springer Science+Business Media; Linguagem: Inglês

10.1007/978-3-030-30244-3_9

1611-3349

Henrique de Castro Neto, Rita Maria Silva Julia, Elaine R. Faria, André C. P. L. F. de Carvalho, Anisio Pereira Santos, Diansley Raphael Santos Peres, Etienne Silva Julia, Jony Teixeira de Melo, Umberto Maia Barcelos, Josiane Esteves de Assis,

Consumer Retail Behavior Studies

Several companies have been taking advantage of Data Mining (DM) techniques to improve their operations by retrieving relevant knowledge from available raw data. Due to the high impact of the internet in all domains of modern life, such data exploration is particularly important in the context of Internet Service Providers (ISP). In these companies, significant information can be extracted from raw data available in Data Warehouses built from sequences of dialogues between customers and attendants, recorded in their Customer Relationship Management (CRM) system. These data are collected every time the clients contact the call center to report problems. Thus, these data have several scripts, each with a set of questions and actions that must be carried out over the course of the dialogues. Two parameters are very relevant in such attendance process: the Average Handle Time (AHT), representing the average time spent to solve the problems of the clients; and the costs with technical visits required whenever the attendant can not remotely solve such problems through the call center scripts. This paper proposes to enhance the customer service of an ISP company through the following strategy: firstly, performing a modelling of its CRM Data Warehouse; and secondly, using such model to improve the call center scripts, so as to reduce the AHT. The modelling proposed here uses DM to induce classification rules able to predict the need for technical visit taking into account the client problems. In the experiments carried out, rules with high predictive accuracy were generated. The results confirm that the generated rules allow for reducing the call AHT and can also be used as a tool for reducing the need for technical visits.