APPLYING DATA MINING APPROACH  TO  ANALYZE  CALL  CENTER PERFORMANCE: THE CASE OF ETHIO TELECOM

BELISA, MULUGETA

dc.contributor.author	BELISA, MULUGETA
dc.date.accessioned	2022-03-09T06:56:38Z
dc.date.available	2022-03-09T06:56:38Z
dc.date.issued	2021-08
dc.identifier.uri	http://ir.bdu.edu.et/handle/123456789/13183
dc.description.abstract	The aim of any organization in world is to deliver quality of service and handling customer issues timely without any interruptions. The call center channels are the comfortable one to handle customer issues quickly. But, it needs well known data analyzing methods because, huge amount of data is generated from this section. And also evaluating the performance of this section is good for the company for further decision making purposes. When we say call center performance, it is the technical implementation of a simple necessity like evaluating channel accessibility, customer satisfaction and overall success rate of the domain. Telecom companies use data mining algorithms and tools to analysis call center performances based on historical data. In our case, the analysis is based on key performance indicator (KPI) of international and Ethio telecom (ET) benchmarks. Three data mining algorithms were used: namely Probabilistic Neural Network, K-Nearest Neighbors and Decision Tree. ET call center data which consisted of 62,956 instances and 17 attributes without including target class were used for building and testing the algorithms. The classification is based on overall performance scores rules (not met, some met, met and exceed) against KPI benchmarks settled by the company. For all experiments, the Konstanz Information Miner tool was used with 10-fold cross validation and percentage split test options. Accuracy, Cohen’s kappa, recall, precision and f-measure was among performance of model evaluation methods including subjective evaluation of domain experts. As a result a classification model built on k-nearest neighbor with 10-fold cross validation has got the best classification accuracy by correctly classifying 99.98 % of the data in to their classes and 0.999 Cohen’s kappa which indicates very good degrees. The model built with decision tree has an accuracy of 99.88% and 0.997 Cohen’s Kappa with 10-fold cross validation test option. Whereas Probabilistic classifiers recoded an accuracy of 97.15% and 0.938 Cohen’s Kappa with percentage split test option. Keywords: Call Center Performance, Data Mining, Key Performance Indicator, Konstanz Information Miner, Classification Algorithms.	en_US
dc.language.iso	en_US	en_US
dc.subject	computer science	en_US
dc.title	APPLYING DATA MINING APPROACH TO ANALYZE CALL CENTER PERFORMANCE: THE CASE OF ETHIO TELECOM	en_US
dc.type	Thesis	en_US