Development of QoS Predictive Model for Major Network Operators in Nigeria using C4.5 and ID3 Algorithms.
Keywords:
Crowdsourcing, Quality of Service, C4.5, ID3, Algorithms, Mobile Network OperatorAbstract
The use of mobile telecommunication in Nigeria, especially in the era of COVID-19, is vital; it is also an agent of the country's socio-political and economic advancement. The number of subscriptions grew from one million to one hundred and eighty million between 2002 and 2019. However, increase in the number of users without the corresponding increase in the expansion of network resource is a challenge, which raises Quality of Service (QoS) issue in the country, especially in the COVID-19 era hence the justification for this research. This work aims to develop a predictive model for audio QoS for the major network operators in Nigeria. The major service providers in Nigeria are labeled X1, X2, X3, and X4, the other minor operators are labeled X0. The crowdsourcing technique employed to collect primary data for this research. Machine learning approach, specifically ID3 and C4.5 algorithms, were used to develop models for each of four major mobile network operators in Nigeria. The telecom sector generates large and complex data that need to be analyzed and draw the inference to help the stakeholder. Hence the justification for the use of machine learning paradigm. Congestion rate, received signal strength, call success setup rate, and call drop rate are the four QoS parameters applied for this work. Furthermore, the performance evaluation metrics used are precision, accuracy, false alarm rate, and true positive rate. The result shows that the ID3 had better accuracy than C4.5 in only one of the datasets. The C4.5 and ID3 decision trees algorithm had an equal performance on the X3, X4, and other network datasets. The C4.5 Area under ROC is 0.998 for X1, X2, X3 and X4. Similarly, ID3 Area under ROC for X1, X2, X3, and X4 are 0.995, 0.998, 0.993 0.990 respectively. Since the results show that, the ID3 and C4.5 accuracy, precision and ROC are high; hence; they are good classifiers for determining the QoS of major mobile network operators.