Network Disruption Prediction Using Naïve Bayes Classifier

The most crucial challenge of internet service providers is to assure the availability and reliability of their services to their customers. The companies should prevent the customer's complaint by recognizing a potential disruption for the customers, especially in the category ‘under spec' condition (potentially impaired service). This study proposed and implemented a model using the Naive Bayes classifier to classify and detect the potential disruption of network services to prevent customer's complaints about their service. The criteria for this model prediction are revenue number of each customer (REVENUE), recurrent disruption value of ODP (N_Q), attenuation value in ODP (OLT), and attenuation value in customer (ONU). The data classified into three classes or conditions, namely GREEN representing no network disruption, YELLOW is representing low-level disruption, and RED representing high-level disruption, which needs more attention to follow up. The result obtained 91.89% accuracy of the model performance using WEKA Tool.