Data Mining Applied to the HFC Network to Analyze the Availability of Telecommunication Services

The failures that affect the telecommunications service in hybrid fiber-coaxial networks (HFC) can be anticipated, through data analysis, to minimize the impact on service availability and users. This work seeks to analyze the causes that affect the performance of HFC networks in Ecuador, through a predictive model for their timely detection and application of anticipated works. The base studied included the components and causes that affected the availability of television and telephony service in a HFC network in 2020, from a sample of 17707 records. For the research, the dependent variable CATEGORY CAUSE is considered, and the objective was to analyze the causes that affect the telecommunications service. The KDD methodology was used and, from the use of WEKA software, classification algorithms were established to determine suitable predictive features, being the causes, the type of event, and the network components, the most significant. The results showed that, for the television service, the best classification algorithm was J48, with a precision value of 68.668% and an area under the curve (ROC Area) of 0.913. For the telephony service, the best classification algorithm was Random Tree, with a precision of 73.666% and an area under the curve of 0.969. The conducted research demonstrated the importance of data mining in the process of analyzing the causes of the impact on television and telephone services.