Modeling downtime severity of telecommunication networks using discrete time Markov chains

Telecommunication network reliability remains a top priority for both customers and service providers. Downtime can result in revenue loss for providers and productivity loss for customers. Thus, accurately predicting downtime severity can help providers plan and respond effectively. This study models telecommunication network downtime severity using discrete-time Markov chains (DTMC). The data used consists of 1,211 daily network downtime records, in minutes, recorded by Ghana's national communications authority (NCA) from August 1, 2015, to April 30, 2021. The severity of daily downtime was categorized into 5 categories based on duration. Results indicate that the majority (n=905) of daily network downtime was negligible, while only 25 outages were severe. The transition probability matrix indicates that if the present network downtime severity is negligible, there is an 81% chance that the next network downtime severity may also be negligible, a 12% chance of minimal severity, a 4% chance of significant severity, a 2% chance of serious severity, and a 1% chance of severe severity. The steady-state distribution indicates that over the long term (n ≥ 17), 74% of network downtime severity is expected to be negligible, while only 2% is expected to be severe. Based on probability simulations for 12 steps, it is evident that the 'negligible' category is the most probable network downtime severity, regardless of the initial severity category. These findings can assist telecommunication providers in better planning and delivering more reliable services to their customers.