A Multivariate Approach to Predicting Quantity of Failures in Broadband Networks Based on a Recurrent Neural Network

In this paper, we present a multivariate recurrent neural network model for short-time prediction of the number of failures that are expected to be reported by users of a broadband telecommunication network. An accurate prediction of the expected number of reported failures is becoming increasingly important to service providers. It enables proactive actions and improves the decision-making process, operational network maintenance, and workforce allocation. Our previous studies have shown that the recursive neural network is flexible enough to approximate the dynamics of the failure reporting process. Development of the model is based on long-term monitoring of failure-reporting processes and experience gained through fault management related to the network of one of the leading Croatian telecom providers (T-HT). Many factors, both in the network and outside the network, influence the time series representing failure reporting. The model encompasses the most important predictor variables and their logical and temporal dependencies. Predictor variables represent internal factors such as profiles of past and current quantities of failures as well as external factors like weather forecasts or announced activities (scheduled maintenance) in the network. External factors have a strong effect on fault occurrence, which finally results in failures reported by users. These factors are quantified and included as input variables to our model. The model is fitted to the data from different sources like an error-logging database, a trouble-ticket archive, announced settings logs and a meteo-data archive. The accuracy of the model is examined on simulation tests varying the prediction horizons. Assessment of the model’s accuracy is made by comparing results obtained by prediction and the actual data. This research represents a real-world case study from telecom operations. The developed prediction model is scalable and adaptable so that other relevant input factors can be added as needed. Hence, the proposed prediction approach based on the model can be efficiently implemented as a functionality in real fault-management processes where a variety of available input data of different volumes exist.