LAUSR

Current fixed and mobile networks’ behavior is rapidly changing, which calls for flexible monitoring approaches to avoid loosing track with such a fast evolutionary pace. Due to the many challenges that this scenario is posing to network managers, we propose the exploration of Functional Data Analysis (FDA) techniques as a mean to easily deal with network management and analysis issues. Specifically, we describe and evaluate several FDA methods with applications to network measurement preprocessing and clustering, bandwidth allocation, and anomaly and outlier detection. Our work focuses on how these FDA-based tools serve to improve the outcomes of traffic data mining and analysis, providing easy-to-understand and comprehensive outputs for network managers. We present the results that we have obtained from real case studies in the Spanish Academic network using throughput time series, comparing them with other alternatives of the state of the art. With this com- parative, we have qualitatively and quantitatively evaluated the advantages of FDA-methods in the networking area.