Abstract: A method for optimizing performance analysis of a plurality of network hosts associated with a communications network includes aggregating captured network performance data including a plurality of captured network performance metrics for a plurality of network flows. Each one of the plurality of network flows is associated with a plurality of network hosts. The aggregated captured network performance data is encoded by employing at least one data modification function. Dimensionality of the encoded captured network performance data is reduced using a neural network model. One or more reduced-dimensional clusters of the encoded captured network performance data are generated. Each of the one or more reduced-dimensional clusters is grouping one or more hosts of the plurality of network hosts based on the captured network performance metrics.

Abstract: A method for optimizing performance analysis of a plurality of network hosts associated with a communications network includes aggregating captured network performance data including a plurality of captured network performance metrics for a plurality of network flows. Each one of the plurality of network flows is associated with a plurality of network hosts. The aggregated captured network performance data is encoded by employing at least one data modification function. Dimensionality of the encoded captured network performance data is reduced using a neural network model. One or more reduced-dimensional clusters of the encoded captured network performance data are generated. Each of the one or more reduced-dimensional clusters is grouping one or more hosts of the plurality of network hosts based on the captured network performance metrics.