Dimensioning of telecommunication infrastructure

Abstract

The application relates to method for determining resource needs needed in a telecommunications network for providing a network function in the telecommunications network, comprising the steps of—determining training data describing a network traffic, determining domain specific features of the telecommunications network to be used as predictors of a regression model in which the predictors are weighted by model weights to determine the resource needs, wherein the training data are described by the regression model, the regression model being configured to predict the resource needs, determining domain knowledge of the telecommunications network on at least one of expected values of the model weights, expected boundaries of the model weights and constraints between the model weights, encoding the domain knowledge as at least one of a prior probability distribution, constraints between model weights and boundaries of the model weights, encoding target values of the resource needs as a likelihood probability distribution centered on outcomes of the regression model, determining the model weights of the regression model through Bayesian modeling, taking into account the likelihood probability distribution, the prior probability distribution, optionally at least one of the constraints between the model weights and boundaries of the model weights, and determining the resource needs based on the determined regression model.

Dependencies are inferred from the claim text.