Kasper Apajalahti

StaRe: Statistical Reasoning Tool for 5G Network Management

In operations of increasingly complex telecommunication networks, characterization of a system state and choosing optimal operation in it are challenges. One possible approach is to utilize statistical and uncertain information in the network management. This paper gives an overview of our work in which a Markov Logic Network model (MLN) is used for mobile network analysis with an RDF-based faceted search interface to monitor and control the behavior of the MLN reasoner. Our experiments, based on a prototype implementation, gives promising results of utilizing an ontology and MLN model in network status characterization, optimization and visualization.

Combining ontological modelling and probabilistic reasoning for network management

Advanced automation is needed in future mobile networks to provide adequate service quality economically and with high reliability. In this paper, a system is presented that takes into account the network context, analyses uncertain information, and infers network configurations by means of probabilistic reasoning. The system introduced in this paper is an experimental platform integrating a mobile network simulator, a Markov Logic Network (MLN) model, and an OWL 2 ontology into a runtime environment that can be monitored via a Resource Description Framework (RDF) – based user interface. In this approach, the OWL ontology contains a semantic representation of the relevant concepts, and the MLN model evaluates elements of uncertain information. Experiments based on a prototype implementation demonstrate the value of semantic modelling and probabilistic reasoning in network status characterization, optimization, and visualization.

Sharing performance measurement events across domains

Network management activities, such as fault analysis and configuration management, are eventually related to changes in network measurements. Some measurement event might be either a trigger or objective of a management activity. We argue that sharing the semantics of performance data across networks provides a basis for more advanced automation. This paper presents an ontology-based system for sharing information about network measurements across network domains. The represented information contains correlations and human-defined mappings between network measurements and the system is based on semantic reasoning that identifies dependencies which arise by combining local and shared information. We demonstrate the usage of the system in a Long Term Evolution (LTE) network domain. Our experiments from an LTE simulator and LTE test network show that a combination of correlations, human-defined mappings, and ontological reasoning produces useful cross-domain information that can be accessed with ontology queries.

Using Biographical Texts as Linked Data for Prosopographical Research and Applications

This paper argues that representing texts as semantic Linked Data provides a useful basis for analyzing their contents in Digital Humanities research and for Cultural Heritage application development. The idea is to transform Cultural Heritage texts into a knowledge graph and a Linked Data service that can be used flexibly in different applications via a SPARQL endpoint. The argument is discussed and evaluated in the context of biographical and prosopographical research and a case study where over 13 000 life stories form biographical collections of Biographical Centre of the Finnish Literature Society were transformed into RDF, enriched by data linking, and published in a SPARQL endpoint. Tools for biography and prosopography, data clustering, network analysis, and linguistic analysis were created with promising first results.

Creating Time Series-Based Metadata for Semantic IoT Web Services

In the near future, the Internet of things (IoT) will rapidly change and automate tasks in our everyday life. IoT networks have sensors measuring the environment and automated agents changing it with respect to predefined objectives. Modeling agents as web services requires lots of metadata from the environment in order to define the desired performance in a specific context. For this purpose, we propose an automatic measurement-based metadata creation method that analyses multivariate time series gathered from the sensors during agents change the environment. The time series analysis uses a cumulative sum algorithm (CuSum) to detect events and association rule learning to find temporal patterns. We evaluate our system with a Long-Term Evolution (LTE) simulator having mobile phones corresponding to IoT devices, LTE macro cells as the data source, and the Self-Organised Network (SON) functions as the automated agents in the network. Our experiments give promising results and show that the metadata creation process can be utilised to characterise IoT agents.

Automatic Definition and Application of Similarity Measures for Self-Operation of Network

Self-operation concept is proposed to learn the past experiences of network operations and apply the learned operation experiences to solve new but similar problems. It works based upon the observation that actions appropriate for achieving an objective resemble each other in similar network contexts. Plenty of such similarities exist at the level of network elements, functions, and their relations. Similarity measure definition and application are essential components for the self-operation to apply the learned operation experiences. This paper provides a solution for self-operation to define and apply two types of similarity measures for two self-operation use cases. The first use case answers how to select a best suitable function to achieve any given objective. The second use case tells how the selected function should be configured with the most optimal parameter values so that the given objective could be achieved. This solution is realized on a demonstrator implementing the self-operation concept. Corresponding experiments are made with the demonstrator. The experimental results show that the self-operation solution works well.