Cristina Laura Stolojescu, du département Logique des usages, sciences sociales et sciences de l’information et laboratoire Lab-STICC, a présenté ses travaux de thèse réalisés à l’Université Polithnica de Timisoara (Roumanie) et à la  Faculté d’ Électronique et de Télécommunications de Timisoara (Roumanie) le 13 janvier

Cristina Laura Stolojescu doctorante au département Logique des usages, sciences sociales et sciences de l'information de Télécom Bretagne et laboratoire Lab-STICC.

Résumé : This thesis has a practical objective. It consists in finding an answer to the following question : “Is it possible to identify the base stations (BS) which are bad positioned in a Worldwide Interoperability for Microwave Access (WiMAX) network topology by traffic analysis ?”. The question is important for the planning and exploitation of WiMAX networks. This thesis is based on the research of time series analysis. Time series analysis has been an area of considerable activity in recent years. The work of this thesis evaluates a set of time series conceived by monitoring the traffic developed in a WiMAX network, composed by 67 BSs, for a time interval of eight weeks, developed by Alcatel Lucent Timisoara, Romania. Taking into consideration the high volume of information contained in this database, a data-mining approach was preferred. Assuming that the traffic associated with a BS bad positioned is heavier than the traffic associated with a BS well positioned, two approaches for the appreciation of the heaviness of the traffic were developed. The first approach is based on the supposition that a BS with heavy traffic has a reduced risk of saturation. Hence, it is necessary to appreciate the risk of saturation of each BS. This is equivalent with the estimation of the moment when the BS will saturate. So, the first objective of this thesis is to propose an approach for predicting time series. This approach is based on a multiple resolution decomposition of the signal using the Stationary Wavelet Transform (SWT) and Autoregressive integrated moving average (ARIMA) model. The second approach for the appreciation of the heaviness of the traffic is based on Long Range Dependence (LRD) analysis. This is a relative new statistical concept in communication traffic analysis and can be implemented using wavelets as well. The estimation of LRD degree is realized through the estimation of the Hurst parameter of the time-series under analysis. THis property has important implications on the performance, design and dimensioning of the network. Our objective is to highlight the particularities of WiMAX traffic from a LRD perspective and to analyze the positioning of BSs in the architecture of a WiMAX network. The results show which BSs have a good localization in the topology of the network and which BSs have a bad localization in the topology of the network and must be repositioned when the next session of network maintenance will take place. The application of both data mining techniques (forecasting and LRD analysis) in the wavelets domain is decisive for their performance, improving the speed and the precision of the developed algorithms.

Mots-clés : ondelettes, séries temporelles, data mining.

Membres du jury : Alexandru Isar, professeur, Université Politehnica de Timisoara, Roumanie – Philippe Lenca, maître de conférences, Télécom Bretagne, Brest – Lotfi Senhadjl, professeur, LTSI, Université de Rennes 1 – Monica Borda, Professeur, Universitatea Tecnica Cluj – Napoca, Roumanie – Thierry Dhorne, professeur, Université de Bretagne Sud, Vannes – Adrian Ivan, ingénieur, Alcatel Lucent, Timisoara, Roumanie – Corneliu Rusu, professeur, Universitatea Tecnica Cluj – Napoca, Roumanie – Ioan Nafornita, professeur, Université Politehnica de Timisoara, Roumanie – Marius Ostesteanu, professeur, Université Politehnica de Timisoara, Roumanie – Sorin Moga, maître de conférences, Télécom Bretagne, Brest.

Voir les publications de Cristina Laura Stolojescu dans la base de l’École.