search query: @supervisor Tirkkonen, Olav / total: 43
reference: 4 / 43
| Author: | Manavi Nezhad, Abbas |
| Title: | Smart Content Delivery in cellular Network |
| Publication type: | Master's thesis |
| Publication year: | 2016 |
| Pages: | 50 s. + liitt. 7 Language: eng |
| Department/School: | Sähkötekniikan korkeakoulu |
| Main subject: | Radio Communications (S3019) |
| Supervisor: | Tirkkonen, Olav |
| Instructor: | Ruttik, Kalle |
| Electronic version URL: | http://urn.fi/URN:NBN:fi:aalto-201611025264 |
| Location: | P1 Ark Aalto 5401 | Archive |
| Keywords: | anticipatory mobile network content delivery Gaussian mixture model (GMM) EM algorithm linear prediction |
| Abstract (eng): | With the appearance of smartphones and a growing number of mobile subscribers, more media content is delivered to users. The growing volume of demanding media content causes challenges in usage of network resources. Deep understanding of changes in the temporal and spatial behavioral trends of users, can lead telecom companies and service provider to implement more optimized methods for resource allocation to the users. One of the methods that can help is to schedule data delivery based on the user location and network statistics. In another words, predicting user mobility behavior and network resources can help to optimize data delivery scheduling. Anticipatory mobile communication is a field which uses different methods to provide more optimized decisions regarding content delivery, based on the prediction of the users' location and network resources. In this thesis a system is modeled based on the user's location and available data rate prediction, by analyzing historical statistics of the network. Based on the assumption that a user's life pattern follows a cyclic daily pattern, following a Gaussian distribution, a Gaussian mixture model and expectation maximization (EM) algorithm have been used to predict the user mobility pattern in future. In addition, linear prediction is chosen to predict the network spectral efficiency and available bandwidth in future, through a regression matrix based on historical statistics. Applying this methodology, the available data rate to the user at different times and locations in future can be predicted. Merging this data with the user presence probability at different locations in future would result in deriving the probability distribution function of the required time duration to deliver a specific amount of data to the user. |
| ED: | 2016-11-13 |
INSSI record number: 54798
+ add basket
INSSI