| dc.description.abstract | Small cell networks are complements for existing networks to improve quality of service (QoS), capacity, and coverage. The primary purpose of this thesis is to mine mobile network data to provide an algorithm that mobile network operators can use to determine the best small cell network topology automatically instead of manually. The main drawbacks for deploying topology manually is the cost and time the effort consumes. Therefore, we have developed our algorithm based on a real dataset collected from the city of Milan, Italy, to show our approach for automating the task of identifying the best small cell network topology to implement for specific situations.
First, we designed an algorithm to adjust all types of call detail records (CDRs) for small cells at mmWave frequencies. Moreover, the information produced by this algorithm together with spatio-temporal mobile data reflected a pattern of user activity in our sample city.
Second, we compared k-means, density-based spatial clustering of applications with noise (DBSCAN), hierarchical algorithm, and two more clustering algorithms to find the best clustering method for small cell network topology. In addition, we developed an ant colony optimization algorithm to produce spatio-temporal mobile dataset and provide a novel small cell network planning solution.
Finally, we ascertained the best topology by using machine learning clustering and an optimization algorithm. Our topology came up with 2097 mmWave cell sites that covered 1,853.28 sq. km, 424 small cell sites for that covered 11,276.3 sq. km, and 25 macro cell sites that covered 22,090 sq. km. | en_US |
