부산대학교 도서관

In this paper, we propose a user centric network parameter optimization approach that utilizes the information contained in subscriber call detail records generated in a cellular network. To be able to maximize average user throughput on a cell level, we perform optimization of sector azimuth angles based on user centric weighted grids. The grid pattern is formulated by identifying spatially distributed points that correspond to user activity quantified by user location and service utilization information obtained from the call detail records. In this study, we present numerical and cell level simulation over a system of 30 cells to evaluate the proposed solution. In comparison to an optimal brute-force method that takes into account the location of every user to optimize the azimuth angles in a cell, we show that the proposed grid based self-optimization approach yields matching results in performance with substantial reduction in computational complexity.