Page 153 - Kỷ yếu hội thảo quốc tế: Ứng dụng công nghệ mới trong công trình xanh - lần thứ 9 (ATiGB 2024)
P. 153

144                              TRƯỜNG ĐẠI HỌC SƯ PHẠM KỸ THUẬT - ĐẠI HỌC ĐÀ NẴNG





















                                                                   Fig. 6. The GUI interface of the proposed
                                                                    Python program for clustering TWAMP's
                                                                          performance data of BSs.
                                                                 On  the  graph  in  Fig.  7,  each  centroid  of  each
                       Fig. 4. A segment of data employed     cluster is visualized by the red dot.
                         for feeding into the algorithm.
                  4.2. Result analysis
                  In  this  section,  the  results  of  experiments  are
               shown.  Firstly,  the  optimal  cluster  number  is
               determined  by  k-means  clustering  and  the  Elbow
               method  as  a  plot  in  Fig.  5.  Then,  invisible
               performance patterns of MBH are exposed through k-
               means clustering algorithm with the optimal  k .





                                                                       Fig. 7. Visualization of clustering
                                                                        MBH dataset into six groups.
                                                                 The corresponding pair of values for each centroid
                                                              is calculated and summarized in Table 1.
                                                                    Table 1. Centroid values of each cluster.
                                                                 Order of centroids   TWAMP SR(%)  Packet Loss (#)
                                                               The 1st cluster’s centroid   100   41.66
                                                               The 2nd cluster’s centroid   0.0   863988.0
                                                               The 3rd cluster’s centroid   0.0   683069.0
                       Fig. 5. Elbow method applied into
                          MBH performance dataset.             The 4th cluster’s centroid   85.67   82423.0
                                                               The 5th cluster’s centroid   98.03   16729.42
                  In the plot above, we can visually perceive that the
               “Elbow” is the number 6 which is the optimum value   The 6th cluster’s centroid   99.61   3426.71
               of k for the MBH dataset. In case of k>6, the value of   Based  on  the  clustering  results  of  the  proposed
               SSE  does  not  change  significantly.  Therefore,  a  k-  optimal  k-means  model,  base  stations'  backhauls  in
               means algorithm using as k=6 will be executed in the   the  2nd  and  3rd  clusters  will  be  prioritized  for
               final stage. As shown in Fig. 6, the GUI interface of   processing  and  restoring  the  backhaul  transmission
               the  Python  program  implementing  the  proposed   quality  to  improve  overall  availability  of  mobile
               algorithm  flow  chart  depicted  in  Fig.  3  with  the   communication  services  delivered  to  subscribers.
               support of the Tkinter library is displayed.   MNOs always keep their backhaul network stable to
                                                              support  these  complex  heterogeneous  networks  with

               ISBN: 978-604-80-9779-0
   148   149   150   151   152   153   154   155   156   157   158