這篇論文中，我們使用三種機器學習的方法去預測行動用戶和哪些基地台連線：分別為網路嵌入方法、Principal Components Analysis (PCA)及Long Short-Term Memory (LSTM)。論文中的實驗僅使用參考訊號接收功率 (Reference Signal Received Power, RSRP)作為訓練資料。為了使實驗更貼近真實狀況，我們實際在清華大學附近的社區收集訊號資訊，一共收集24344筆RSRP值。
使用Simulation of Urban Mobility (SUMO) 軟體生成行動用戶的移動軌跡。我們將行動用戶前30秒的RSRP資訊當成訓練集 (training set)，第31秒的RSRP資訊當成測試集 (testing set)。根據實驗結果，在所有的機器學習方法之中，LSTM的方法表現最好。

In this thesis, we consider the link prediction problem that predicts the connections between mobile users
and base stations. For such a problem, we use three machine learning methods: network embedding approach, Principal Components Analysis (PCA), and Long Short-Term Memory (LSTM). Only the Reference Signal Received Power (RSRP) is used as the training data. In order to make the experiment closer to the real situation, we collected signal information in a community near the National Tsing Hua University. The dataset consists of a total of 24344 RSRP values. The traffic pattern of mobile users is generated by using the Simulation of Urban Mobility (SUMO) software. We regard the RSRP information of the mobile users in the first 30 seconds as the training set and the RSRP information in the 31st second as the testing set. According to the experimental results, the LSTM method performs the best among all the three machine learning methods.

Contents 1
List of Figures 3
1 Introduction 4
2 Data Collection 8
3 Machine Learning Approaches 16
3.1 Network Embedding Approach 16
3.1.1 Construction of a similarity matrix 16
3.1.2 The CAFE algorithm 20
3.1.3 Using the embedding vectors for the link prediction problem 23
3.2 Principal Components Analysis (PCA) 24
3.3 Long Short-term Memory (LSTM) 25
4 Experiments 27
4.1 Last RSRP 28
4.2 Ridge (L=2) 28
4.3 Distance CAFE (L=2, 3) 28
4.4 Jaccard CAFE (L=2, 3) 29
4.5 Distance PCA (6, L=2) and Jaccard PCA (6, L=2) 29
4.6 LSTM 30
4.7 Performance Metrics 30
5 Conclusion 35
A RSRP distributions for each PCI with each ARFCN 39

[1] Y.-S. Chen, Efficient Algorithms for Multiple Concentrator Placement and Handover Management in 5G Mobile Networks. Ph.D. Dissertation, Inst. Commun. Eng., National Tsing Hua Univ., Hsinchu, Taiwan, June 2022.
[2] B. Mahdy, H. Abbas, H. S. Hassanein, A. Noureldin, and H. Abou-zeid, “A clusteringdriven approach to predict the traffic load of mobile networks for the analysis of base stations deployment,” Journal of Sensor and Actuator Networks, vol. 9, no. 4, p. 53, 2020.
[3] C.-H. Yeh, A Network Embedding Approach for Tracking and Predicting Mobile Users. M.S. thesis, Inst. Commun. Eng., National Tsing Hua Univ., Hsinchu, Taiwan, Aug. 2021.
[4] P.-E. Lu, C.-H. Yeh, and C.-S. Chang, “Explainable, stable, and scalable network embedding algorithms for unsupervised learning of graph representations,” IEEE Transactions on Computational Social Systems, 2022.
[5] “Lte; evolved universal terrestrial radio access (e-utra); physical layer; measurements
(3gpp ts 36.214 version 11.1.0 release 11),” 2013.
[6] M. Tayyab, X. Gelabert, and R. Jantti, “A survey on handover management: From lte to nr,” IEEE Access, vol. 7, pp. 118 907–118 930, 2019.
[7] S. Hochreiter and J. Schmidhuber, “Long short-term memory,” Neural computation, vol. 9, no. 8, pp. 1735–1780, 1997.
[8] “Lte; evolved universal terrestrial radio access (e-utra); user equipment (ue) radio transmission and reception (3gpp ts 36.101 version 14.3.0 release 14),” 2017.
[9] M. Belkin and P. Niyogi, “Laplacian eigenmaps and spectral techniques for embedding and clustering,” in Advances in neural information processing systems, 2002, pp. 585–591.
[10] A. Ahmed, N. Shervashidze, S. Narayanamurthy, V. Josifovski, and A. J. Smola, “Distributed large-scale natural graph factorization,” in Proceedings of the 22nd international
conference on World Wide Web, 2013, pp. 37–48.
[11] S. Cao, W. Lu, and Q. Xu, “Grarep: Learning graph representations with global structural information,” in Proceedings of the 24th ACM International on Conference on Information and Knowledge Management. ACM, 2015, pp. 891–900.
[12] M. Ou, P. Cui, J. Pei, Z. Zhang, and W. Zhu, “Asymmetric transitivity preserving graph embedding,” in Proceedings of the 22nd ACM SIGKDD international conference on
Knowledge discovery and data mining, 2016, pp. 1105–1114.
[13] B. Perozzi, R. Al-Rfou, and S. Skiena, “Deepwalk: Online learning of social representations,” in Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining. ACM, 2014, pp. 701–710.
[14] A. Grover and J. Leskovec, “node2vec: Scalable feature learning for networks,” in Proceedings of the 22nd ACM SIGKDD international conference on Knowledge discovery and data mining. ACM, 2016, pp. 855–864.
[15] P.-E. Lu and C.-S. Chang, “Explainable, stable, and scalable graph convolutional networks for learning graph representation,” arXiv preprint arXiv:2009.10367, 2020.
[16] T. N. Kipf and M. Welling, “Semi-supervised classification with graph convolutional networks,” arXiv preprint arXiv:1609.02907, 2016.
[17] C.-S. Chang, C.-C. Huang, C.-T. Chang, D.-S. Lee, and P.-E. Lu, “Generalized modularity embedding: a general framework for network embedding,” arXiv preprint arXiv:1904.11027, 2019.
[18] P. A. Lopez, M. Behrisch, L. Bieker-Walz, J. Erdmann, Y.-P. Flotter ¨ od, R. Hilbrich, ¨ L. Lucken, J. Rummel, P. Wagner, and E. Wießner, “Microscopic traffic simulation using ¨sumo,” in The 21st IEEE International Conference on Intelligent Transportation Systems. IEEE, 2018. [Online]. Available: https://elib.dlr.de/124092/