Algorithm Study Based On Machine Learning In Handover Process Of Femtocell

With the popularization of 4G network, people pay more and more attention to the key technologies of LTE system. Femtocell emerges as the times require in order to improve the communication quality of the indoor user. Femtocell has the advantages of small volume, low power consumption, plug and play. But the coverage of Femtocell is limited and Femtocell is usually applied to the indoor environment. Because the indoor user frequently switches between Femtocell and Macrocell, the thesis focuses on the research of handover process from Femtocell to Macrocell and proposes an appropriate optimization algorithm.Due to relatively complex indoor wireless environment, the ping-pong effect occurs frequently in the handover process from Femtocell to Macrocell. In order to reduce the number of ping-pong handover, this thesis proposes the self-organizing map based on machine learning to optimize the handover process. According to the actual handover surroundings between Macrocell and Femtocell, this thesis proposes the self-organizing map based on the kernel function. The self-organizing map combines with the K-means, forming a kind of two-stage clustering algorithm based on machine learning. Through the optimization of the two-stage clustering algorithm, the handover system of Femtocell will can judge the region of mobile terminal independently and make the decision of the handover request according to the measurement report. Finally, the algorithm will improve the performance of the handover system of Femtocell and inhibit the most of unnecessary handover. At the same time, Femtocell can reduce the whole handover times and the drop phenomenon will decrease in handover process. Femtocell system reduces the signaling cost of the whole system and improves work efficiency of the system.