Cell Outage Detection And Flow Analysis In Hetnet

To meet the overwhelming requirement for data traffic and achieve dramatic coverage enhancement, broadband mobile communication technologies get increasingly improved. In LTE-Advanced system, a het-erogeneous networking strategy through complementing the mac-ro-base-station with low-power nodes is proposed, which is also called heterogeneous network (HetNet). The strategy can increase network ca-pacity substantially, while also causing great challenges on network plan-ning and maintenance. Operators desperately eager to provide the best service for users without additional operating expenses. Especially in HetNet, the small cell is more prone to failures due to its low physical size and small coverage features. One potential solution is to introduce the extensively attracted self-organizing network (SON), in which cell outage detection is able to detect anomaly and trigger outage compensa-tion functionality automatically. After all, It is really meaningful for net-work maintenance.In the meantime, the development for broadband mobile communi-cation also promotes the integration of mobile network and the Internet. In the mobile Internet, the traditional major business like voice and short message service have been replaced by abundant inexpensive Internet business, which results the proliferation of data traffic. While, operators do not see a breakthrough in the profit, and have possibility to be only a pipeline for data transmission. Fortunately, operators own mass reliable data that cannot be matched by other industries. To deal with the compe-tition from Internet, operators need to promote the development and ap-plication for big data, making the data pipeline more intelligent and prof-itable. Based on the above background, the article was committed to re-search on cell outage detection and flow analysis from theoretical and practical perspectives.In the HetNet, a small cell outage detection mechanism was pro-posed. The collection for performance indicators were triggered by peri-odical events. Firstly, KNN was implemented to construct a classifier based on training data. Then, the classifier was applied to classify the testing data into two categories, one stands for normal periodical data, the other denotes the anomaly. Further, localization for anomaly data and performance verification for classification accuracy were done to demon-strate the feasibility.In the practical operating networks where GSM and TD-SCDMA are co-existed, the network flow was analyzed through a clustering algorithm. Firstly, a typical research area was determined through statistical analysis about user complaints. Then, based on a certain network type (GSM/TD-SCDMA), a rasterized scheme against the network coverage was proposed, in order to reduce the analyzing complexity. Also, the network flow was mapped into each grid averagely. Finally, the k-means clustering algorithm was applied to classify the grid data into three cate-gories, in which the anomaly grids were particularly studied and a feasi-ble solution was proposed.