| Presentation | 2023-03-02 Feature Selection Method for Predicting Network Failures on CNF 5GC Using Machine Learning with Low Layer Log Data Takeru Hakii, Norihiro Fukumoto, Akihiro Nakao, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | In complex networks, once a failure occurs, it takes a long time to identify and recover from the cause of the failure, and the failure can be large enough to have a huge impact on society as a whole. However, storing all eBPF metrics and using them to train models increases the processing load and resource consumption of the models, which negatively affects performance and development efficiency. Therefore, we propose a feature selection method based on outlier processing using the ratio of the mean values under fault and normal conditions. The proposed method reduces the number of metrics from 3,325 to 320. We train a model that predicts whether a failure has occurred after 600 seconds of simulation using the proposed method, and compare its F1 score with that of a model that does not use the proposed method. The results show that the model with the proposed method predicts an F1 score of 0.93 at 140 seconds into the simulation, and an F1 score of 0.90 at 130 seconds. These results exceed those of the model without the proposed method. Therefore, our proposed method is not only effective in reducing the computational complexity by reducing the size of the model, but also in improving the accuracy and speed of prediction. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | 5G / CNF / core network / machine learning / failure prediction |
| Paper # | NS2022-192 |
| Date of Issue | 2023-02-23 (NS) |
| Conference Information | |
| Committee | IN / NS |
|---|---|
| Conference Date | 2023/3/2(2days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | Okinawa Convention Centre + Online |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | General |
| Chair | Kunio Hato(Internet Multifeed) / Tetsuya Oishi(NTT) |
| Vice Chair | Tsutomu Murase(Nagoya Univ.) / Takumi Miyoshi(Shibaura Insti of Tech.) |
| Secretary | Tsutomu Murase(KDDI Research) / Takumi Miyoshi(Nagaoka Univ. of Tech.) |
| Assistant | / Kotaro Mihara(NTT) |
| Paper Information | |
| Registration To | Technical Committee on Information Networks / Technical Committee on Network Systems |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Feature Selection Method for Predicting Network Failures on CNF 5GC Using Machine Learning with Low Layer Log Data |
| Sub Title (in English) | |
| Keyword(1) | 5G |
| Keyword(2) | CNF |
| Keyword(3) | core network |
| Keyword(4) | machine learning |
| Keyword(5) | failure prediction |
| Keyword(6) | |
| 1st Author's Name | Takeru Hakii |
| 1st Author's Affiliation | The University of Tokyo(UTokyo) |
| 2nd Author's Name | Norihiro Fukumoto |
| 2nd Author's Affiliation | The University of Tokyo(UTokyo) |
| 3rd Author's Name | Akihiro Nakao |
| 3rd Author's Affiliation | The University of Tokyo(UTokyo) |
| Date | 2023-03-02 |
| Paper # | NS2022-192 |
| Volume (vol) | vol.122 |
| Number (no) | NS-406 |
| Page | pp.pp.145-150(NS), |
| #Pages | 6 |
| Date of Issue | 2023-02-23 (NS) |