| Presentation | 2019-12-05 [Poster Presentation] Quality state analysis of eNodeB log data by semi-supervised learning using Self training Shouta Yoshida, Atsushi Morohoshi, Kohei Shiomoto, Chin Lam Eng, Sebastian Backstad, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | In an LTE network where traffic is increasing year by year. It is important to quickly find the cause when a failure occurs in the eNodeB base station. Therefore, using the eNodeB log data extracted from the base station, 13 types of base station states are clasified by machine learning, and high accuracy is achieved while a few labeled data that is expensive to create. In this paper, semi-supervised learning is performed by using self-training, which considers unlabeled data with high confidence as label data, and the accuracy is improved from 86.73％ to 90.13％ compared to normal supervised learning. Also, using two types of methods, Active learning to add labels to data with low confidence, and unlabeled loss function, which is an effective loss function for unlabeled data, Accuracy has improved to 90.96％. The advantage of semi-supervised learning in learning with a few label data was clarified. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | machine learning / semi-supervised learning / Self training / Active learning / unlabeled loss function |
| Paper # | SR2019-92 |
| Date of Issue | 2019-11-28 (SR) |
| Conference Information | |
| Committee | SR |
|---|---|
| Conference Date | 2019/12/5(2days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | Ishigaki City Hall (Ishigaki Island) |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | cognitive radio, machine learning application, heterogeneous network, SDN, IoT etc. |
| Chair | Masayuki Ariyoshi(NEC) |
| Vice Chair | Suguru Kameda(Tohoku Univ.) / Osamu Takyu(Shinshu Univ.) / Kentaro Ishidu(NICT) |
| Secretary | Suguru Kameda(ATR) / Osamu Takyu(Univ. of Electro-Comm.) / Kentaro Ishidu(Mie Univ.) |
| Assistant | Mai Ohta(Fukuoka Univ.) / Teppei Oyama(Fujitsu Lab.) / Kentaro Kobayashi(Nagoya Univ.) |
| Paper Information | |
| Registration To | Technical Committee on Smart Radio |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | [Poster Presentation] Quality state analysis of eNodeB log data by semi-supervised learning using Self training |
| Sub Title (in English) | |
| Keyword(1) | machine learning |
| Keyword(2) | semi-supervised learning |
| Keyword(3) | Self training |
| Keyword(4) | Active learning |
| Keyword(5) | unlabeled loss function |
| 1st Author's Name | Shouta Yoshida |
| 1st Author's Affiliation | Tokyo City University(TCU) |
| 2nd Author's Name | Atsushi Morohoshi |
| 2nd Author's Affiliation | Fujitsu Fsas Inc.(Fujitsu Fsas) |
| 3rd Author's Name | Kohei Shiomoto |
| 3rd Author's Affiliation | Tokyo City University(TCU) |
| 4th Author's Name | Chin Lam Eng |
| 4th Author's Affiliation | Ericsson Japan(Ericsson Japan) |
| 5th Author's Name | Sebastian Backstad |
| 5th Author's Affiliation | Ericsson Japan(Ericsson Japan) |
| Date | 2019-12-05 |
| Paper # | SR2019-92 |
| Volume (vol) | vol.119 |
| Number (no) | SR-325 |
| Page | pp.pp.29-36(SR), |
| #Pages | 8 |
| Date of Issue | 2019-11-28 (SR) |