| Presentation | 2019-01-17 Study on Extraction of Important Data Based on Feature Selection Ensemble for Real-time Predictive Information Delivery Takumi Sakai, Ryoichi Shinkuma, Yuichi Inagaki, Takehiro Sato, Eiji Oki, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | Recently, the demands on the services that predict and deliver real-time spatial information, such as road-traffic volume, have been increasing. Mobile IoT devices that collect data for these services cannot necessarily transmit all of these data because of the bandwidth limitation in mobile networks. Some previous works have reduced the volume of data transmission while maintaining prediction accuracy by prioritizing data transmission based on data importance. They used feature selection methods to extract data importance from the machine learning model for the prediction. However, some feature selection methods may deteriorate the prediction accuracy because what feature selection method achieves the best prediction accuracy depends on what dataset is used. Therefore, this work proposes a method to ensemble data importance extracted by multiple feature selection methods to reduce the deterioration of the prediction accuracy. Evaluation with real-world datasets shows that the proposed system suppresses the deterioration of the prediction accuracy by using ensembled data importance. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | feature selection ensemble / IoT / machine learning / real-time prediction / priority control |
| Paper # | MoNA2018-67 |
| Date of Issue | 2019-01-09 (MoNA) |
| Conference Information | |
| Committee | MoNA |
|---|---|
| Conference Date | 2019/1/16(2days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | T. B. D. |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Chair | Ryoichi Shinkuma(Kyoto Univ.) |
| Vice Chair | Shigeaki Tagashira(Kansai Univ.) / Gen Kitagata(Tohoku Univ.) |
| Secretary | Shigeaki Tagashira(Kyushu Univ.) / Gen Kitagata(NEC) |
| Assistant | Ken Usui(KDDI Research) / Kenji Kanai(Waseda Univ.) |
| Paper Information | |
| Registration To | Technical Committee on Mobile Network and Applications |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Study on Extraction of Important Data Based on Feature Selection Ensemble for Real-time Predictive Information Delivery |
| Sub Title (in English) | |
| Keyword(1) | feature selection ensemble |
| Keyword(2) | IoT |
| Keyword(3) | machine learning |
| Keyword(4) | real-time prediction |
| Keyword(5) | priority control |
| 1st Author's Name | Takumi Sakai |
| 1st Author's Affiliation | Kyoto University(Kyoto Univ) |
| 2nd Author's Name | Ryoichi Shinkuma |
| 2nd Author's Affiliation | Kyoto University(Kyoto Univ) |
| 3rd Author's Name | Yuichi Inagaki |
| 3rd Author's Affiliation | Kyoto University(Kyoto Univ) |
| 4th Author's Name | Takehiro Sato |
| 4th Author's Affiliation | Kyoto University(Kyoto Univ) |
| 5th Author's Name | Eiji Oki |
| 5th Author's Affiliation | Kyoto University(Kyoto Univ) |
| Date | 2019-01-17 |
| Paper # | MoNA2018-67 |
| Volume (vol) | vol.118 |
| Number (no) | MoNA-389 |
| Page | pp.pp.57-61(MoNA), |
| #Pages | 5 |
| Date of Issue | 2019-01-09 (MoNA) |