WEKO3
アイテム
尤度比を用いたモデル数削減と予測精度の維持
https://tama.repo.nii.ac.jp/records/912
https://tama.repo.nii.ac.jp/records/912a65d571b-e6de-40ae-a7f8-c30e87874cd2
| 名前 / ファイル | ライセンス | アクション |
|---|---|---|
201702100202 (10.9 MB)
|
|
| Item type | 紀要論文 / Departmental Bulletin Paper(1) | |||||
|---|---|---|---|---|---|---|
| 公開日 | 2017-05-18 | |||||
| タイトル | ||||||
| タイトル | 尤度比を用いたモデル数削減と予測精度の維持 | |||||
| タイトル | ||||||
| タイトル | Method of keeping model accuracy while reducing number of models by likelihood ratio test | |||||
| 言語 | en | |||||
| 言語 | ||||||
| 言語 | jpn | |||||
| キーワード | ||||||
| 言語 | en | |||||
| 主題Scheme | Other | |||||
| 主題 | Canonical regression analysis | |||||
| キーワード | ||||||
| 言語 | en | |||||
| 主題Scheme | Other | |||||
| 主題 | SVD | |||||
| キーワード | ||||||
| 言語 | en | |||||
| 主題Scheme | Other | |||||
| 主題 | Text Mining | |||||
| キーワード | ||||||
| 言語 | en | |||||
| 主題Scheme | Other | |||||
| 主題 | stock price prediction | |||||
| 資源タイプ | ||||||
| 資源タイプ識別子 | http://purl.org/coar/resource_type/c_6501 | |||||
| 資源タイプ | departmental bulletin paper | |||||
| 著者 |
津田, 高治
× 津田, 高治× 今泉, 忠× TSUDA, Takaharu× IMAIZUMI, Tadashi |
|||||
| 抄録 | ||||||
| 内容記述タイプ | Abstract | |||||
| 内容記述 | There are cases where news and its responses lead to debate and its shift explains socioeconomic changes. Tsuda（2015） introduced a method to: （1） numerically represent tendency of comments about chosen subject, and （2） extract differential features of news and its quotation in social media. In this paper, we propose an approach to: （1） predict stock price using the differential features, （2） keep predictive accuracy with reduced number of predictive models, and （3） adopt as simple method as possible for predictive models. So far many relevant studies have been carried out to predict stock price by statistical model utilizing publicly available text data such as those on internet, but while they have achieved predictive accuracy, they are not ready to use for real world purpose due primarily to lack of real-world applicability. In this paper we suggest an approach to apply likelihood ratio test to 2 periods of explanatory variables’ differential features to identify significant change between the 2 periods so as to limit frequency of rebuilding of predictive model, and the approach is applied to publicly available text data of news and its responses regarding Honda motors as well as its stock price data. The result turns out that while limiting number of models, predictive accuracy keeps as good as those reported in past relevant studies. It is also numerically represented that the more number of models the more accurate prediction becomes. |
|||||
| 書誌情報 |
経営・情報研究 多摩大学研究紀要 en : Tama University Journal of Management and Information Sciences 巻 21, p. 45-60, 発行日 2017-02-01 |
|||||
| 出版者 | ||||||
| 出版者 | 多摩大学経営情報学部 | |||||
| ISSN | ||||||
| 収録物識別子タイプ | ISSN | |||||
| 収録物識別子 | 13429507 | |||||
| 書誌レコードID | ||||||
| 収録物識別子タイプ | NCID | |||||
| 収録物識別子 | AA11140277 | |||||
