Publikationsserver
Machine Learning Methods for Fuzzy Pattern Tree Induction
This thesis elaborates on a novel approach to fuzzy machine learning, that is, the combination of machine learning methods with mathematical tools for modeling and information processing based on fuzzy logic. More specifically, the thesis is devoted to so-called fuzzy pattern trees, a model class th...
保存先:
| 第一著者: | |
|---|---|
| その他の著者: | |
| フォーマット: | Dissertation |
| 言語: | 英語 |
| 出版事項: |
Philipps-Universität Marburg
2014
|
| 主題: | |
| オンライン･アクセス: | PDFフルテキスト |
| タグ: |
タグ追加
タグなし, このレコードへの初めてのタグを付けませんか!
|
| 要約: | This thesis elaborates on a novel approach to fuzzy machine learning, that is, the combination of machine learning methods with mathematical tools for modeling and information processing based on fuzzy logic. More specifically, the thesis is devoted to so-called fuzzy pattern trees, a model class that has recently been introduced for representing dependencies between
input and output variables in supervised learning tasks, such as classification and regression. Due to its hierarchical, modular structure and the use of different types of (nonlinear) aggregation operators, a fuzzy pattern tree has the ability to represent such dependencies in a very exible and compact way, thereby offering a reasonable balance between accuracy and model transparency.
The focus of the thesis is on novel algorithms for pattern tree induction, i.e., for learning fuzzy pattern trees from observed data. In total, three new algorithms are introduced and compared to an existing method for the data-driven construction of pattern trees. While the first two algorithms are mainly geared toward an improvement of predictive accuracy, the last one focuses on eficiency aspects and seeks to make the learning process faster. The description and discussion of each algorithm is complemented with theoretical analyses and empirical studies in order to show the effectiveness of the proposed solutions. |
|---|---|
| DOI: | 10.17192/z2014.0391 |