| Summary: | 碩士 === 國立暨南國際大學 === 資訊管理學系 === 99 === The support vector machines (SVMs) has been widely used in the various fields, and that the classification performance and the accuracy rate has performed a satisfactory job in many research domains. Unfortunately, SVMs still be considered a 'black-box', because the SVMs could not prove or explained their solutions to approach. Therefore, in this paper proposes a method to extraction rule from SVMs, first of all, get the SVs and predicted label from the SVMs, and sequentially provide to RST mechanism to derive the comprehensive decision rules. To sound the model we proposed, we further compare the performance of raw data and SVs data, and the result undergo the ANOVA to examine the decision rules is significant or not.
The data used in this paper is to study the male adolescents in different social environments (such as family, school, etc.) and their psychological and physiological factors such as the results generated by the impact of deviant behaviors. To prevent the outlier or missing data deviate the result, the raw data undergo the data preprocessing procedure. Data preprocessed would eliminate the computation complexity and enhance the prediction performance. Sequentially, the SVM model utilized the preprocessed data to construct the prediction model. The parameters determination is very essential for SVM model construction, thus the meta-heuristic approach was utilized to discover the suitable parameters. To overcome the opaque nature of SVM, the investigation applied RST to yield comprehensive decision rules. The decision rules were easy for users to make a proper judgment. Finally, the decision rules undergo the ANOVA test to sound the findings.
|
|---|
