Applying Dynamic Adjusting NGHS-ANN in Predicting the Recidivism of Taiwanese Commuted Prisoners
| Autor: | Po-Chou Shih, 施柏州 |
|---|---|
| Rok vydání: | 2018 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 107 This research proposes a new algorithm called the Dynamic Adjusting Novel Global Harmony Search (DANGHS), which combines the concept of Dynamic Adjusting Parameters in a Novel Global Harmony Search (NGHS), and uses 14 famous benchmark continuous optimization problems to verify that the method proposed by this research is superior to 4 other algorithms. However, as there is more than one method for adjusting parameters, 16 dynamic adjustment strategies are also applied for experimental analysis. Then, taking the recidivism rate of Taiwanese commuted prisoners as the research topic, this research combined DANGHS and Artificial Neural Network (ANN) to construct a set of DANGHS-ANN prediction model for the recidivism rate of commuted prisoners. Subsequently, the k-fold method was adopted for repeated experiments to verify that the prediction model proposed by this research is superior to 5 other prediction models. Finally, 3 conclusions were found from the research data. (1) Different problems are applicable to different parameter adjustment strategies. (2) In terms of the DANGHS algorithm, as proposed by this research, its solving ability and efficiency are superior to other 4 algorithms. (3) For the DANGHS-ANN prediction model regarding the recidivism rate of commuted prisoners, as proposed by this research, its prediction error rate and robustness are superior to 5 other prediction models. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|