The Knowledge Discovery of the Nuclear Power Issue Using the Artificial Intelligence Model: An Example of the CART and the SVM

10.6138/JIT.2016.17.6.20160524

Yueh-Shiun Chung；Mei Hsiu-Ching Ho；Sun-Jen Huang；Shih-Wu Liang

Artificial intelligence ； Classification and regression tree ； Support vector machine ； Logistic regression ； Nuclear power issue

網際網路技術學刊

17卷6期（2016 / 11 / 01）

1071 - 1082

英文

After the nuclear accidents in the Fukushima Daiichi Nuclear Power Plant, the public has become increasingly concerned about nuclear power policies. We employed the CART and the SVM methods to establish and compare classification patterns in order to reveal obscure information in the survey data of Nuclear Power Policy^1. Our results showed that the total classification accuracy of the CART is higher than that of SVM or logistic regression. 48.87% of the Taiwanese public opposed nuclear power, 34.16% supported nuclear power, and 16.97% did not respond to the question. However, the CART analysis showed that among the non-respondents, 56% supported nuclear power and 44% opposed its use. The two primary factors influencing the opinion for nuclear power are ＂satisfaction with the safety of nuclear power＂ and ＂improvement in safety inspection and disaster prevention standards in nuclear power plants.＂ Therefore, by improving the safety of nuclear power plant, the government might regain public support for nuclear power.