應用ANFIS於毒品施用傾向評估之研究

Research on Applying ANFIS to Risk Assessment of Drug Recidivism

曾淑萍(Shu-Ping Tzeng) ； 蔡修毓(Hsiu-Yu Tsai)

《犯罪學期刊》 13卷2期 (2010/12) Pp. 41-68

https://doi.org/10.29607/ZHWHGX.201012.0002

摘要

根據統計，近年受勒戒人釋放後於4年內再犯毒品相關罪名之比率高達33%~42%，肇致現行評估方式之準確性屢受質疑。本文以模糊理論(Fuzzy Theory)爲基礎，透過自組織映射圖(SOM)求取模糊規則數量，據此建構自適應類神經模糊推論系統(ANFIS)，以協助判斷受勒戒者是否再犯。針對台中看守所民國90年7月至12月出所之345名成年男性受勒戒人爲樣本進行驗證結果，本研究所建議方法在訓練組之正確率達94.20%，在測試組之正確率亦有83.80%，不僅正確率優於現行評估方式，且具有快速、客觀、自動化等優點，應有深入研究應用的價值。

關鍵字

自適應網路模糊推論系統 ； 自組織映射圖 ； 吸毒犯 ； 藥物濫用 ； 再犯危險評估

並列摘要

According to the official data, in recent years, the proportion that drug abusers receiving rehabilitation treatment recommit drug-related offenses within four years after release is about 33% to 42%. This high percentage has resulted in doubt about accuracy of the current assessment tool. Based on Fuzzy Theory, this study obtains the quantity of fuzzy rules through Self-Organizing MAP (SOM) and constructs an Adaptive Neuro-Fuzzy Inference System (ANFIS) with an attempt to assess whether the drug abusers would recidivate after rehabilitation treatment. This study utilized 345 male adult drug offenders receiving rehabilitation treatment who were released from Taichung Detention Center during July and December, 2001 as the sample. The assessment method proposed in this study reached an accuracy rate of 94.20% in the training group, and an accuracy rate of 83.80% in the test group. This suggested process does not only have a higher accuracy rate than the current assessment method, but it also has the advantages of swiftness, objectivity, and automation. Therefore, it is worthy of further research and application.

並列關鍵字

ANFIS ； SOM ； drug offender ； drug abuse ； risk assessment of recidivism