藥物濫用預警模型取樣與建置之研究

張智欽

指導教授 ： 曾淑萍

國立中正大學/教育學院/犯罪防治學系/碩士(2016年)

摘要

目的：本研究透過資料採礦來建立藥物濫用預警模型，提供研究者有效規則來判斷有／無使用毒品者，亦可做為日後從事相關研究者作為參考之用。
方法：本研究乃使用2009年「國民健康訪問暨藥物濫用調查國民健康訪問暨藥物濫用調查」，利用該調查進行次級資料分析，抽取研究樣本為年滿18歲以上之人口，共計16,723人。經由集群分析選擇藥物濫用高危險因子，再透過決策樹建立預警模型。
結果：以「有使用毒品」及「沒有使用毒品」1：2 之超取樣比例搭決策樹所建構模型，有最佳正確率，是為本次研究評估後之最佳模型，提供7條規則，相關因子為吸菸程度高 、熟友中曾有用毒、年齡小於 47 歲、第一次嚼檳榔年齡小於 26 歲、母為本省閩南人、家人親戚曾有用毒、認為藥物對身心有較少傷害、發現親友用毒採取規勸、第一次抽菸年齡小於 14 歲。
結論：本研究利用這些規則初步檢視一般民眾在藥物濫用情形，並搭配相關毒品與「入門物質」宣導，強化或導正對於藥物的認知，以提供制定相關決策之參考。

關鍵字

藥物濫用 ； 資料採礦 ； 預警模型

並列摘要

Objectives：The purpose of this study was to construct warning models of drug abuse through data mining. The warning models are effective in differentiating drug abusers from non-drug abusers as well as with references for the future selection of analytical tools.
Method：This study a secondary-data analysis was conducted with data derived from 2009 National Health Interview and Drug Abuse Survey; the data were collected by Bureau of Health Promotion, Department of Health, R.O.C. The subjects were consisted of 16,723 persons over 18 years old in the survey database. We employed cluster analysis to select the risk factors of drug abuse and obtained the warning model by using the methods of decision tree.
Results：The results showed that, the model that was constructed with decision tree’ analysis and were " drug abusers " and " non-drug abusers " of 1: 2 ratio was the best model in over-sampling. The model provided 7 rules and found that various factors were associated with drug abuse symptoms by applying C5.0 algorithms. Specifically, the model found that higher scores of smoking, have close friend used to take drug, less than 47 years old, the first contact with betel less than 26 years old, mother’s provincial being Taiwanese, having family or relative used to take drug, thinking that drugs have fewer physical and mental harm, persuading family member when finding they were using drug, the first contact with smoke less than 14 years old were the faction associated with drug-abuse symptoms.
Conclusion：This study utilized the rules to preliminary screen drug abuse. The model constructed in the study can be useful in strengthening and correcting incorrect cognition of drug abuse and can be used as reference of the authorities in making drug-abuse policies.

參考文獻

謝邦昌、張瓈文（民102）。資料採礦技術應用於全民健保資料庫分析腦中風病患死亡相關因素之研究。數據分析，8(6)，165-194。

童俊榮、劉志光（民102）。應用資料採礦技術分析臺灣健保資料庫－以攝護腺癌病人為例. 數據分析，8(5)，67-83。

侯藹玲、朱國豪、蘇志雄（民98）。資料採礦應用於乳癌患者之遺傳基因及生活因素探討。數據分析，4(3)，131-158.

李建明、李景美（民101）。臺北地區大學生社會人口學變項對成癮物質使用之研究。中華科技大學學報，51，149-159。

楊士隆、曾淑萍、戴伸峰（民100）。台灣地區收容少年入院前非法藥物使用盛行率調查研究。青少年犯罪防治研究期刊，3(2)， 73-104。