資料探勘應用於憂鬱症與自殺之因素分析與預測

目的：本研究以臺北市政府照顧服務管理資訊平台的資料，利用資料探勘技術，建構憂鬱及負面想法和自傷行為及自殺（包含意念及行為）的特性模型，提供照管人員後續關懷追蹤之參考。方法：收集2018年個案資料，共7653筆，利用SPSS 18.0進行描述性統計分析個案基本特性（如年齡、性別、婚姻狀況等），再透過類神經網路計算出可能影響的變數，接著以SPSS Modeler 18版進行關聯規則Apriori的建模分析，最後使用Tableau資料分析軟體，將分析結果以視覺化呈現。結果：由本研究結果可知性別、高血壓、恐懼焦慮、年齡群組、照顧者性別、照顧者年齡群組、婚姻狀況、日夜顛倒以及教育程度等，為憂鬱及負面想法較需注意的變項，而關聯規則結果顯示女性、高齡、低教育程度、喪偶、罹患高血壓及恐懼焦慮等，是較常出現在關聯組合中之變項。而有關自殺（包含意念及行為）部分，較相關的變項為憂鬱及負面想法、性別、恐懼焦慮、婚姻狀況、照顧者性別、年齡群組、照顧者年齡群組、高血壓以及教育程度等，根據關聯規則結果顯示男性、已婚、女性照顧者、憂鬱及負面想法和恐懼焦慮等，是較常出現在關聯組合中之變項。結論：本研究為找出有關自殺（包含意念及行為）或憂鬱負面想法之特性，且需與系統結合才能發揮價值，故未來將積極與政府單位合作將挖掘出重要條件寫入平台中，往後若有相關申請人符合這些條件，則系統自動跳出警示，即可超前佈署給予適當關心與預防，減少不幸事件的發生。

關鍵字

長期照護；資料探勘；自殺；憂鬱； Tableau ；視覺化

並列摘要

Objective: By data mining techniques, this study used the data from the care service management information platform of the Taipei City Government to construct a characteristic model of negative thoughts, self-harm, and suicidal behaviors (including ideations and behaviors), to be used as a reference for caregivers in subsequent follow-up care. Method: We used the SPSS 18.0 to conduct descriptive statistical analysis of the basic characteristics (such as age, gender, marital status, etc.) of 7653 cases collected in 2018, and calculated possible influencing variables through the neural network, and used the SPSS Modeler 18 to conduct the apriori modeling of association rules. Lastly, we used Tableau data analysis software to visualize the results. Results: The results showed that variables that need attentions for depression/negative thoughts were gender, high blood pressure, fear and anxiety, age group, caregiver gender, caregiver age group, marital status, day and night reversal, and education levels. The association rule analysis showed that women, advanced age, low education level, widowed, suffering from high blood pressure, fear and anxiety, etc., were variables that often appeared in the association assemblies. For the suicide part, the more relevant variables were depression/ negative thoughts, gender, fear and anxiety, marital status, caregiver gender, age group, caregiver age group, hypertension and education level, etc. The association rule analysis showed that male, married, female caregivers, melancholic/ negative thoughts, fear and anxiety, etc., were the more common variables in the association assemblies. Conclusion: We identified characteristics related to suicide (including ideations and behaviors) or depression/negative thoughts. The findings are useful if incorporated into a central system. We will actively collaborate with the government to merge the essential information into the central platform so that alerts would jump out when applicants who meet the terms appear, and thus allow an earlier deployment to reduce unfortunate events.

並列關鍵字

long-term care ； data mining ； suicide ； depression ； tableau ； visualization

國際替代計量

資料探勘應用於憂鬱症與自殺之因素分析與預測

全文下載

主題瀏覽