本研究是以多層感知機類神經網路為基礎,以發展出一套氣喘發病之決策系統。 利用氣象及空氣環境因子及氣喘就醫資料,輸入層包括氣象及空氣因子,輸出層則為氣喘就醫人次,並以前饋倒傳遞演算法訓練,以交差驗證評估法來衡量。其結果顯示,特別的是以一週平均值來進行處理時,預測的準確率高達96%以上。 經由本研究,未來可以建置一氣喘發病預警網路服務,氣喘病患透過GPS裝置連接,服務中心伺服器可偵測病患所在位置,並即時取得該地監測站環境資料,經系統決策判斷,回傳給病患發病之危險警示,以期達到降低發病之可能。
A decision support system based on multiplayer perceptron was developed to indicate the possible morbidity of asthma diseases. The inputs of the system included indices of weather and air pollution. The outputs were levels of asthma outpatients. The cross validation assessment method was applied to assess the generalization of the system. In particular, we investigated the accuracy of the result, when medical records were handled using one week mean data. The results showed that the proposed MLP-based decision support system could achieve very high prediction accuracy (>96%). It is promising on indication of ambient air pollutants and weather for asthma patients. Furthermore, we builded a medical web service in this thesis as an application of the Global Positioning System (GPS). For an asthmatic patient, he or she can carry a portable GPS device to prevent possible morbidity of asthma during his or her outdoor activities of daily livings. To reduce possible allergy asthma, the GPS-enable device which sending user’s position continuously consults a remote server to decide whether current ambient air quality will threaten user’s health or not. The server of the detecting system collects real-time data from the network of national air quality monitoring stations. For the response to a remote query, the server makes decision of warning messages according to a proposed asthma neural network model. The proposed system is hopeful for asthmatic patients.