應用TSTO-HANN-GA框架建立心血管疾病風險之類神經網路精準預測模型

Taipei Medical University Institutional Repository > 醫學院 > 人工智慧醫療碩士在職專班 > 博碩士論文 > Item 987654321/65014

請使用永久網址來引用或連結此文件: http://libir.tmu.edu.tw/handle/987654321/65014

題名:	應用TSTO-HANN-GA框架建立心血管疾病風險之類神經網路精準預測模型 Utilizing the TSTO-HANN-GA Framework to Develop an Artificial Neural Network Model for the Precise Prediction of Cardiovascular Disease Risk
作者:	林家銘 Lin, Chia-Ming
貢獻者:	醫學院人工智慧醫療碩士在職專班林于翔
關鍵詞:	心血管疾病;兩階段田口優化方法;類神經網路;基因演算法;定點醫療診斷 cardiovascular disease;two-stage Taguchi optimization method;artificial neural network;genetic algorithm;point-of-care testing
日期:	2024-05-29
上傳時間:	2025-01-06 09:19:50 (UTC+8)
摘要:	心血管疾病（CVD）風險的早期預測對於預防治療至關重要。因此本研究目的在於使用個人電腦設備去提高預測心血管疾病之準確率，以期達到定點醫療診斷（Point-of-care testing, POCT）之目標。本研究提出了全新的TSTO-HANN-GA架構。此架構可持續不斷地調整類神經網絡（ANN）的超參數，並顯著地提高CVD預測之準確性。TSTO-HANN-GA架構整合了田口方法、類神經網絡與基因演算法（GA），可以更有效地探索超參數空間。相較於傳統的網格搜尋方法，它只需要少於40倍以上的實驗次數。因此，此架構適用於資源有限的環境，例如低功耗設備的環境。該架構成功地找出了ANN模型超參數的最佳設定：4個隱藏層、tanh激活函數、SGD優化器、0.23425849學習率、0.75462782動量率和7個隱藏節點數。此最佳設定在預測心血管疾病上達到了74.25%之平均準確率，此表現也較文獻所提的GA-ANN模型來的好。此一改善結果有望可以在定點醫療診斷（POCT）上進行客制化的心血管疾病預測，也讓每個人具有掌握自己健康的能力，進而對患者的健康產生重大影響。 Early and accurate prediction of cardiovascular disease risk is crucial for prevention and intervention. This study aims to enhance disease prediction accuracy using personal devices, aligning with point-of-care testing (POCT) objectives. This study introduces TSTO -HANN-GA, a novel framework that continuously refines hyperparameters for an Artificial Neural Network (ANN) model, significantly boosting its accuracy in CVD prediction. TSTO-HANN-GA leverages the Taguchi method, artificial neural network and genetic algorithm (GA) to efficiently explore hyperparameter space. Compared to traditional methods, it requires 40 times fewer experiments, making it suitable for resource-constrained environments, such as those involving low-power devices. The framework successfully identified the optimal configurations for the ANN model's hyperparameters, resulting in a hidden layer of 4, tanh activation function, SGD optimizer, learning rate of 0.23425849, momentum rate of 0.75462782, and 7 hidden nodes. This optimized model achieved an impressive average accuracy of 74.25%, outperforming existing GA-ANN models. This improvement has the potential to significantly impact patient outcomes by enabling personalized CVD prediction at the point of care, empowering individuals to take charge of their heart health.
描述:	碩士指導教授：林于翔口試委員：彭徐鈞口試委員：蘇家玉口試委員：林于翔
附註:	論文公開日期：2024-07-03
資料類型:	thesis
顯示於類別:	[人工智慧醫療碩士在職專班] 博碩士論文

文件中的檔案:

檔案	描述	大小	格式	瀏覽次數
index.html		0Kb	HTML	181	檢視/開啟

在TMUIR中所有的資料項目都受到原著作權保護.

TAIR相關文章

資料載入中.....