Skripsi
PREDIKSI POLA HENTI JANTUNG BERDASARKAN DELAPAN DATA TANDA VITAL MENGGUNAKAN METODE RECURRENT NEURAL NETWORK
The ability to predict patterns of Cardiac Arrest with high accuracy can significantly contribute to more effective prevention and treatment. This research focuses on predicting patterns of Cardiac Arrest based on eight vital sign data using Recurrent Neural Network (RNN) methods with LSTM and Bi-LSTM architectures. In order to address the issue of imbalanced data, undersampling techniques and class weighting with cost-sensitive learning approach are employed. To fill in missing values in the dataset, this study utilizes Linear Interpolation as well as Deep Learning techniques such as Autoencoder and U-Net for data imputation. The best performance is achieved by the cost-sensitive Bi-LSTM (CSBi-LSTM) model without undersampling the majority class. Linear Interpolation is applied for data imputation with a total data duration and prediction range of 60 minutes. The evaluation results of the CSBi-LSTM model on accuracy, sensitivity, precision, f1-score, and specificity metrics are 95%, 95%, 5%, 10%, and 100% respectively.
Inventory Code | Barcode | Call Number | Location | Status |
---|---|---|---|---|
2307002410 | T113569 | T1135692023 | Central Library (Referens) | Available |
No other version available