Laser & Optoelectronics Progress, Volume. 58, Issue 4, 0410003(2021)
Self-Att-BiLSTM: A Multitask Prediction Method for Business Process Activities and Time
Figures & Tables(7)
Fig. 4. Comparison of prediction accuracy of repeated activities on different datasets
Fig. 5. Model training loss under different datasets. (a) BPI12; (b) BPI12_O; (c) BPI12_A; (d) BPI12_W; (e) Helpdesk
Table 1. Experimental parameters
View tableTable 1. Experimental parameters
Parameter Content Learning rate 0.001 Loss function 1 Categorical_crossentropy Loss function 2 MAE Optimizer Nadam Dropout 0.2 Activation function softmax, ReLU Number of LSTM units 50 Batch_size 128 Epoch 50
Table 2. Result comparison of different methods
View tableTable 2. Result comparison of different methods
Dataset Method Accuracy /% MAE /day BPI12 Self-Att-BiLSTM 80.57 1.39 Method in Ref.[16] Method in Ref.[9] BPI12_W Self-Att-BiLSTM 78.21 1.45 Method in Ref.[16] 76.00 1.56 Method in Ref.[9] 71.90 BPI12_O Self-Att-BiLSTM 81.64 2.36 Method in Ref.[16] Method in Ref.[9] 81.10 BPI12_A Self-Att-BiLSTM 81.33 0.99 Method in Ref.[16] Method in Ref.[9] 80.10 Helpdesk Self-Att-BiLSTM 81.13 2.05 Method in Ref.[16] 71.23 3.75 Method in Ref.[9]
Tools
Get Citation
Copy Citation Text
Qi He, Qiaoqing Yang, Dongmei Huang, Wei Song, Yanling Du. Self-Att-BiLSTM: A Multitask Prediction Method for Business Process Activities and Time[J]. Laser & Optoelectronics Progress, 2021, 58(4): 0410003
Paper Information
Category: Image Processing
Received: Jun. 19, 2020
Accepted: Aug. 3, 2020
Published Online: Feb. 24, 2021
The Author Email: Song Wei (wsong@shou.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20