Table 1.
Compare related work with our research.
Fig 1.
Extraction methodology of time series data for training the model.
Table 2.
Google trends keywords used to train the model.
Fig 2.
LSTM model structure.
Fig 3.
Seq2Seq model structure.
Fig 4.
Seq2Seq model structure.
Fig 5.
LSTM, Seq2Seq + Attention model structure (LSTM: Top, Seq2Seq + Attention: Bottom).
Fig 6.
Proposed model architecture.
Fig 7.
Learning error and verification data error according to the number of model trainings (LSTM).
Fig 8.
Learning error and verification data error according to the number of model trainings (Seq2Seq + Attention).
Table 3.
Hyperparameter of model.
Fig 9.
Week 1 prediction using LSTM model.
Fig 10.
Week 2 prediction using LSTM model.
Fig 11.
Week 3 prediction using LSTM model.
Fig 12.
Week 1 prediction using Seq2Seq model.
Fig 13.
Week 2 prediction using Seq2Seq model.
Fig 14.
Week 3 prediction using Seq2Seq model.
Table 4.
LSTM model test set prediction accuracy evaluation.
Table 5.
Seq2Seq model test set prediction accuracy evaluation.