| Week |
Tasks |
Completion
|
| Week 4
|
- Paper reading.
- News dataset and financial labels exploring.
|
✓
|
| Week 5
|
- Learning NLP data preprocessing.
- Finishing data preprocessing on New York Times dataset.
|
✓
|
| Week 6
|
- Initially defining the entire project's workflow and extracting key metrics.
- Finishing data preprocessing on New York Times dataset.
|
✓
|
| Week 7
|
- Completing the feature engineering construction and basic pipeline for the TF-IDF based model.
- Completing the feature engineering construction and basic pipeline for the sentiment dictionary-based model.
|
✓
|
| Weeks 8–9
|
- Modifying model details based on the sequential characteristics of time-based prediction.
- Improving various models, experimenting with different economic indicators to enhance accuracy.
|
✓
|
| Week 10
|
- Reading papers to understand how to integrate semantic connections into the sentiment analysis model.
- Prepare for the midterm presentation.
- Starting to write the Wikipedia page.
|
✓
|
| Week 11
|
- Reconstructing the dataset based on the suggestions from the mid-term report.
- Utilizing four different pre-trained models for text sentiment detection: DistilBERT-base-uncased, FinBERT, Twitter-roBERTa-base, and FinBERT-tone.
|
|
| Week 12
|
- Applying and tweaking other time series models to enhance model accuracy.
- Compare the results and analyze the correlations between sentiment scores and different financial indicators.
|
|
| Week 13
|
- Decision fusion for enhancing model accuracy.
- Achieving visual representation to display news trends, sentiment analysis outcomes, and predictive metrics in real-time.
- Depending on the situation: considering implementing incremental learning.
|
|
| Week 14
|
- Write the report.
- Prepare for the final presentation.
|
|