How zhn66776/tidal_try works
This repository is a notebook-based tide-level forecasting research project. Its central workflow is to prepare gauge observations, optionally derive UTide features, train BiLSTM or UTide-augmented UBiLSTM models, and compare predictions with observed levels. The README identifies the paper notebooks and data collections; sampled Python source confirms a separate CNN-BiLSTM training/evaluation path. Notebook internals were not sampled, so connections beyond the documented workflow are kept at the subsystem level.
Main entry points: Final_data Paper Data · run1.py CNN-BiLSTM Run · dataProcessed Processed Data · multi_utide25-2.ipynb UTide BiLSTM
Components 19
Tide Data
- Gauge Observations
dataProcessed - Processed Data
dataProcessed - Paper Data
Final_data - Experiment Data
all_data
Feature Preparation
- Data Preparation
dataprocess_code.ipynb - UTide Analysis
utide.ipynb - Windowed Datasets
utils/dataTools.py
Forecast Experiments
- Forecast Config
config.py - BiLSTM Forecast
multi25-2.ipynb - UTide BiLSTM
multi_utide25-2.ipynb - CNN-BiLSTM Run
run1.py - Neural Models
utils/models.py - Train And Evaluate
utils/trainer.py - Autoregressive Forecast
utils/trainer_autoreg.py
Evaluation And Analysis
- Model Comparison
notebook/compare_models.ipynb - UTide Comparison
notebook/compare_with_utide.ipynb - Data Visualization
data_visual.ipynb - Experiment Results
all_data
Other components
- Researcher
Connections 20
- Researcher → Data Preparation: runs analysis
- Gauge Observations → Data Preparation: reads observations
- Data Preparation → Processed Data: writes prepared data
- Processed Data → UTide Analysis: provides water levels
- UTide Analysis → Paper Data: produces paper data
- Processed Data → Windowed Datasets: loads series
- Windowed Datasets → CNN-BiLSTM Run: provides data splits
- Forecast Config → CNN-BiLSTM Run: supplies settings
- CNN-BiLSTM Run → Neural Models: instantiates model
- CNN-BiLSTM Run → Train And Evaluate: calls training and evaluation
- Paper Data → UTide BiLSTM: provides experiment data
- UTide BiLSTM → UTide Analysis: uses UTide features
- Paper Data → BiLSTM Forecast: provides experiment data
- Experiment Data → Model Comparison: supplies experiment data
- Experiment Data → UTide Comparison: supplies experiment data
- BiLSTM Forecast → Experiment Results: records experiment outputs
- UTide BiLSTM → Experiment Results: records experiment outputs
- Train And Evaluate → Autoregressive Forecast: supports iterative prediction
- Paper Data → Data Visualization: provides plotted data
- Experiment Results → Data Visualization: provides experiment results