-
austria_NNSI.ipynb: For reproducing legacy spatial interaction model in http://openjournals.wu.ac.at/region/paper_175_revised/175.html
-
SI.ipynb: Data collection and exploration for this project.
-
Taxi_NNSI.ipynb: Prepare NYC taxi data for NNSI and visualization
-
MainPlayground: Stage for running Spatial dependence SI model. (OLS/GLM/MLP)
-
LSTM_NNSI.ipynb: Stage for running Temporal dependence SI model. (OLS/MLP/RNN/LSTM)
- Clean notebooks
- Visualization from streamlit
- OLS/NN/RNN/LSTM SI model for flow time series
- Interpretation template for PDP/SHAP
- Interpretation template for counterfactual explanation
- tuning and compare RNN/LSTM SI model
- Real world-- Taxi data on NN run init result
- Real world-- Taxi data on NN run parameters comparison
- LSTM SI model
- Interpretation template of NN and LSTM model
- Markdown description
- Function definition position
- Clear legacy code in notebook
- Clear ERROR
- Use test data in comparison
- Upload data file
- Make simulated SI data
- Real world data run
Add mainPlayground notebook with all data/computation modules wrapped in function
study real SI data to get valid output in NN model.
Simple linear regression can be realized in Step0 neural network.
introduce NNSI
NYC Taxi/Bike data processing (SI.ipynb)
Gravity spatial interaction prototyping (austria.ipynb)
BP neural network prototyping (austria.ipynb)
Newly added Taxi_nn.ipynb:
Part 1
Processing and generate Data of NYC Taxi
Almost done. Is ready for use in part 2 and part 3
Part 2
Using PyTorch to implement LSTM model
Too advanced. Put away for now
Part 3 Manual simple NN
Tuning now Hand writing numpy based Neural Net with 3 input and 1 output features.
Data formating from Taxi dataframe to LSTM input features
(partial done)LSTM neural network prototyping (LSTM.ipynb)
Calibration of models
Explanation of models