my_project:PNC_Estimate

🚩 For example on a small dataset that can run on a personal laptop in example_test. The ./example_test/dataset/ can download from dataset.

1.Description

Machine Learning-Enhanced High-Resolution Exposure Assessment of Ultrafine Particles

2.File Structure

finalcode
├─ CAMS_downscale
│  ├─ CAMSpollutionDistribution.m
│  ├─ CAMS_downscale_plot.ipynb
│  ├─ cams_train.py
│  ├─ cams_train_pred.ipynb
│  ├─ downScaleConc.m
│  ├─ preprocessing.py
├─ PNC_estimate
│  ├─ diffplot.m
│  ├─ PNC_estimate_plot.ipynb
│  └─ swissPNCDistribution.m
├─ PNC_pred
│  ├─ model_train.ipynb
│  ├─ pnc_pred.py
│  ├─ PNC_pred_plot.ipynb
│  ├─ test_pnc_model
│  │  ├─ model_test.py
│  └─ trained model
├─ pop_PNC
│  ├─ pop_data_processing.py
│  ├─ pop_plot_final.ipynb
└─ SI_figure

3.File Function

2.1 CAMS_downscale

1. downScaleConc.m : If predictionFlag = 0 can get the pollution_trainData.csv, and if predictionFlag = 1 get 8760 hourId_pollutant_predData.bin files.
2. preprocessing.py : Contains functions cams_train.py may used.
3. cams_train.py : Use pollution_trainData.csv and hourId_pollutant_predData.bin files to the annual pollution high-resolution 1km grid data used machine learning methods.
4. cams_train_pred.ipynb: Get site-specific prediction results.
5. CAMS_downscale_plot.ipynb : Figure plot code for pollution compares.
6. CAMSpollutionDistribution.m : Figure plot code for DownScales pollution.

2.2 PNC_estimate

1. swissPNCDistribution.m : The code to plot 2020_DownScale_PNC.png.
2. PNC_estimate_plot.ipynb : The code to plot other figures.
3. diffplot.m : The code plot diff between matflies.

2.3 PNC_pred

1. test_pnc_model/model_test.py : DL code for pnc_pred.py.
2. pnc_pred.py : Get the pnc_estimate mat files.
3. model_train.ipynb : Train the PNC models.

2.4 pop_PNC

1. pop_data_processing.py: The functions pop_plot_final.ipynb used.
2. pop_plot_final.ipynb: Get the population related figures.

4.Software and Hardware Environment and Version

4.1 Hardware

4.1.1 Computer

• CPU： Intel(R) Core(TM) i7-10750H
• GPU： NVIDIA Geforce GTX 1650 Ti
• RAM： 16GB

4.1.2 Workstation CPU

• CPU： Intel(R) Xeon(R) Platinum 8383C
• GPU： NVIDIA Geforce GTX 3090
• RAM： 256GB

4.1.3 Workstation GPU

• CPU： Intel(R) Xeon(R) Platinum 8488C 2.40 GHz
• GPU： 4 x NVIDIA RTX A6000
• RAM： 512GB

4.1 Software

4.2.1 Computer

• Operating System： Windows 11
• Python Version： Anaconda Python 3.9.7
• CUDA Version： 11.7

4.2.2 Workstation CPU

• Operating System： Windows 10
• Python Version： Anaconda Python 3.11.4
• CUDA Version： 12.2

4.2.3 Workstation GPU

• Operating System： Windows 10
• Python Version： Python 3.12.6
• CUDA Version： 12.6

5.Usage

5.1 Environment configuration commands

• For convenience, execute the following command.

    pip install -r requirements.txt

5.2 Develop your own model.

• Start the dataset. A small dataset ./example_test/dataset/ can download from dataset as an example.

• Use the ./PNC_pred/model_train.ipynb to train your own model.

6.Citation

If you find this repo useful, please cite our paper. Yudie Jianyao, Machine Learning-Enhanced High-Resolution Exposure Assessment of Ultrafine Particles, 10.5281/zenodo.14554168 , 2024.

7.Contact

If you have any questions or suggestions, feel free to contact:

• Yudie Jianyao (Ph.D. student, [email protected])