GNNWR

https://github.com/zjuwss/gnnwr
Admin
GNNWR is a PyTorch-based package for spatiotemporal intelligent regression modeling, designed to...
Tokens:29,302
Snippets:201
Trust Score:6
License:GPL-3.0
Update:3 days ago
Show doc for...
Tokens:
Raw
### Initialize dataset and run GNNWR model

Source: https://github.com/zjuwss/gnnwr/blob/main/README.md

Example of loading data with pandas, initializing the dataset with spatial and target columns, and running the GNNWR regression model.

```python
from gnnwr import models,datasets
import pandas as pd

data = pd.read_csv('your_data.csv')

train_dataset, val_dataset, test_dataset = datasets.init_dataset(data=data,
                                                                 test_ratio=0.2, valid_ratio=0.1,
                                                                 x_column=['x1', 'x2'], y_column=['y'],
                                                                 spatial_column=['u', 'v'])

gnnwr = models.GNNWR(train_dataset, val_dataset, test_dataset)

gnnwr.run(100)
```

--------------------------------

### Install GNNWR via pip

Source: https://github.com/zjuwss/gnnwr/blob/main/README.md

Standard installation command for the GNNWR package.

```bash
pip install gnnwr
```

--------------------------------

### Verify PyTorch CUDA installation

Source: https://github.com/zjuwss/gnnwr/blob/main/README.md

Check the installed version of PyTorch to ensure it includes CUDA support for GPU acceleration.

```bash
> pip list | grep torch
torch                   1.13.1+cu117
```

--------------------------------

### Initialize GNNWR model with configuration

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_zh_CN.ipynb

Creates a GNNWR model instance with specified dense layers, learning rate, optimizer, activation function, and paths for saving model, logs, and writes.

```python
gnnwr_load = models.GNNWR(train_dataset=train_dataset_load,
                          valid_dataset=val_dataset_load,
                          test_dataset=test_dataset_load,
                          dense_layers=[512, 256, 128],
                          start_lr=0.2,
                          optimizer="Adadelta",
                          activate_func=nn.PReLU(init=0.1),
                          model_name="GNNWR_PM25",
                          model_save_path="../demo_result/gnnwr/models/",
                          log_path="../demo_result/gnnwr/logs/",
                          write_path="../demo_result/gnnwr/writes/"
                          )
```

--------------------------------

### Initialize prediction dataset

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_zh_CN.ipynb

Creates a prediction dataset using datasets.init_predict_dataset with the prediction data, training dataset, feature columns, and spatial columns.

```python
pred_dataset = datasets.init_predict_dataset(data=pred_data, train_dataset=train_dataset, x_column=[
                                             'dem', 'w10', 'd10', 't2m', 'aod_sat', 'tp'],
                                             spatial_column=['proj_x', 'proj_y'])
```

--------------------------------

### Load and Preview Data from CSV

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gtnnwr_en_US.ipynb

Reads the demo data CSV file and displays the first few rows to inspect the dataset structure.

```python
data = pd.read_csv('../data/demo_data_gtnnwr.csv')
data.head()
```

```text
Result:
   refl_b01  refl_b02  refl_b03  refl_b04  refl_b05  refl_b07       proj_x  \
0      1895      1251      1737      1872       543       280  736685.4179   
1      1621      1066      1435      1665       757       511  736685.4179   
2      1587      1141      1807      1864      1211      1024  736685.4179   
3      1121      1012      1411      1324       711       485  768193.4945   
4       913       203       840      1035       114        63  736685.4179   

        proj_y  day   SiO3  
0  3314205.295  113  1.049  
1  3314205.295  121  1.209  
2  3314205.295  152  1.102  
3  3402138.581  168  0.945  
4  3314205.295  200  1.350  
```

--------------------------------

### Initialize Dataset with init_dataset

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gtnnwr_en_US.ipynb

Adds an ID column and initializes training, validation, and test datasets using the init_dataset function with specified ratios, feature columns, spatial and temporal columns, and batch size.

```python
data["id"] = np.arange(len(data))
train_dataset, val_dataset, test_dataset = init_dataset(data=data,
                                                        test_ratio=0.15,
                                                        valid_ratio=0.1,
                                                        x_column=['refl_b01', 'refl_b02',
                                                                  'refl_b03', 'refl_b04', 'refl_b05',
                                                                  'refl_b07'],
                                                        y_column=['SiO3'],
                                                        spatial_column=[
                                                            'proj_x', 'proj_y'],
                                                        temp_column=['day'],
                                                        id_column=['id'],
                                                        use_model="gtnnwr",
                                                        sample_seed=48,
                                                        batch_size=1024)
```

--------------------------------

### Initialize Visualizer with GNNWR data

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_zh_CN.ipynb

Creates a Visualizer instance using the GNNWR data object, specifying longitude and latitude columns.

```python
visualizer = utils.Visualize(data=gnnwr, lon_lat_columns=['lng', 'lat'])
```

--------------------------------

### Initialize dataset with init_dataset

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_zh_CN.ipynb

Initializes train, validation, and test datasets using the init_dataset function. Specifies feature columns (dem, w10, d10, t2m, aod_sat, tp), target column (PM2_5), spatial coordinates, and a batch size of 64.

```python
train_dataset, val_dataset, test_dataset = datasets.init_dataset(data=data,
                                                                 test_ratio=0.15,
                                                                 valid_ratio=0.15,
                                                                 x_column=[
                                                                     'dem', 'w10', 'd10', 't2m', 'aod_sat', 'tp'],
                                                                 y_column=[
                                                                     'PM2_5'],
                                                                 spatial_column=[
                                                                     'proj_x', 'proj_y'],
                                                                 sample_seed=23,
                                                                 batch_size=64)
```

--------------------------------

### Create 3D scatter plots for reflection coefficients

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gtnnwr_zh_CN.ipynb

Generates a 3x3 grid of 3D scatter plots showing 'coef_refl_b0' coefficients (for bands 1,2,3,4,5,7) colored by value, with axes proj_x, proj_y, and day. The figure size is 30x30 inches.

```python
fig = plt.figure(figsize=(30, 30))
plt.rcParams['font.sans-serif'] = ['Arial Unicode MS']
id = 1
for i in [1, 2, 3, 4, 5, 7]:
    ax = fig.add_subplot(3, 3, id, projection='3d')
    fig.colorbar(ax.scatter(ana_data['proj_x'], ana_data['proj_y'], ana_data['day'],
                            c=ana_data['coef_refl_b0'+str(i)], cmap='GnBu'),
                 shrink=0.8, aspect=30, orientation='horizontal', pad=0.05)
    ax.set_xlabel('proj_x')
    ax.set_ylabel('proj_y')
    ax.set_zlabel('day')
    ax.set_title('coef_refl_b0'+str(i), fontsize=15, fontweight='bold')
    id += 1
```

--------------------------------

### Display dataset 'all' with PM2.5 values

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_zh_CN.ipynb

Displays the full dataset on a map, coloring points by PM2.5 values using a blue-green-yellow-red gradient.

```python
visualizer.display_dataset(name='all', y_column='PM2_5', colors=[
                           'blue', 'green', 'yellow', 'red'])
```

--------------------------------

### Display dataset 'test' with PM2.5 values

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_zh_CN.ipynb

Displays the test dataset on a map, coloring points by PM2.5 values.

```python
visualizer.display_dataset(name='test', y_column='PM2_5', colors=[
                           'blue', 'green', 'yellow', 'red'])
```

--------------------------------

### Display dataset 'train' with PM2.5 values

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_zh_CN.ipynb

Displays the training dataset on a map, coloring points by PM2.5 values.

```python
visualizer.display_dataset(name='train', y_column='PM2_5', colors=[
                           'blue', 'green', 'yellow', 'red'])
```

--------------------------------

### Create GNNWR model and add graph

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_zh_CN.ipynb

Creates a GNNWR model with specified dense layers, learning rate, optimizer (Adadelta), activation function (PReLU), and output paths. Then calls add_graph() to build the computational graph, which prints 'Add Graph Successfully'.

```python
gnnwr = models.GNNWR(train_dataset=train_dataset,
                     valid_dataset=val_dataset,
                     test_dataset=test_dataset,
                     dense_layers=[512, 256, 128],
                     start_lr=0.2,
                     optimizer="Adadelta",
                     activate_func=nn.PReLU(init=0.1),
                     model_name="GNNWR_PM25",
                     write_path="../demo_result/gnnwr/runs/",
                     model_save_path="../demo_result/gnnwr/models/",
                     log_path="../demo_result/gnnwr/logs/",
                     )
gnnwr.add_graph()
```

```text
Output:
Add Graph Successfully
```

--------------------------------

### Load trained GNNWR model and print results

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_zh_CN.ipynb

Loads a previously trained GNNWR model from a .pkl file and prints the result, which includes model information and performance metrics.

```python
gnnwr_load.load_model('../demo_result/gnnwr/models/GNNWR_PM25.pkl')
print(gnnwr_load.result())
```

--------------------------------

### Configure and Instantiate GTNNWR Model with add_graph

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gtnnwr_en_US.ipynb

Sets optimizer parameters and creates a GTNNWR model instance with specified architecture, dropout, optimizer, and save paths. The add_graph() method saves the model graph for TensorBoard.

```python
optimizer_params = {
    "scheduler": "MultiStepLR",
    "scheduler_milestones": [1000, 2000, 3000, 4000],
    "scheduler_gamma": 0.8,
}
gtnnwr = GTNNWR(train_dataset, val_dataset, test_dataset, [[3], [512, 256, 64]], drop_out=0.4, optimizer='Adadelta', optimizer_params=optimizer_params,
                write_path="../demo_result/gtnnwr/runs/",  # 这里需要修改
                model_save_path="../demo_result/gtnnwr/models/",
                log_path="../demo_result/gtnnwr/logs/",
                model_name="GTNNWR_DSi")
gtnnwr.add_graph()
```

```text
Output:
Add Graph Successfully

```

--------------------------------

### Run GNNWR Training

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_en_US.ipynb

Executes the training process for the initialized GNNWR model with specified epochs and early stopping criteria.

```python
gnnwr.run(max_epoch=200, early_stop=50)
```

```text
Output:
100%|██████████| 200/200 [00:29<00:00,  6.71it/s, Train Loss=55.007241, Train R2=0.726874, Train AIC=6957.4873, Valid Loss=58.3, Valid R2=0.692, Best Valid R2=0.768, Learning Rate=0.2]
```

--------------------------------

### Import GTNNWR and Required Libraries

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gtnnwr_en_US.ipynb

Imports the GTNNWR model, dataset initialization function, and common data processing libraries (numpy, pandas).

```python
from gnnwr.models import GTNNWR
from gnnwr.datasets import init_dataset
import numpy as np
import pandas as pd
```

--------------------------------

### Load PM2.5 training data

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_en_US.ipynb

Reads the PM2.5 dataset from '../data/pm25_data.csv' using pandas and displays the first 5 rows. The output shows columns including station_id, lng, lat, date, PM2_5, and various environmental features.

```python
data = pd.read_csv('../data/pm25_data.csv')
data.head(5)
```

```text
Result:
  station_id      lng      lat      date      PM2_5  row_index  col_index  \
0      1001A  116.366  39.8673  20170601  54.733894       2201       6867   
1      1002A  116.170  40.2865  20170601  48.080737       2134       6835   
2      1003A  116.434  39.9522  20170601  54.898592       2188       6877   
3      1004A  116.434  39.8745  20170601  52.266382       2200       6877   
4      1005A  116.473  39.9716  20170601  53.189076       2185       6884   

        proj_x       proj_y  dem  ...         t2m            sp        tp  
0  1650847.552  1370268.366   46  ...  284.561066  100809.27340  0.001006   
1  1625003.973  1416959.964  420  ...  282.907684   97125.08594  0.001044   
2  1653776.710  1381524.305   48  ...  284.492249  100830.96880  0.001002   
3  1655828.045  1372270.098   45  ...  284.636200  100936.80470  0.001010   
4  1656224.681  1384491.842   40  ...  284.506561  100880.17970  0.001019   

          blh         e          r       u10       v10   aod_sat  ndvi  
0  134.995636 -0.000007  46.315975  0.425366  0.170262  0.870967  2401  
1  157.775970 -0.000006  53.605503  0.211734 -0.676848  0.712080  5255  
2  129.971298 -0.000007  45.537464  0.266666  0.069172  0.875811  2609  
3  138.793961 -0.000007  45.387913  0.299403  0.227950  0.869679  2420  
4  130.520599 -0.000007  44.790119  0.169121  0.079546  0.873232  3296  

[5 rows x 22 columns]
```

--------------------------------

### Import gnnwr and dependencies

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_en_US.ipynb

Imports the gnnwr library modules (models, datasets, utils) along with pandas, numpy, folium, torch.nn, sklearn.metrics.r2_score, and matplotlib.pyplot for the PM2.5 estimation workflow.

```python
from gnnwr import models, datasets, utils
import pandas as pd
import numpy as np
import folium
import torch.nn as nn
from sklearn.metrics import r2_score as r2
import matplotlib.pyplot as plt
```

--------------------------------

### Display dataset distribution for 'all'

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_en_US.ipynb

Calls visualizer.display_dataset with name='all' and y_column='PM2_5' to generate a folium map. The output is a folium Map object.

```python
visualizer.display_dataset(name='all', y_column='PM2_5', colors=[
                           'blue', 'green', 'yellow', 'red'])
```

```text
Result:
<folium.folium.Map at 0x1d9b771e350>
```

--------------------------------

### Print model results with result()

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_zh_CN.ipynb

Prints the model information and evaluation metrics, including OLS coefficients, test loss, R2, RMSE, AIC, and other parameters.

```python
print(gnnwr.result())
```

```text
Output:
--------------------Model Information-----------------
Model Name:           | GNNWR_PM25
independent variable: | ['dem', 'w10', 'd10', 't2m', 'aod_sat', 'tp']
dependent variable:   | ['PM2_5']

OLS coefficients: 
x0: 4.84385
x1: -8.85402
x2: 0.17612
x3: 18.69575
x4: 35.19490
x5: -23.06419
Intercept: 22.06353

--------------------Result Information----------------
Test Loss: |                  55.22187
Test R2  : |                   0.74461
Train R2 : |                   0.80714
Valid R2 : |                   0.76311
RMSE: |                        7.43114
AIC:  |                     1451.93323
AICc: |                     1453.84448
F1:   |                        0.23711
F2:   |                       -0.72962
f3_param_0: |                  1.42917
f3_param_1: |                  4.78019
f3_param_2: |                  0.37658
f3_param_3: |                 39.44110
f3_param_4: |                 86.51201
f3_param_5: |                103.20709
f3_param_6: |                 73.79106
```

--------------------------------

### Heatmap for coefficient 'coef_w10'

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_zh_CN.ipynb

Generates a heatmap for the 10-meter wind speed coefficient.

```python
visualizer.coefs_heatmap('coef_w10', colors=['blue', 'green', 'yellow', 'red'])
```

--------------------------------

### Display dataset distribution for 'test'

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_en_US.ipynb

Calls visualizer.display_dataset with name='test' and y_column='PM2_5' to generate a folium map. The output is a folium Map object.

```python
visualizer.display_dataset(name='test', y_column='PM2_5', colors=[
                           'blue', 'green', 'yellow', 'red'])
```

```text
Result:
<folium.folium.Map at 0x1d9b775bce0>
```

--------------------------------

### Predict and visualize validation set

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_zh_CN.ipynb

Initializes a prediction dataset from the validation dataframe, runs GNNWR prediction, and creates a dot map of the predicted results with specified value range.

```python
# 验证集预测
valid_pred_dst = datasets.init_predict_dataset(data=val_dataset.dataframe,
                                               train_dataset=train_dataset,
                                               x_column=[
                                                   'dem', 'w10', 'd10', 't2m', 'aod_sat', 'tp'],
                                               spatial_column=['lng', 'lat'])
valid_pred_res = gnnwr.predict(valid_pred_dst)
visualizer.dot_map(data=valid_pred_res, lon_column='lng', lat_column='lat',
                   y_column='pred_result', colors=['blue', 'green', 'yellow', 'red'], vmin=4, vmax=81)
```

--------------------------------

### Heatmap for coefficient 'coef_d10'

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_zh_CN.ipynb

Generates a heatmap for the 10-meter wind direction coefficient.

```python
visualizer.coefs_heatmap('coef_d10', colors=['blue', 'green', 'yellow', 'red'])
```

--------------------------------

### Heatmap for coefficient 'coef_t2m'

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_zh_CN.ipynb

Generates a heatmap for the 2-meter temperature coefficient.

```python
visualizer.coefs_heatmap('coef_t2m', colors=['blue', 'green', 'yellow', 'red'])
```

--------------------------------

### Display dataset distribution for 'train'

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_en_US.ipynb

Calls visualizer.display_dataset with name='train' and y_column='PM2_5' to generate a folium map. The output is a folium Map object.

```python
visualizer.display_dataset(name='train', y_column='PM2_5', colors=[
                           'blue', 'green', 'yellow', 'red'])
```

```text
Result:
<folium.folium.Map at 0x1db1acc1950>
```

--------------------------------

### Train GNNWR model with early stopping

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_zh_CN.ipynb

Runs the training loop for up to 200 epochs with early stopping after 50 epochs without improvement. The output shows progress bars and training metrics, and stops when the model has not improved for over 50 epochs.

```python
gnnwr.run(max_epoch=200, early_stop=50)
```

```text
Output:
  0%|          | 0/200 [00:00<?, ?it/s]
```

```text
Output:
 84%|████████▍ | 169/200 [00:25<00:04,  6.50it/s, Train Loss=52.647901, Train R2=0.738589, Train AIC=6916.417, Valid Loss=54.8, Valid R2=0.71, Best Valid R2=0.763, Learning Rate=0.2]       
```

```text
Output:
Training stop! Model has not been improved for over 50 epochs.
```

--------------------------------

### Create 3D Scatter Plots for Coefficients

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gtnnwr_en_US.ipynb

Creates a 3x3 grid of 3D scatter plots showing the relationship between projected coordinates and day, colored by different coefficient values (coef_refl_b01 to coef_refl_b07). Sets font to Arial Unicode MS and uses a color bar.

```python
fig = plt.figure(figsize=(30, 30))
plt.rcParams['font.sans-serif'] = ['Arial Unicode MS']
id = 1
for i in [1, 2, 3, 4, 5, 7]:
    ax = fig.add_subplot(3, 3, id, projection='3d')
    fig.colorbar(ax.scatter(ana_data['proj_x'], ana_data['proj_y'], ana_data['day'],
                            c=ana_data['coef_refl_b0'+str(i)], cmap='GnBu'),
                 shrink=0.8, aspect=30, orientation='horizontal', pad=0.05)
    ax.set_xlabel('proj_x')
    ax.set_ylabel('proj_y')
    ax.set_zlabel('day')
    ax.set_title('coef_refl_b0'+str(i), fontsize=15, fontweight='bold')
    id += 1
```

```text
Result:
<Figure size 3000x3000 with 12 Axes>
```

--------------------------------

### Save train, validation, and test datasets

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_zh_CN.ipynb

Saves the train, validation, and test datasets to the specified paths using the save method with exist_ok=True.

```python
# 保存数据集到指定路径
train_dataset.save('../demo_result/gnnwr/dataset/train_dataset/', exist_ok=True)
val_dataset.save('../demo_result/gnnwr/dataset/val_dataset/', exist_ok=True)
test_dataset.save('../demo_result/gnnwr/dataset/test_dataset/', exist_ok=True)
```

--------------------------------

### Display dataset 'valid' with PM2.5 values

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_zh_CN.ipynb

Displays the validation dataset on a map, coloring points by PM2.5 values.

```python
visualizer.display_dataset(name='valid', y_column='PM2_5', colors=[
                           'blue', 'green', 'yellow', 'red'])
```

--------------------------------

### Draw coefficient heatmap for 'coef_w10'

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_en_US.ipynb

Calls visualizer.coefs_heatmap with 'coef_w10' and specified colors to generate a folium map. The output is a folium Map object.

```python
visualizer.coefs_heatmap('coef_w10', colors=['blue', 'green', 'yellow', 'red'])
```

```text
Result:
<folium.folium.Map at 0x1d9b77cbc50>
```

--------------------------------

### gnnwr.models.GNNWR.run()

Source: https://github.com/zjuwss/gnnwr/blob/main/doc/build/html/genindex.html

Runs the GNNWR model training or prediction process. This method is part of the GNNWR class in gnnwr.models and is the main entry point for executing the model.

```APIDOC
## GNNWR.run()

### Description
Runs the GNNWR model training or prediction process.

### Class
gnnwr.models.GNNWR

### Usage
GNNWR.run()
```

--------------------------------

### gnnwr.utils.Visualize

Source: https://github.com/zjuwss/gnnwr/blob/main/doc/build/html/genindex.html

Class for visualization utilities. Refer to the utils documentation for usage details.

```APIDOC
## gnnwr.utils.Visualize

### Description
Class for visualization utilities.

### Class
Visualize

### Methods
- weights_heatmap()

### Parameters
None documented.

### Returns
None documented.
```

--------------------------------

### Create dot map for prediction results

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_zh_CN.ipynb

Creates a dot map of prediction results using longitude and latitude columns, coloring by predicted values.

```python
visualizer.dot_map(data=pred_res, lon_column='lng', lat_column='lat',
                   y_column='pred_result', colors=['blue', 'green', 'yellow', 'red'])
```

--------------------------------

### Predict with GNNWR model and display first rows

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_zh_CN.ipynb

Uses the loaded GNNWR model to predict on the prediction dataset and displays the first 5 rows of the result.

```python
pred_res = gnnwr_load.predict(pred_dataset)
pred_res.head(5)
```

--------------------------------

### Heatmap for coefficient 'coef_tp'

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_zh_CN.ipynb

Generates a heatmap for the total precipitation coefficient.

```python
visualizer.coefs_heatmap('coef_tp', colors=['blue', 'green', 'yellow', 'red'])
```

--------------------------------

### Train GTNNWR Model with run

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gtnnwr_en_US.ipynb

Trains the model for 200 epochs with a validation frequency of 50, displaying training loss, R2, AIC, and validation metrics.

```python
gtnnwr.run(200, 50)
```

```text
Output:
100%|██████████| 200/200 [01:06<00:00,  3.01it/s, Train Loss=0.652585, Train R2=-0.133355, Train AIC=5353.5864, Valid Loss=0.464, Valid R2=0.13, Best Valid R2=0.141, Learning Rate=0.1]    
```

--------------------------------

### gnnwr.models.GNNWR.load_model()

Source: https://github.com/zjuwss/gnnwr/blob/main/doc/build/html/genindex.html

Loads a trained GNNWR model from disk. This method is part of the GNNWR class in gnnwr.models and is used to restore a previously saved model.

```APIDOC
## GNNWR.load_model()

### Description
Loads a trained GNNWR model from disk.

### Class
gnnwr.models.GNNWR

### Usage
GNNWR.load_model()
```

--------------------------------

### Load saved datasets using datasets.load_dataset

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_zh_CN.ipynb

Loads the previously saved train, validation, and test datasets using the datasets.load_dataset function.

```python
train_dataset_load = datasets.load_dataset(
    '../demo_result/gnnwr/dataset/train_dataset/')


val_dataset_load = datasets.load_dataset(
    '../demo_result/gnnwr/dataset/val_dataset/')


test_dataset_load = datasets.load_dataset(
    '../demo_result/gnnwr/dataset/test_dataset/')
```

--------------------------------

### Retrieve Model Evaluation Results

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_en_US.ipynb

Displays the model information and performance metrics including R2, RMSE, and AIC values.

```python
print(gnnwr.result())
```

```text
Output:
--------------------Model Information-----------------
Model Name:           | GNNWR_PM25
independent variable: | ['dem', 'w10', 'd10', 't2m', 'aod_sat', 'tp']
dependent variable:   | ['PM2_5']

OLS coefficients: 
x0: 4.84385
x1: -8.85402
x2: 0.17612
x3: 18.69575
x4: 35.19490
x5: -23.06419
Intercept: 22.06353

--------------------Result Information----------------
Test Loss: |                  56.67216
Test R2  : |                   0.73790
Train R2 : |                   0.82596
Valid R2 : |                   0.76837
RMSE: |                        7.52809
AIC:  |                     1452.48474
AICc: |                     1449.55029
F1:   |                        0.14694
F2:   |                        0.79863
f3_param_0: |                  1.39256
f3_param_1: |                  4.65791
f3_param_2: |                  0.36700
f3_param_3: |                 38.43280
f3_param_4: |                 84.29873
f3_param_5: |                100.56641
f3_param_6: |                 71.90481
```

--------------------------------

### Draw coefficient heatmap for 'coef_d10'

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_en_US.ipynb

Calls visualizer.coefs_heatmap with 'coef_d10' and specified colors to generate a folium map. The output is a folium Map object.

```python
visualizer.coefs_heatmap('coef_d10', colors=['blue', 'green', 'yellow', 'red'])
```

```text
Result:
<folium.folium.Map at 0x1d9a9f16950>
```

--------------------------------

### Draw coefficient heatmap for 'coef_aod_sat'

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_en_US.ipynb

Calls visualizer.coefs_heatmap with 'coef_aod_sat' and specified colors to generate a folium map. The output is a folium Map object.

```python
visualizer.coefs_heatmap('coef_aod_sat', colors=[
                         'blue', 'green', 'yellow', 'red'])
```

```text
Result:
<folium.folium.Map at 0x1d9b7810e20>
```

--------------------------------

### Display dataset distribution for 'valid'

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_en_US.ipynb

Calls visualizer.display_dataset with name='valid' and y_column='PM2_5' to generate a folium map. The output is a folium Map object.

```python
visualizer.display_dataset(name='valid', y_column='PM2_5', colors=[
                           'blue', 'green', 'yellow', 'red'])
```

```text
Result:
<folium.folium.Map at 0x1d9b775b820>
```

--------------------------------

### Display validation dataset with comment

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_zh_CN.ipynb

Displays the validation dataset on a map, with a comment indicating it is the validation set.

```python
# 验证集
visualizer.display_dataset(name='valid', y_column='PM2_5', colors=[
                           'blue', 'green', 'yellow', 'red'])
```

--------------------------------

### Draw coefficient heatmap for 'coef_dem'

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_en_US.ipynb

Calls visualizer.coefs_heatmap with 'coef_dem' and specified colors to generate a folium map. The output is a folium Map object.

```python
visualizer.coefs_heatmap('coef_dem', colors=['blue', 'green', 'yellow', 'red'])
```

```text
Result:
<folium.folium.Map at 0x1d9b78118c0>
```

--------------------------------

### Draw coefficient heatmap for 'coef_tp'

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_en_US.ipynb

Calls visualizer.coefs_heatmap with 'coef_tp' and specified colors to generate a folium map. The output is a folium Map object.

```python
visualizer.coefs_heatmap('coef_tp', colors=['blue', 'green', 'yellow', 'red'])
```

```text
Result:
<folium.folium.Map at 0x1d9b77ca750>
```

--------------------------------

### gnnwr.datasets.load_dataset()

Source: https://github.com/zjuwss/gnnwr/blob/main/doc/build/html/genindex.html

Loads a dataset for GNNWR modeling. This function is part of the gnnwr.datasets module and is used to prepare data for training or prediction.

```APIDOC
## load_dataset()

### Description
Loads a dataset for GNNWR modeling.

### Module
gnnwr.datasets

### Usage
load_dataset()
```

--------------------------------

### Draw coefficient heatmap for 'coef_t2m'

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_en_US.ipynb

Calls visualizer.coefs_heatmap with 'coef_t2m' and specified colors to generate a folium map. The output is a folium Map object.

```python
visualizer.coefs_heatmap('coef_t2m', colors=['blue', 'green', 'yellow', 'red'])
```

```text
Result:
<folium.folium.Map at 0x1d9b77cb750>
```

--------------------------------

### Heatmap for coefficient 'coef_dem'

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_zh_CN.ipynb

Generates a heatmap for the DEM coefficient.

```python
visualizer.coefs_heatmap('coef_dem', colors=['blue', 'green', 'yellow', 'red'])
```

--------------------------------

### gnnwr.utils.Visualize.weights_heatmap

Source: https://github.com/zjuwss/gnnwr/blob/main/doc/build/html/genindex.html

Method to generate a heatmap of weights. Refer to the utils documentation for usage details.

```APIDOC
## gnnwr.utils.Visualize.weights_heatmap

### Description
Method to generate a heatmap of weights.

### Method
weights_heatmap

### Signature
weights_heatmap()

### Parameters
None documented.

### Returns
None documented.
```

--------------------------------

### Heatmap for bias coefficient

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_zh_CN.ipynb

Generates a heatmap for the bias coefficient.

```python
visualizer.coefs_heatmap('bias', colors=['blue', 'green', 'yellow', 'red'])
```

--------------------------------

### Draw dot map for prediction results

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_en_US.ipynb

Calls visualizer.dot_map with prediction data and specified columns to generate a folium map. The output is a folium Map object.

```python
visualizer.dot_map(data=pred_res, lon_column='lng', lat_column='lat',
                   y_column='pred_result', colors=['blue', 'green', 'yellow', 'red'])
```

```text
Result:
<folium.folium.Map at 0x1d9b78571d0>
```

--------------------------------

### Heatmap for coefficient 'coef_aod_sat'

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_zh_CN.ipynb

Generates a heatmap for the AOD satellite coefficient.

```python
visualizer.coefs_heatmap('coef_aod_sat', colors=[
                         'blue', 'green', 'yellow', 'red'])
```

--------------------------------

### Display dataset distribution for 'valid' (second occurrence)

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_en_US.ipynb

Calls visualizer.display_dataset with name='valid' and y_column='PM2_5' to generate a folium map. The output is a folium Map object.

```python
visualizer.display_dataset(name='valid', y_column='PM2_5', colors=[
                           'blue', 'green', 'yellow', 'red'])
```

```text
Result:
<folium.folium.Map at 0x1db1dbac250>
```

--------------------------------

### Print model result with gtnnwr.result()

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gtnnwr_en_US.ipynb

Call result() on the fitted GTNNWR model to display model information (name, variables, OLS coefficients) and test set performance metrics (loss, R2, RMSE, AIC, etc.). The output is printed to the console.

```python
print(gtnnwr.result())
```

```text
Output:
--------------------Model Information-----------------
Model Name:           | GTNNWR_DSi
independent variable: | ['refl_b01', 'refl_b02', 'refl_b03', 'refl_b04', 'refl_b05', 'refl_b07']
dependent variable:   | ['SiO3']

OLS coefficients: 
x0: 7.70582
x1: 0.80789
x2: 0.35238
x3: -7.27008
x4: 2.71207
x5: -2.70717
Intercept: 1.14341

--------------------Result Information----------------
Test Loss: |                   0.46818
Test R2  : |                   0.14439
Train R2 : |                   0.15684
Valid R2 : |                   0.14058
RMSE: |                        0.68424
AIC:  |                      910.35614
AICc: |                      916.41449
F1:   |                        0.73564
F2:   |                       23.04567
f3_param_0: |                  3.03797
f3_param_1: |                  6.53450
f3_param_2: |                  0.88140
f3_param_3: |                  0.05762
f3_param_4: |                  2.91409
f3_param_5: |                 25.02167
f3_param_6: |                 30.42470
```

--------------------------------

### Draw coefficient heatmap for 'bias'

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_en_US.ipynb

Calls visualizer.coefs_heatmap with 'bias' and specified colors to generate a folium map. The output is a folium Map object.

```python
visualizer.coefs_heatmap('bias', colors=['blue', 'green', 'yellow', 'red'])
```

```text
Result:
<folium.folium.Map at 0x1db1dbac150>
```

--------------------------------

### Export and View Regression Results

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_en_US.ipynb

Exports the regression results to a CSV file and displays the sorted dataframe.

```python
gnnwr.reg_result('../demo_result/gnnwr/gnnwr_result.csv').sort_values(by='id')
```

```text
Result:
       coef_dem   coef_w10  coef_d10   coef_t2m  coef_aod_sat    coef_tp  \
525  -13.982780   0.986766 -0.265440  15.096186     39.772655   0.168308   
856  -13.017211   1.010324 -0.244473  15.638182     43.804943  -0.079256   
1106 -13.462659   1.519648 -0.253571  14.013595     43.117420   0.434180   
878  -13.738739   1.328917 -0.260308  14.333707     41.565952   0.365591   
1224 -13.282292   1.723946 -0.250305  13.519067     44.411713   0.508260   
...         ...        ...       ...        ...           ...        ...   
1371 -17.246046   6.366604 -0.478529  -1.594304     52.899521  10.622722   
961    1.734755 -22.650177 -1.014575   9.799455     89.167091 -34.360985   
925  -17.335573   6.256605 -0.480732  -1.363827     52.114891  10.563044   
1088  -8.700379  10.593680 -0.705613  25.247837     32.304390   4.642990   
245  -24.186794  -3.212118 -0.623428  21.558348     -1.865212  -1.452510   

           bias  Pred_PM2_5    id dataset_belong  denormalized_pred_result  
525   10.864908   50.459122     0          train                 50.459122  
856    7.333379   41.463264     1          train                 41.463264  
1106   8.655427   50.600433     2          valid                 50.600433  
878    9.796055   50.511223     3          train                 50.511223  
1224   7.894271   50.550869     4           test                 50.550869  
...         ...         ...   ...            ...                       ...  
1371  14.197436   53.354580  1403           test                 53.354580  
961   14.354010   12.014944  1404          train                 12.014944  
925   14.693017   53.442284  1405          train                 53.442284  
1088   6.718304   12.019615  1406          valid                 12.019615  
245   48.198952   57.349991  1407          train                 57.349991  

[1408 rows x 11 columns]
```

--------------------------------

### Save regression results to CSV

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gnnwr_zh_CN.ipynb

Saves the regression results to a CSV file using reg_result() and sorts the rows by the 'id' column.

```python
gnnwr.reg_result('../demo_result/gnnwr/gnnwr_result.csv').sort_values(by='id')
```

--------------------------------

### gnnwr.datasets.baseDataset.read()

Source: https://github.com/zjuwss/gnnwr/blob/main/doc/build/html/genindex.html

Reads data from a file or source. This method is part of the baseDataset class in gnnwr.datasets and is used to load raw data into the dataset object.

```APIDOC
## baseDataset.read()

### Description
Reads data from a file or source.

### Class
gnnwr.datasets.baseDataset

### Usage
baseDataset.read()
```

--------------------------------

### gnnwr.datasets.baseDataset.rescale()

Source: https://github.com/zjuwss/gnnwr/blob/main/doc/build/html/genindex.html

Rescales the dataset. This method is part of the baseDataset class in gnnwr.datasets and is used to normalize or standardize the data.

```APIDOC
## baseDataset.rescale()

### Description
Rescales the dataset.

### Class
gnnwr.datasets.baseDataset

### Usage
baseDataset.rescale()
```

--------------------------------

### Evaluate predictions with R2 and scatter plot

Source: https://github.com/zjuwss/gnnwr/blob/main/demo/demo_gtnnwr_zh_CN.ipynb

Sorts result and original data by 'id', computes R2 score between actual 'SiO3' and predicted 'Pred_SiO3', and creates a scatter plot with a diagonal reference line. Output shows R2: 0.3518267064128938 and a figure.

```python
import matplotlib.pyplot as plt
from sklearn.metrics import r2_score as r2
result_data.sort_values(by='id', inplace=True)
data.sort_values(by='id', inplace=True)
print('R2:', r2(data['SiO3'], result_data['Pred_SiO3']))
plt.scatter(data['SiO3'], result_data['Pred_SiO3'])
plt.plot([i for i in range(9)], [i for i in range(9)], color='black')
plt.xlabel('label')
plt.ylabel('pred')
plt.show()
```

--------------------------------

### gnnwr.networks.weight_share

Source: https://github.com/zjuwss/gnnwr/blob/main/doc/build/html/genindex.html

Function to share weights across network layers. Refer to the networks documentation for usage details.

```APIDOC
## gnnwr.networks.weight_share

### Description
Function to share weights across network layers.

### Method
weight_share

### Signature
weight_share()

### Parameters
None documented.

### Returns
None documented.
```
Tokens:
Raw