[1]:

%matplotlib inline

D:\C\Anaconda3\envs\tfcpu27_py39\lib\site-packages\numpy\_distributor_init.py:30: UserWarning: loaded more than 1 DLL from .libs:
D:\C\Anaconda3\envs\tfcpu27_py39\lib\site-packages\numpy\.libs\libopenblas.EL2C6PLE4ZYW3ECEVIV3OXXGRN2NRFM2.gfortran-win_amd64.dll
D:\C\Anaconda3\envs\tfcpu27_py39\lib\site-packages\numpy\.libs\libopenblas.GK7GX5KEQ4F6UYO3P26ULGBQYHGQO7J4.gfortran-win_amd64.dll
  warnings.warn("loaded more than 1 DLL from .libs:"

[ ]:

try:
    import ai4water
except (ImportError, ModuleNotFoundError):
    !pip install ai4water[tf2]

LSTM for time series forecasting

[2]:

from ai4water import Model
from ai4water.datasets import MtropicsLaos
from ai4water.utils.utils import get_version_info

D:\C\Anaconda3\envs\tfcpu27_py39\lib\site-packages\sklearn\experimental\enable_hist_gradient_boosting.py:16: UserWarning: Since version 1.0, it is not needed to import enable_hist_gradient_boosting anymore. HistGradientBoostingClassifier and HistGradientBoostingRegressor are now stable and can be normally imported from sklearn.ensemble.
  warnings.warn(

[3]:

for k,v in get_version_info().items():
    print(f"{k} version: {v}")

python version: 3.9.7 | packaged by conda-forge | (default, Sep 29 2021, 19:20:16) [MSC v.1916 64 bit (AMD64)]
os version: nt
ai4water version: 1.06
lightgbm version: 3.3.1
tcn version: 3.4.0
catboost version: 0.26
xgboost version: 1.5.0
easy_mpl version: 0.21.2
SeqMetrics version: 1.3.3
tensorflow version: 2.7.0
keras.api._v2.keras version: 2.7.0
numpy version: 1.21.0
pandas version: 1.3.4
matplotlib version: 3.4.3
h5py version: 3.5.0
sklearn version: 1.0.1
shapefile version: 2.3.0
xarray version: 0.20.1
netCDF4 version: 1.5.7
optuna version: 2.10.1
skopt version: 0.9.0
hyperopt version: 0.2.7
plotly version: 5.3.1
lime version: NotDefined
seaborn version: 0.11.2

[4]:

dataset = MtropicsLaos(save_as_nc=True,  # if set to True, then netcdf must be installed
                       convert_to_csv=False,
                       path="F:\\data\\MtropicsLaos",
                      )

lookback = 20

data = dataset.make_regression(lookback_steps=lookback)
data.shape


    Not downloading the data since the directory
    F:\data\MtropicsLaos already exists.
    Use overwrite=True to remove previously saved files and download again

Value based partial slicing on non-monotonic DatetimeIndexes with non-existing keys is deprecated and will raise a KeyError in a future Version.

[4]:

(4071, 9)

[5]:

inputs = data.columns.tolist()[0:-1]
inputs

[5]:

['air_temp',
 'rel_hum',
 'wind_speed',
 'sol_rad',
 'water_level',
 'pcp',
 'susp_pm',
 'Ecoli_source']

[6]:

layers ={
    "Input": {'config': {'shape': (lookback, len(inputs))}},
    "LSTM": {'config': {'units': 14, 'activation': "elu"}},
    "Dense": 1
}
model = Model(
    input_features=inputs,
    output_features=data.columns.tolist()[-1:],
    model={'layers':layers},
    ts_args={"lookback": lookback},
    lr=0.009919,
    batch_size=8,
    split_random=True,
    train_fraction=1.0,
    x_transformation="zscore",
    y_transformation={"method": "log", "replace_zeros": True, "treat_negatives": True},
    epochs=200
)


            building DL model for
            regression problem using Model
Model: "model"
_________________________________________________________________
 Layer (type)                Output Shape              Param #
=================================================================
 Input (InputLayer)          [(None, 20, 8)]           0

 LSTM (LSTM)                 (None, 14)                1288

 Dense (Dense)               (None, 1)                 15

=================================================================
Total params: 1,303
Trainable params: 1,303
Non-trainable params: 0
_________________________________________________________________

[7]:

h = model.fit(data=data)


********** Removing Examples with nan in labels  **********

***** Training *****
input_x shape:  (206, 20, 8)
target shape:  (206, 1)

********** Removing Examples with nan in labels  **********

***** Validation *****
input_x shape:  (52, 20, 8)
target shape:  (52, 1)
Epoch 1/200
assigning name Input to IteratorGetNext:0 with shape (None, 20, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 20, 8)
18/26 [===================>..........] - ETA: 0s - loss: 100.7893assigning name Input to IteratorGetNext:0 with shape (None, 20, 8)
26/26 [==============================] - 2s 11ms/step - loss: 78.0383 - val_loss: 26.9827
Epoch 2/200
26/26 [==============================] - 0s 5ms/step - loss: 25.4774 - val_loss: 18.8905
Epoch 3/200
26/26 [==============================] - 0s 5ms/step - loss: 14.9665 - val_loss: 8.7782
Epoch 4/200
26/26 [==============================] - 0s 5ms/step - loss: 7.0218 - val_loss: 3.4659
Epoch 5/200
26/26 [==============================] - 0s 5ms/step - loss: 3.7423 - val_loss: 1.9051
Epoch 6/200
26/26 [==============================] - 0s 4ms/step - loss: 3.4310 - val_loss: 1.7568
Epoch 7/200
26/26 [==============================] - 0s 4ms/step - loss: 2.7946 - val_loss: 1.5086
Epoch 8/200
26/26 [==============================] - 0s 4ms/step - loss: 2.5614 - val_loss: 1.8149
Epoch 9/200
26/26 [==============================] - 0s 4ms/step - loss: 2.5903 - val_loss: 1.5388
Epoch 10/200
26/26 [==============================] - 0s 4ms/step - loss: 2.5322 - val_loss: 1.4675
Epoch 11/200
26/26 [==============================] - 0s 4ms/step - loss: 2.3507 - val_loss: 1.4359
Epoch 12/200
26/26 [==============================] - 0s 4ms/step - loss: 2.3415 - val_loss: 1.5000
Epoch 13/200
26/26 [==============================] - 0s 4ms/step - loss: 2.2680 - val_loss: 1.3931
Epoch 14/200
26/26 [==============================] - 0s 4ms/step - loss: 2.2826 - val_loss: 1.5056
Epoch 15/200
26/26 [==============================] - 0s 4ms/step - loss: 2.2283 - val_loss: 1.3600
Epoch 16/200
26/26 [==============================] - 0s 4ms/step - loss: 2.1756 - val_loss: 1.3777
Epoch 17/200
26/26 [==============================] - 0s 4ms/step - loss: 2.2203 - val_loss: 1.4283
Epoch 18/200
26/26 [==============================] - 0s 4ms/step - loss: 2.1854 - val_loss: 1.5904
Epoch 19/200
26/26 [==============================] - 0s 4ms/step - loss: 2.3765 - val_loss: 1.3355
Epoch 20/200
26/26 [==============================] - 0s 4ms/step - loss: 2.1578 - val_loss: 1.5137
Epoch 21/200
26/26 [==============================] - 0s 4ms/step - loss: 2.1805 - val_loss: 1.4412
Epoch 22/200
26/26 [==============================] - 0s 4ms/step - loss: 2.1697 - val_loss: 1.4324
Epoch 23/200
26/26 [==============================] - 0s 4ms/step - loss: 2.1543 - val_loss: 1.4875
Epoch 24/200
26/26 [==============================] - 0s 4ms/step - loss: 2.0792 - val_loss: 1.5516
Epoch 25/200
26/26 [==============================] - 0s 4ms/step - loss: 2.2077 - val_loss: 1.6654
Epoch 26/200
26/26 [==============================] - 0s 4ms/step - loss: 2.2215 - val_loss: 1.5029
Epoch 27/200
26/26 [==============================] - 0s 4ms/step - loss: 2.0324 - val_loss: 1.8807
Epoch 28/200
26/26 [==============================] - 0s 4ms/step - loss: 2.0951 - val_loss: 1.7207
Epoch 29/200
26/26 [==============================] - 0s 4ms/step - loss: 2.0782 - val_loss: 1.7571
Epoch 30/200
26/26 [==============================] - 0s 4ms/step - loss: 2.0621 - val_loss: 1.7785
Epoch 31/200
26/26 [==============================] - 0s 4ms/step - loss: 2.0790 - val_loss: 1.7358
Epoch 32/200
26/26 [==============================] - 0s 4ms/step - loss: 2.1032 - val_loss: 1.6230
Epoch 33/200
26/26 [==============================] - 0s 4ms/step - loss: 2.0606 - val_loss: 1.5836
Epoch 34/200
26/26 [==============================] - 0s 4ms/step - loss: 2.0150 - val_loss: 1.7484
Epoch 35/200
26/26 [==============================] - 0s 4ms/step - loss: 2.0663 - val_loss: 1.7289
Epoch 36/200
26/26 [==============================] - 0s 5ms/step - loss: 2.0914 - val_loss: 1.6749
Epoch 37/200
26/26 [==============================] - 0s 6ms/step - loss: 2.0429 - val_loss: 1.6325
Epoch 38/200
26/26 [==============================] - 0s 4ms/step - loss: 1.9600 - val_loss: 1.8131
Epoch 39/200
26/26 [==============================] - 0s 4ms/step - loss: 1.9286 - val_loss: 1.7169
Epoch 40/200
26/26 [==============================] - 0s 4ms/step - loss: 1.9502 - val_loss: 1.8640
Epoch 41/200
26/26 [==============================] - 0s 4ms/step - loss: 1.9467 - val_loss: 1.9784
Epoch 42/200
26/26 [==============================] - 0s 4ms/step - loss: 1.9145 - val_loss: 1.8156
Epoch 43/200
26/26 [==============================] - 0s 4ms/step - loss: 1.8927 - val_loss: 1.8794
Epoch 44/200
26/26 [==============================] - 0s 4ms/step - loss: 1.9059 - val_loss: 1.7426
Epoch 45/200
26/26 [==============================] - 0s 4ms/step - loss: 1.8567 - val_loss: 1.9001
Epoch 46/200
26/26 [==============================] - 0s 4ms/step - loss: 1.8493 - val_loss: 1.9401
Epoch 47/200
26/26 [==============================] - 0s 4ms/step - loss: 2.0169 - val_loss: 1.9715
Epoch 48/200
26/26 [==============================] - 0s 4ms/step - loss: 1.9213 - val_loss: 2.1020
Epoch 49/200
26/26 [==============================] - 0s 4ms/step - loss: 1.8113 - val_loss: 1.9011
Epoch 50/200
26/26 [==============================] - 0s 4ms/step - loss: 1.7978 - val_loss: 1.8903
Epoch 51/200
26/26 [==============================] - 0s 4ms/step - loss: 1.8025 - val_loss: 1.9471
Epoch 52/200
26/26 [==============================] - 0s 4ms/step - loss: 1.7739 - val_loss: 1.7793
Epoch 53/200
26/26 [==============================] - 0s 4ms/step - loss: 1.8192 - val_loss: 2.1097
Epoch 54/200
26/26 [==============================] - 0s 4ms/step - loss: 1.8179 - val_loss: 1.9036
Epoch 55/200
26/26 [==============================] - 0s 4ms/step - loss: 1.7332 - val_loss: 1.8925
Epoch 56/200
26/26 [==============================] - 0s 4ms/step - loss: 1.7616 - val_loss: 1.8577
Epoch 57/200
26/26 [==============================] - 0s 4ms/step - loss: 1.8437 - val_loss: 2.0430
Epoch 58/200
26/26 [==============================] - 0s 4ms/step - loss: 1.7640 - val_loss: 2.3366
Epoch 59/200
26/26 [==============================] - 0s 4ms/step - loss: 1.7109 - val_loss: 2.0880
Epoch 60/200
26/26 [==============================] - 0s 5ms/step - loss: 1.7181 - val_loss: 1.9773
Epoch 61/200
26/26 [==============================] - 0s 4ms/step - loss: 1.6474 - val_loss: 2.1006
Epoch 62/200
26/26 [==============================] - 0s 4ms/step - loss: 1.6376 - val_loss: 1.9836
Epoch 63/200
26/26 [==============================] - 0s 4ms/step - loss: 1.6859 - val_loss: 1.9822
Epoch 64/200
26/26 [==============================] - 0s 4ms/step - loss: 1.6723 - val_loss: 1.8385
Epoch 65/200
26/26 [==============================] - 0s 4ms/step - loss: 1.5259 - val_loss: 2.1795
Epoch 66/200
26/26 [==============================] - 0s 4ms/step - loss: 1.4792 - val_loss: 2.1116
Epoch 67/200
26/26 [==============================] - 0s 4ms/step - loss: 1.4955 - val_loss: 1.9483
Epoch 68/200
26/26 [==============================] - 0s 4ms/step - loss: 1.6099 - val_loss: 2.2973
Epoch 69/200
26/26 [==============================] - 0s 4ms/step - loss: 1.5268 - val_loss: 1.8486
Epoch 70/200
26/26 [==============================] - 0s 4ms/step - loss: 1.6557 - val_loss: 2.0699
Epoch 71/200
26/26 [==============================] - 0s 4ms/step - loss: 1.6293 - val_loss: 2.0666
Epoch 72/200
26/26 [==============================] - 0s 4ms/step - loss: 1.4971 - val_loss: 1.8501
Epoch 73/200
26/26 [==============================] - 0s 4ms/step - loss: 1.5565 - val_loss: 2.3651
Epoch 74/200
26/26 [==============================] - 0s 4ms/step - loss: 1.4828 - val_loss: 1.7770
Epoch 75/200
26/26 [==============================] - 0s 4ms/step - loss: 1.5315 - val_loss: 2.0530
Epoch 76/200
26/26 [==============================] - 0s 4ms/step - loss: 1.6978 - val_loss: 1.9833
Epoch 77/200
26/26 [==============================] - 0s 4ms/step - loss: 1.4611 - val_loss: 2.4645
Epoch 78/200
26/26 [==============================] - 0s 4ms/step - loss: 1.4407 - val_loss: 2.0058
Epoch 79/200
26/26 [==============================] - 0s 4ms/step - loss: 1.4570 - val_loss: 2.2919
Epoch 80/200
26/26 [==============================] - 0s 4ms/step - loss: 1.4043 - val_loss: 2.1670
Epoch 81/200
26/26 [==============================] - 0s 4ms/step - loss: 1.3432 - val_loss: 2.2628
Epoch 82/200
26/26 [==============================] - 0s 4ms/step - loss: 1.3707 - val_loss: 1.9328
Epoch 83/200
26/26 [==============================] - 0s 4ms/step - loss: 1.3362 - val_loss: 2.2226
Epoch 84/200
26/26 [==============================] - 0s 4ms/step - loss: 1.4187 - val_loss: 2.6026
Epoch 85/200
26/26 [==============================] - 0s 4ms/step - loss: 1.5127 - val_loss: 1.9446
Epoch 86/200
26/26 [==============================] - 0s 4ms/step - loss: 1.4494 - val_loss: 2.0469
Epoch 87/200
26/26 [==============================] - 0s 4ms/step - loss: 1.4525 - val_loss: 2.1871
Epoch 88/200
26/26 [==============================] - 0s 5ms/step - loss: 1.3628 - val_loss: 2.0229
Epoch 89/200
26/26 [==============================] - 0s 4ms/step - loss: 1.3952 - val_loss: 2.4309
Epoch 90/200
26/26 [==============================] - 0s 4ms/step - loss: 1.3025 - val_loss: 1.9898
Epoch 91/200
26/26 [==============================] - 0s 4ms/step - loss: 1.2165 - val_loss: 2.2065
Epoch 92/200
26/26 [==============================] - 0s 4ms/step - loss: 1.4830 - val_loss: 1.8223
Epoch 93/200
26/26 [==============================] - 0s 4ms/step - loss: 1.5106 - val_loss: 2.2720
Epoch 94/200
26/26 [==============================] - 0s 4ms/step - loss: 1.4248 - val_loss: 2.0414
Epoch 95/200
26/26 [==============================] - 0s 4ms/step - loss: 1.4237 - val_loss: 1.9304
Epoch 96/200
26/26 [==============================] - 0s 4ms/step - loss: 1.2869 - val_loss: 2.0285
Epoch 97/200
26/26 [==============================] - 0s 4ms/step - loss: 1.2413 - val_loss: 2.0274
Epoch 98/200
26/26 [==============================] - 0s 4ms/step - loss: 1.2791 - val_loss: 2.2272
Epoch 99/200
26/26 [==============================] - 0s 4ms/step - loss: 1.4163 - val_loss: 2.1095
Epoch 100/200
26/26 [==============================] - 0s 4ms/step - loss: 1.3964 - val_loss: 2.2720
Epoch 101/200
26/26 [==============================] - 0s 4ms/step - loss: 1.6256 - val_loss: 1.8829
Epoch 102/200
26/26 [==============================] - 0s 5ms/step - loss: 1.4604 - val_loss: 2.1002
Epoch 103/200
26/26 [==============================] - 0s 4ms/step - loss: 1.4628 - val_loss: 2.6744
Epoch 104/200
26/26 [==============================] - 0s 4ms/step - loss: 1.3415 - val_loss: 1.9412
Epoch 105/200
26/26 [==============================] - 0s 4ms/step - loss: 1.2682 - val_loss: 2.2500
Epoch 106/200
26/26 [==============================] - 0s 4ms/step - loss: 1.2563 - val_loss: 2.5064
Epoch 107/200
26/26 [==============================] - 0s 4ms/step - loss: 1.2008 - val_loss: 2.0823
Epoch 108/200
26/26 [==============================] - 0s 4ms/step - loss: 1.2272 - val_loss: 2.3500
Epoch 109/200
26/26 [==============================] - 0s 4ms/step - loss: 1.3757 - val_loss: 1.9903
Epoch 110/200
26/26 [==============================] - 0s 4ms/step - loss: 1.2150 - val_loss: 2.6523
Epoch 111/200
26/26 [==============================] - 0s 4ms/step - loss: 1.2578 - val_loss: 2.1164
Epoch 112/200
26/26 [==============================] - 0s 4ms/step - loss: 1.3046 - val_loss: 2.2229
Epoch 113/200
26/26 [==============================] - 0s 4ms/step - loss: 1.2665 - val_loss: 2.1473
Epoch 114/200
26/26 [==============================] - 0s 4ms/step - loss: 1.1936 - val_loss: 2.1517
Epoch 115/200
26/26 [==============================] - 0s 4ms/step - loss: 1.1240 - val_loss: 2.3406
Epoch 116/200
26/26 [==============================] - 0s 5ms/step - loss: 1.2132 - val_loss: 1.9678
Epoch 117/200
26/26 [==============================] - 0s 4ms/step - loss: 1.1991 - val_loss: 2.4041
Epoch 118/200
26/26 [==============================] - 0s 4ms/step - loss: 1.0964 - val_loss: 2.8420
Epoch 119/200
26/26 [==============================] - 0s 4ms/step - loss: 1.0876 - val_loss: 1.9345

../../_images/_notebooks_model_lstm_rgr_9_1.png

********** Successfully loaded weights from weights_019_1.33551.hdf5 file **********

[8]:

type(h)

[8]:

keras.callbacks.History

[9]:

p = model.predict(data=data)

***** Test *****
input_x shape:  (0,)
target shape:  (0,)

********** Removing Examples with nan in labels  **********

***** Validation *****
input_x shape:  (52, 20, 8)
target shape:  (52, 1)
assigning name Input to IteratorGetNext:0 with shape (None, 20, 8)

No test data found. using validation data instead

2/2 [==============================] - 0s 2ms/step

../../_images/_notebooks_model_lstm_rgr_11_3.png

../../_images/_notebooks_model_lstm_rgr_11_4.png

../../_images/_notebooks_model_lstm_rgr_11_5.png

../../_images/_notebooks_model_lstm_rgr_11_6.png

../../_images/_notebooks_model_lstm_rgr_11_7.png

../../_images/_notebooks_model_lstm_rgr_11_8.png

[10]:

p = model.predict_on_training_data(data=data)


********** Removing Examples with nan in labels  **********

***** Training *****
input_x shape:  (206, 20, 8)
target shape:  (206, 1)
7/7 [==============================] - 0s 1ms/step

../../_images/_notebooks_model_lstm_rgr_12_1.png

../../_images/_notebooks_model_lstm_rgr_12_2.png

../../_images/_notebooks_model_lstm_rgr_12_3.png

../../_images/_notebooks_model_lstm_rgr_12_4.png

../../_images/_notebooks_model_lstm_rgr_12_5.png

../../_images/_notebooks_model_lstm_rgr_12_6.png

[13]:

train_x, train_y = model.training_data(data=data)


********** Removing Examples with nan in labels  **********

***** Training *****
input_x shape:  (206, 20, 8)
target shape:  (206, 1)

[15]:

p = model.predict(x=train_x, y=train_y, plots=['residual'],  verbose=0)

../../_images/_notebooks_model_lstm_rgr_14_0.png

[12]:

t, p = model.predict_on_all_data(data=data, return_true=True, process_results=False)
t.shape, p.shape


********** Removing Examples with nan in labels  **********

***** Training *****
input_x shape:  (206, 20, 8)
target shape:  (206, 1)

********** Removing Examples with nan in labels  **********

***** Validation *****
input_x shape:  (52, 20, 8)
target shape:  (52, 1)
***** Test *****
input_x shape:  (0,)
target shape:  (0,)
9/9 [==============================] - 0s 1ms/step

[12]:

((258, 1), (258, 1))

[ ]: