[ ]:
%matplotlib inline
Comparison of LSTM with different transformations
[1]:
from ai4water.models import LSTM
from ai4water.utils.utils import get_version_info
from ai4water.experiments import TransformationExperiments
from ai4water.hyperopt import Categorical, Integer
from ai4water.utils.utils import dateandtime_now
from ai4water.datasets import MtropicsLaos
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:"
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(
[2]:
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
[3]:
dataset = MtropicsLaos(save_as_nc=True, # if set to True, then netcdf must be installed
convert_to_csv=False,
path="F:\\data\\MtropicsLaos", # path where the data is saved, if not availble leave it to None
)
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.
(3691, 9)
[7]:
data.head()
[7]:
| air_temp | rel_hum | wind_speed | sol_rad | water_level | pcp | susp_pm | Ecoli_source | Ecoli_mpn100 | |
|---|---|---|---|---|---|---|---|---|---|
| 2011-05-25 14:54:00 | 22.538134 | 96.169885 | 4.3 | 167.8 | 133.000000 | 0.0 | 0.9 | 2.868160e+15 | NaN |
| 2011-05-25 15:00:00 | 22.538271 | 96.170115 | 4.0 | 158.0 | 133.000000 | 0.0 | 0.9 | 2.868160e+15 | NaN |
| 2011-05-25 15:06:00 | 22.538407 | 96.170345 | 3.7 | 150.0 | 133.000000 | 0.0 | 0.9 | 2.868160e+15 | NaN |
| 2011-05-25 15:12:00 | 22.538544 | 96.170575 | 3.4 | 142.0 | 132.988889 | 0.0 | 0.9 | 2.868160e+15 | NaN |
| 2011-05-25 15:18:00 | 22.538680 | 96.170805 | 3.1 | 134.0 | 132.977778 | 0.0 | 0.9 | 2.868160e+15 | NaN |
[8]:
print(data.isna().sum())
air_temp 0
rel_hum 0
wind_speed 0
sol_rad 0
water_level 0
pcp 0
susp_pm 0
Ecoli_source 0
Ecoli_mpn100 3433
dtype: int64
[15]:
input_features = data.columns.tolist()[0:-1]
input_features
[15]:
['air_temp',
'rel_hum',
'wind_speed',
'sol_rad',
'water_level',
'pcp',
'susp_pm',
'Ecoli_source']
[16]:
output_features = data.columns.tolist()[-1:]
output_features
[16]:
['Ecoli_mpn100']
[17]:
class MyTransformationExperiments(TransformationExperiments):
def update_paras(self, **kwargs):
_layers = LSTM(units=kwargs['units'],
input_shape=(lookback, len(input_features)),
activation=kwargs['activation'])
y_transformation = kwargs['y_transformation']
if y_transformation == "none":
y_transformation = None
return {
'model': _layers,
'batch_size': int(kwargs['batch_size']),
'lr': float(kwargs['lr']),
'y_transformation': y_transformation
}
[18]:
cases = {
'model_None': {'y_transformation': 'none'},
'model_minmax': {'y_transformation': 'minmax'},
'model_zscore': {'y_transformation': 'zscore'},
'model_robust': {'y_transformation': 'robust'},
'model_quantile': {'y_transformation': 'quantile'},
'model_log': {'y_transformation': {'method':'log', 'treat_negatives': True, 'replace_zeros': True}},
"model_pareto": {"y_transformation": "pareto"},
"model_vast": {"y_transformation": "vast"},
"model_mmad": {"y_transformation": "mmad"},
'model_box_cox': {'y_transformation': {'method': 'box-cox', 'treat_negatives': True, 'replace_zeros': True}},
'model_yeo-johnson': {'y_transformation': 'yeo-johnson'},
}
[19]:
search_space = [
Integer(low=10, high=30, name='units', num_samples=10),
Categorical(categories=['relu', 'elu', 'tanh', "linear"], name='activation'),
Categorical(categories=[4, 8, 12, 16, 24, 32], name='batch_size'),
Categorical(categories=[0.05, 0.02, 0.009, 0.007, 0.005,
0.003, 0.001, 0.0009, 0.0007, 0.0005, 0.0003,
0.0001, 0.00009, 0.00007, 0.00005], name='lr'),
]
[20]:
x0 = [16, "relu", 32, 0.0001]
experiment = MyTransformationExperiments(
x_transformation="minmax",
cases=cases,
input_features=input_features,
output_features = output_features,
param_space=search_space,
x0=x0,
verbosity=0,
epochs=500,
exp_name = f"ecoli_lstm_y_exp_{dateandtime_now()}",
ts_args={"lookback": lookback},
save=False
)
[21]:
experiment.fit(data = data, run_type='dry_run')
********** Removing Examples with nan in labels **********
***** Training *****
input_x shape: (144, 18, 8)
target shape: (144, 1)
********** Removing Examples with nan in labels **********
***** Validation *****
input_x shape: (36, 18, 8)
target shape: (36, 1)
********** Removing Examples with nan in labels **********
***** Test *****
input_x shape: (78, 18, 8)
target shape: (78, 1)
running None model
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
divide by zero encountered in log
divide by zero encountered in true_divide
running minmax model
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
divide by zero encountered in log
invalid value encountered in log
invalid value encountered in log
divide by zero encountered in true_divide
running zscore model
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
divide by zero encountered in log
invalid value encountered in log
invalid value encountered in log
divide by zero encountered in true_divide
running robust model
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
divide by zero encountered in log
invalid value encountered in log
invalid value encountered in log
divide by zero encountered in true_divide
running quantile model
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
divide by zero encountered in log
divide by zero encountered in true_divide
running log model
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
overflow encountered in square
overflow encountered in reduce
running pareto model
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
divide by zero encountered in log
invalid value encountered in log
invalid value encountered in log
divide by zero encountered in true_divide
running vast model
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
invalid value encountered in log
invalid value encountered in log
running mmad model
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
divide by zero encountered in log
divide by zero encountered in true_divide
running box_cox model
overflow encountered in power
overflow encountered in multiply
overflow encountered in reduce
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
running yeo-johnson model
overflow encountered in power
overflow encountered in multiply
overflow encountered in reduce
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
divide by zero encountered in log
divide by zero encountered in true_divide
[23]:
results = experiment.compare_errors('rmse', data=data)
********** Removing Examples with nan in labels **********
***** Training *****
input_x shape: (144, 18, 8)
target shape: (144, 1)
********** Removing Examples with nan in labels **********
***** Validation *****
input_x shape: (36, 18, 8)
target shape: (36, 1)
********** Removing Examples with nan in labels **********
***** Test *****
input_x shape: (78, 18, 8)
target shape: (78, 1)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
overflow encountered in multiply
overflow encountered in square
overflow encountered in square
overflow encountered in square
overflow encountered in square
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
[24]:
print(results)
train test
log 5409.450619 inf
box_cox 4814.439939 1.662770e+05
yeo-johnson 4757.606795 4.553616e+04
minmax 4226.970310 4.295539e+04
zscore 4187.136969 2.706149e+04
vast 4300.321445 2.651358e+04
pareto 4736.278899 2.515788e+04
None 5282.584216 2.427799e+04
mmad 4424.704330 1.870424e+04
robust 4014.077852 1.541642e+04
quantile 4798.453918 1.030005e+04
[25]:
r_squares = experiment.compare_errors('r2', data=data)
********** Removing Examples with nan in labels **********
***** Training *****
input_x shape: (144, 18, 8)
target shape: (144, 1)
********** Removing Examples with nan in labels **********
***** Validation *****
input_x shape: (36, 18, 8)
target shape: (36, 1)
********** Removing Examples with nan in labels **********
***** Test *****
input_x shape: (78, 18, 8)
target shape: (78, 1)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
overflow encountered in multiply
overflow encountered in square
overflow encountered in square
overflow encountered in square
overflow encountered in square
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
[26]:
print(r_squares)
train test
robust 0.663870 0.676143
mmad 0.594767 0.504128
None 0.423291 0.498426
pareto 0.531722 0.389125
quantile 0.565400 0.281959
vast 0.614166 0.035123
zscore 0.634722 0.007834
minmax 0.626898 0.001608
box_cox 0.547313 0.000283
yeo-johnson 0.567264 0.000166
log 0.438190 0.000000
[27]:
nrmses = experiment.compare_errors('nrmse', data=data)
********** Removing Examples with nan in labels **********
***** Training *****
input_x shape: (144, 18, 8)
target shape: (144, 1)
********** Removing Examples with nan in labels **********
***** Validation *****
input_x shape: (36, 18, 8)
target shape: (36, 1)
********** Removing Examples with nan in labels **********
***** Test *****
input_x shape: (78, 18, 8)
target shape: (78, 1)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
overflow encountered in multiply
overflow encountered in square
overflow encountered in square
overflow encountered in square
overflow encountered in square
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
[28]:
print(nrmses)
train test
log 0.110399 inf
box_cox 0.098330 0.723063
quantile 0.097928 0.210205
yeo-johnson 0.097094 0.197983
minmax 0.086265 0.186763
zscore 0.085452 0.117659
vast 0.087762 0.115276
pareto 0.096659 0.109382
None 0.107808 0.105556
mmad 0.090300 0.081323
robust 0.081920 0.067028
[31]:
********** Removing Examples with nan in labels **********
***** Training *****
input_x shape: (144, 18, 8)
target shape: (144, 1)
********** Removing Examples with nan in labels **********
***** Validation *****
input_x shape: (36, 18, 8)
target shape: (36, 1)
********** Removing Examples with nan in labels **********
***** Test *****
input_x shape: (78, 18, 8)
target shape: (78, 1)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
overflow encountered in multiply
overflow encountered in square
overflow encountered in square
overflow encountered in square
overflow encountered in square
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
divide by zero encountered in true_divide
[37]:
_ = experiment.compare_edf_plots(data=data, exclude=["log"])
********** Removing Examples with nan in labels **********
***** Training *****
input_x shape: (144, 18, 8)
target shape: (144, 1)
********** Removing Examples with nan in labels **********
***** Validation *****
input_x shape: (36, 18, 8)
target shape: (36, 1)
********** Removing Examples with nan in labels **********
***** Test *****
input_x shape: (78, 18, 8)
target shape: (78, 1)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
[42]:
_ = experiment.compare_regression_plots(data=data)
********** Removing Examples with nan in labels **********
***** Training *****
input_x shape: (144, 18, 8)
target shape: (144, 1)
********** Removing Examples with nan in labels **********
***** Validation *****
input_x shape: (36, 18, 8)
target shape: (36, 1)
********** Removing Examples with nan in labels **********
***** Test *****
input_x shape: (78, 18, 8)
target shape: (78, 1)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
[41]:
_ = experiment.compare_residual_plots(data=data)
********** Removing Examples with nan in labels **********
***** Training *****
input_x shape: (144, 18, 8)
target shape: (144, 1)
********** Removing Examples with nan in labels **********
***** Validation *****
input_x shape: (36, 18, 8)
target shape: (36, 1)
********** Removing Examples with nan in labels **********
***** Test *****
input_x shape: (78, 18, 8)
target shape: (78, 1)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
assigning name Input to IteratorGetNext:0 with shape (None, 18, 8)
[40]:
_ = experiment.loss_comparison()
