This Get Started guide is intended as a quick example to start programming with GRFlood .

► Let's install our package.

In : pip install GRFlood

► Upload our class Flood.

In : from GRFlood import Flood

► Creates an instance of the class Flood.

In : model = Flood('Example_Water_Volume.csv')

Out : Example_Water_Volume.csv(application/vnd.ms-excel) - 99369 bytes, last modified: 23/02/2021 - 100% done
Saving Example_Water_Volume.csv to Example_Water_Volume.csv
            new Q (l/s)
Date                   
1999-01-01          910
1999-01-02          900
1999-01-03          900
1999-01-04          890
1999-01-05          890
...                 ...
2016-12-27          290
2016-12-28          290
2016-12-29          290
2016-12-30          290
2016-12-31          290

[6575 rows x 1 columns]

► To view our data you can print.

In : model.data

Out : new Q (l/s)
Date    
1999-01-01  910
1999-01-02  900
1999-01-03  900
1999-01-04  890
1999-01-05  890
... ...
2016-12-27  290
2016-12-28  290
2016-12-29  290
2016-12-30  290
2016-12-31  290
6575 rows × 1 columns

► Normalized our data.

In : model.normalized_features(model.data)

Out: [[0.0155052 ]
 [0.01533481]
 [0.01533481]
 ...
 [0.00494122]
 [0.00494122]
 [0.00494122]]

► To view our normalized data you can print.

In : model.dataset

Out : array([[0.0155052 ],
       [0.01533481],
       [0.01533481],
       ...,
       [0.00494122],
       [0.00494122],
       [0.00494122]])

► Split our data.

In : model.split(model.dataset, ratio = 75)

Out : 6575
train size: 4931 and test size : 1644

► To view more details about split.

In : print("train data :{} \n test data : {} \n train size : {} \n test size : {}  ".format(model.train,model.test,model.train_size,model.test_size))

Out : train data :[[0.0155052 ]
 [0.01533481]
 [0.01533481]
 ...
 [0.00460044]
 [0.00460044]
 [0.00443006]] 
 test data : [[0.00425967]
 [0.00425967]
 [0.00408928]
 ...
 [0.00494122]
 [0.00494122]
 [0.00494122]] 
 train size : 4931 
 test size : 1644  

► Checking stationarity, we’ll be using the rolling statistics plots along with Dickey-Fuller test results.
For Dicky Fuller Test :

Null hypothesis in ADF test is that Data is not stationary.

• It’s better to have the p-value to be smaller than 0.05 inorder to reject the null hypothesis and consider data as stationary.
• If Test Statistic < Critical Values => Rejects the null hypothesis.
• If Test Statistic >Critical Values => failed to reject the null hypothesis

In : model.test_stationarity(model.data['new Q (l/s)'])

► In our example we can understand that my test statistics value = -8.097405e+00 is smaller than critical value(5%) = -2.861982e+00, Thus the data is stationary. If the data is not stationary we have to tranform it to make the data more stationary, using differencing technique for example.

Out : Results of Dickey-Fuller Test:
Test Statistic                -8.097405e+00
p-value                        1.327622e-12
#Lags Used                     3.400000e+01
Number of Observations Used    6.540000e+03
Critical Value (1%)           -3.431350e+00
Critical Value (5%)           -2.861982e+00
Critical Value (10%)          -2.567005e+00
dtype: float64

► We now experiment with the following three models:

1. Artificial Neural Network
2. Recurrent Neural Network
3. Long Short-Term Memory (a RNN with a LSTM layer)

► Plot of the predictions made by the ANN.

In : model.ANN_model(train = model.train, test = model.test, train_size = model.train_size, data = model.data, look_back = 20)

Out : Test Score: 1858.73 RMSE
Normalized RMSE: 0.03167030826699024

► Plot of the predictions made by the LSTM.

In : model.LSTM_model(train = model.train, test = model.test, train_size = model.train_size, data = model.data, look_back = 20)

Out : Test Score: 1828.80 RMSE
Normalized RMSE: 0.031160290158361494 

► Plot of the predictions made by the RNN.

In : model.RNN(train = model.train, test = model.test, train_size = model.train_size, data = model.data, look_back = 20)

Out : Test Score: 1946.10 RMSE
Normalized RMSE: 0.033158903327623425