Corrections added, tensorflow==2.15.0 compatibility

forecasting/app.py

@@ -1,12 +1,13 @@
 import numpy as np
 import pandas as pd
 import streamlit as st
-from figure import *
 from likelihood.tools import *
+from tensorflow.keras.models import load_model
+
+from figure import *
 
 # This files are in the forecasting folder
 from series import *
-from tensorflow.keras.models import load_model
 
 np.random.seed(0)
 neural_network = load_model("forecasting/models/model_tensor.h5")
@@ -50,7 +51,8 @@
 
 # The dataset is generated
 data_ = generate_series(n, n_steps=100, incline=val)
-data_scale, values = rescale(np.copy(data_))
+scaler = DataScaler(np.copy(data_))
+data_scale = scaler.rescale()
 series = convert_to_df(np.copy(data_)).describe()
 train, test = create_train_test_set(np.copy(data_scale), p_train=p_train)
 neural_network.fit(train[0], train[1])
@@ -82,7 +84,7 @@
     """
 )
 
-prediction = forecasting(model, train, data_.shape[1], values)
+prediction = forecasting(model, train, data_.shape[1], scaler)
 
 fig2 = plot_series(convert_to_df(data_), prediction, i=i_serie)
 st.plotly_chart(fig2)

forecasting/model_tensor.py

@@ -1,3 +1,8 @@
+import os
+
+os.environ["TF_CPP_MIN_LOG_LEVEL"] = "2"
+os.environ["TF_ENABLE_ONEDNN_OPTS"] = "0"
+
 import matplotlib.pyplot as plt
 import numpy as np
 import pandas as pd
@@ -58,8 +63,8 @@ def on_epoch_end(self, epoch, logs):
         print(".", end="")
 
 
-def make_predictions(model, input, n_steps):
-    X = input.copy()
+def make_predictions(model, input_model, n_steps):
+    X = input_model.copy()
     for step_ahead in range(n_steps):
         y_pred_one = model.predict(X[:, step_ahead:])
         y_pred_one = y_pred_one[:, :, np.newaxis]
@@ -69,12 +74,13 @@ def make_predictions(model, input, n_steps):
 
 
 if __name__ == "__main__":
-    num_series = 500
+    num_series = 80
     series_size = 100
     n_steps = 5
 
-    input = generate_series(num_series, series_size, incline=False)
-    y_new = input[:, :-n_steps]
+    input_serie = generate_series(num_series, series_size, incline=False)
+    y_new = input_serie[:, :-n_steps]
+    print(y_new.shape)
     scaler = DataScaler(y_new)
     y_new = scaler.rescale()
     size_ = int(0.8 * y_new.shape[0])
@@ -92,7 +98,7 @@ def make_predictions(model, input, n_steps):
     # model(x_train[:, :, np.newaxis])
     model.compile(loss="mse", optimizer="sgd", metrics=["mae"])
     history = model.fit(
-        x_train[:, :, np.newaxis], y_train, epochs=30, validation_split=0.2, callbacks=[PrintDot()]
+        x_train[:, :, np.newaxis], y_train, epochs=100, validation_split=0.2, callbacks=[PrintDot()]
     )
 
     hist = pd.DataFrame(history.history)
@@ -110,7 +116,7 @@ def make_predictions(model, input, n_steps):
     print(X.shape)
     X = scaler.scale(X)
 
-    plt.plot(range(len(input[0, :])), input[0, :], "o-", label="real value")
+    plt.plot(range(len(input_serie[0, :])), input_serie[0, :], "o-", label="real value")
     plt.plot(range(len(X[0, :]))[-n_steps * 4 :], X[0, :][-n_steps * 4 :], "-r", label="prediction")
     plt.xlabel("Time")
     plt.ylabel("Value")