Zad 05.Biblioteki DL - prosta sieć neuronowa

process_dataset.py

@@ -1,25 +1,126 @@
 import pandas as pd
+import numpy as np
 from sklearn.model_selection import train_test_split
+from sklearn import metrics
+import matplotlib.pyplot as plt
+import tensorflow as tf
+import math
 
-energy_data = pd.read_csv('Steel_industry_data.csv')
+from tensorflow import keras
+#from tensorflow.keras import layers
 
-train_data, test_data = train_test_split(energy_data, test_size=7008, random_state=1)
-test_data, dev_data = train_test_split(test_data, test_size=3504, random_state=1)
+
+def process_data_and_get_x_y(data):
+    data.columns = ["date", "Usage_kWh", "Lagging_Current_Reactive.Power_kVarh", "Leading_Current_Reactive_Power_kVarh",
+                    "CO2(tCO2)", "Lagging_Current_Power_Factor", "Leading_Current_Power_Factor", "WeekStatus",
+                    "Day_of_week", "Load_Type"] #without NSM column
+    data = data.set_index('date')
+    data = pd.get_dummies(data, drop_first=True)
+
+    x = data.drop('Usage_kWh', axis=1)
+    #x = data['Lagging_Current_Reactive.Power_kVarh']
+    y = data['Usage_kWh']
+    return data, x, y
+
+
+def plot_loss(history):
+    plt.plot(history.history['loss'], label='loss')
+    plt.plot(history.history['val_loss'], label='val_loss')
+    plt.ylim([0, 10])
+    plt.xlabel('Epoch')
+    plt.ylabel('Error')
+    plt.legend()
+    plt.grid(True)
+    plt.show()
+
+
+def show_result(x, y):
+    plt.title('One variable Model', fontsize=15, color='g', pad=12)
+    plt.plot(x, y, 'o', color='r')
+
+    m, b = np.polyfit(x, y, 1)
+    plt.plot(x, m * x + b, color='darkblue')
+    plt.xlabel('Actual')
+    plt.ylabel('Predicted')
+    plt.show()
+
+
+energy_data_train = pd.read_csv('Steel_industry_data_train.csv')
+energy_data_test = pd.read_csv('Steel_industry_data_test.csv')
+
+energy_data_test, x_test, y_test = process_data_and_get_x_y(energy_data_test)
+energy_data_train, x_train, y_train = process_data_and_get_x_y(energy_data_train)
+
+#x_train, x_test, y_train, y_test = train_test_split(x_data, y_data, test_size=0.2, random_state=1)
+# x_test, x_dev, y_test, y_dev = train_test_split(x_test, y_test, test_size=0.5, random_state=1)
 
 # stats
-print(energy_data.describe(include='all'))
+print(energy_data_test.describe(include='all'))
+print(x_train.describe(include='all'))
+#print(np.array(x_train).reshape(-1, 1))
 
-print('Training set size:')
-print(train_data.shape)
-print('Testing set size:')
-print(test_data.shape)
-print('Dev set size:')
-print(dev_data.shape)
+normalizer = tf.keras.layers.Normalization(axis=1)
+normalizer.adapt(np.array(x_train))
+print(normalizer.mean.numpy())
 
-#print(train_data.describe(include='all'))
-#print(test_data.describe(include='all'))
-#print(dev_data.describe(include='all'))
+# powinno być niezmienione
+print(np.array(x_train[:1]))
 
-test_data.to_csv("steel_industry_data_test.csv", encoding="utf-8", index=False)
-dev_data.to_csv("steel_industry_data_dev.csv", encoding="utf-8", index=False)
-train_data.to_csv("steel_industry_data_train.csv", encoding="utf-8", index=False)
+#usage = np.array(x_train)
+#usage_normalizer = keras.layers.Normalization(input_shape=[14, ], axis=1)
+#usage_normalizer.adapt(usage)
+
+usage_model = tf.keras.Sequential([
+    normalizer,
+    keras.layers.Dense(units=10, activation='relu'),
+    keras.layers.Dense(units=1)
+])
+
+print(usage_model.summary())
+
+usage_model.compile(
+    optimizer=tf.optimizers.Adam(learning_rate=0.1),
+    loss='mean_absolute_error')
+
+history = usage_model.fit(
+    x_train,
+    y_train,
+    epochs=100,
+    # Suppress logging.
+    verbose=0,
+    # Calculate validation results on 20% of the training data.
+    validation_split=0.2)
+
+hist = pd.DataFrame(history.history)
+hist['epoch'] = history.epoch
+print(hist.tail())
+
+plot_loss(history)
+
+y_predicted = usage_model.predict(x_test)
+test_results = {}
+test_results['usage_model'] = usage_model.evaluate(
+    x_test,
+    y_test, verbose=0)
+
+print('Mean Absolute Error : ', metrics.mean_absolute_error(y_test, y_predicted))
+print('Mean Squared Error : ', metrics.mean_squared_error(y_test, y_predicted))
+print('Root Mean Squared Error : ', math.sqrt(metrics.mean_squared_error(y_test, y_predicted)))
+
+print(test_results['usage_model'])
+
+show_result(y_test, y_predicted)
+
+#print('Training set size:')
+#print(x_train.shape)
+#print(y_train.shape)
+#print('Testing set size:')
+#print(x_test.shape)
+#print(y_test.shape)
+# print('Dev set size:')
+# print(x_dev.shape)
+# print(y_dev.shape)
+
+# print(train_data.describe(include='all'))
+# print(test_data.describe(include='all'))
+# print(dev_data.describe(include='all'))