Lab2.py - master - Durbolay/Lab2

Lab2.py
148 строк5 KB
Durbolay
create Lab2.py
17 ноя 2025, 18:21
17 ноя 2025, 18:21c09c7d9
# Лабораторная работа №2: Прогнозирование временных рядов и классификация текстов
import numpy as np
import tensorflow as tf
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import SimpleRNN, LSTM, GRU, Dense, Embedding
from sklearn.preprocessing import MinMaxScaler
from sklearn.metrics import mean_squared_error
import matplotlib.pyplot as plt

print("=== ЧАСТЬ 1: ПРОГНОЗИРОВАНИЕ ВРЕМЕННЫХ РЯДОВ ===")

def generate_time_series(n_steps):
    time = np.linspace(0, 1, n_steps)
    series = 0.5 * np.sin((time) * (10 + 10)) + 0.2 * np.sin((time) * (20 + 20)) + 0.1 * (np.random.rand(n_steps) - 0.5)
    return series

n_steps = 1000
series = generate_time_series(n_steps)

plt.figure(figsize=(12, 4))
plt.plot(series)
plt.title('Сгенерированный временной ряд')
plt.show()

split_time = 800
train_series = series[:split_time]
test_series = series[split_time:]

scaler = MinMaxScaler(feature_range=(0, 1))
train_scaled = scaler.fit_transform(train_series.reshape(-1, 1))
test_scaled = scaler.transform(test_series.reshape(-1, 1))

def create_dataset(data, window_size):
    X, y = [], []
    for i in range(window_size, len(data)):
        X.append(data[i-window_size:i, 0])
        y.append(data[i, 0])
    return np.array(X), np.array(y)

window_size = 20
X_train, y_train = create_dataset(train_scaled, window_size)
X_test, y_test = create_dataset(test_scaled, window_size)

X_train = X_train.reshape((X_train.shape[0], X_train.shape[1], 1))
X_test = X_test.reshape((X_test.shape[0], X_test.shape[1], 1))

def build_model(model_type, window_size):
    model = Sequential()
    if model_type == 'SimpleRNN':
        model.add(SimpleRNN(50, activation='tanh', input_shape=(window_size, 1)))
    elif model_type == 'LSTM':
        model.add(LSTM(50, activation='tanh', input_shape=(window_size, 1)))
    elif model_type == 'GRU':
        model.add(GRU(50, activation='tanh', input_shape=(window_size, 1)))
    model.add(Dense(1))
    model.compile(optimizer='adam', loss='mse')
    return model

model_types = ['SimpleRNN', 'LSTM', 'GRU']
histories = {}
models = {}

for model_type in model_types:
    print(f"Обучение модели {model_type}...")
    model = build_model(model_type, window_size)
    history = model.fit(X_train, y_train, epochs=50, batch_size=32, validation_data=(X_test, y_test), verbose=0)
    models[model_type] = model
    histories[model_type] = history

plt.figure(figsize=(12, 6))
for model_type in model_types:
    plt.plot(histories[model_type].history['val_loss'], label=f'{model_type} Val Loss')
plt.title('Сравнение потерь на валидации (MSE)')
plt.ylabel('MSE')
plt.xlabel('Эпоха')
plt.legend()
plt.show()

plt.figure(figsize=(15, 10))
for i, model_type in enumerate(model_types):
    y_pred = models[model_type].predict(X_test)
    y_test_inv = scaler.inverse_transform(y_test.reshape(-1, 1))
    y_pred_inv = scaler.inverse_transform(y_pred)
    mse = mean_squared_error(y_test_inv, y_pred_inv)

    plt.subplot(3, 1, i+1)
    plt.plot(y_test_inv, label='Истинные значения', linewidth=2)
    plt.plot(y_pred_inv, label='Предсказания', linewidth=1)
    plt.title(f'{model_type} (MSE: {mse:.4f})')
    plt.legend()
plt.tight_layout()
plt.show()

print("=== ЧАСТЬ 2: КЛАССИФИКАЦИЯ ТЕКСТОВ ===")

from tensorflow.keras.datasets import imdb
from tensorflow.keras.preprocessing.sequence import pad_sequences

vocab_size = 10000
max_len = 200
(X_train, y_train), (X_test, y_test) = imdb.load_data(num_words=vocab_size)

X_train = pad_sequences(X_train, maxlen=max_len)
X_test = pad_sequences(X_test, maxlen=max_len)

def build_text_model(model_type):
    model = Sequential()
    model.add(Embedding(vocab_size, 32))
    if model_type == 'LSTM':
        model.add(LSTM(32))
    elif model_type == 'GRU':
        model.add(GRU(32))
    model.add(Dense(1, activation='sigmoid'))
    model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])
    return model

print("Обучение LSTM модели...")
model_lstm = build_text_model('LSTM')
history_lstm = model_lstm.fit(X_train, y_train, epochs=5, batch_size=128, validation_split=0.2, verbose=1)

print("Обучение GRU модели...")
model_gru = build_text_model('GRU')
history_gru = model_gru.fit(X_train, y_train, epochs=5, batch_size=128, validation_split=0.2, verbose=1)

lstm_test_loss, lstm_test_acc = model_lstm.evaluate(X_test, y_test, verbose=0)
gru_test_loss, gru_test_acc = model_gru.evaluate(X_test, y_test, verbose=0)

print(f"LSTM Точность на тесте: {lstm_test_acc:.4f}")
print(f"GRU Точность на тесте: {gru_test_acc:.4f}")

plt.figure(figsize=(12, 5))
plt.subplot(1, 2, 1)
plt.plot(history_lstm.history['val_accuracy'], label='LSTM Val Accuracy')
plt.plot(history_gru.history['val_accuracy'], label='GRU Val Accuracy')
plt.title('Сравнение точности на валидации')
plt.ylabel('Accuracy')
plt.xlabel('Эпоха')
plt.legend()

plt.subplot(1, 2, 2)
plt.plot(history_lstm.history['val_loss'], label='LSTM Val Loss')
plt.plot(history_gru.history['val_loss'], label='GRU Val Loss')
plt.title('Сравнение потерь на валидации')
plt.ylabel('Loss')
plt.xlabel('Эпоха')
plt.legend()
plt.tight_layout()
plt.show()