Preprocess data for LSTM

Module for preparing LSTM input data from preprocessed features. Includes functionality for loading feature arrays, creating sequences, splitting the dataset, and saving the output for training and testing.

This module is intended for use with the Digital Twin of 5G Network project.

lstm_preprocessing.create_sequences(X, y, seq_len)

Vytvorí sekvencie vstupných dát pre LSTM z kĺzavého okna.

Parameters:

  • X (np.ndarray) – Vstupné dáta (features)

  • y (np.ndarray) – Cieľové hodnoty (triedy)

  • seq_len (int) – Dĺžka sekvencie pre LSTM

Returns:

(X_seq, y_seq) ako ndarray

Return type:

tuple

lstm_preprocessing.load_data(X_path, Y_path, scaler_path, features_path, uc_map_path)

Loads the preprocessed data from specified paths.

Args

  • X_path (str): Path to the input features (X)

  • Y_path (str): Path to the target labels (y)

  • scaler_path (str): Path to the scaler object

  • features_path (str): Path to the selected features JSON file

  • uc_map_path (str): Path to the UC map JSON file

Returns

  • tuple: (X, y, scaler, selected_features, uc_map)

lstm_preprocessing.split_and_save_data(X_seq, y_seq, output_dir='preprocessed_data')

Splits the dataset into training and testing sets and saves them to disk.

Args

  • X_seq (np.ndarray): Input features in sequence format

  • y_seq (np.ndarray): Target labels in sequence format

  • output_dir (str): Directory to save the split data

Returns

None