jeevan0704/sequential-model-for-sequential_dataset

1"""2Utility functions for data preprocessing and model handling3"""4import numpy as np5 6def normalize_sequence(sequence, method='minmax'):7    """8    Normalize a sequence using specified method9    10    Args:11        sequence: Input sequence (numpy array or list)12        method: 'minmax' or 'standard'13    14    Returns:15        Normalized sequence16    """17    sequence = np.array(sequence)18    19    if method == 'minmax':20        min_val = np.min(sequence)21        max_val = np.max(sequence)22        if max_val - min_val == 0:23            return sequence24        return (sequence - min_val) / (max_val - min_val)25    26    elif method == 'standard':27        mean = np.mean(sequence)28        std = np.std(sequence)29        if std == 0:30            return sequence31        return (sequence - mean) / std32    33    return sequence34 35def pad_sequence(sequence, max_length, padding_value=0):36    """37    Pad sequence to a fixed length38    39    Args:40        sequence: Input sequence41        max_length: Target length42        padding_value: Value to use for padding43    44    Returns:45        Padded sequence46    """47    sequence = np.array(sequence)48    49    if len(sequence) >= max_length:50        return sequence[:max_length]51    52    padding = np.full(max_length - len(sequence), padding_value)53    return np.concatenate([sequence, padding])54 55def reshape_for_lstm(data, timesteps, features):56    """57    Reshape data for LSTM input (samples, timesteps, features)58    59    Args:60        data: Input data61        timesteps: Number of timesteps62        features: Number of features63    64    Returns:65        Reshaped data66    """67    data = np.array(data)68    69    if len(data.shape) == 1:70        # Flatten array, reshape to (samples, timesteps, features)71        total_elements = data.shape[0]72        samples = total_elements // (timesteps * features)73        return data[:samples * timesteps * features].reshape(samples, timesteps, features)74    75    return data.reshape(-1, timesteps, features)76 77def validate_input_shape(data, expected_shape):78    """79    Validate that input data matches expected shape80    81    Args:82        data: Input data83        expected_shape: Tuple of expected dimensions84    85    Returns:86        Boolean indicating if shape is valid87    """88    data = np.array(data)89    90    if len(data.shape) != len(expected_shape):91        return False92    93    for i, (actual, expected) in enumerate(zip(data.shape, expected_shape)):94        if expected is not None and actual != expected:95            return False96    97    return True98 99def format_prediction_output(prediction, class_names=None):100    """101    Format prediction output for display102    103    Args:104        prediction: Model prediction (numpy array)105        class_names: Optional list of class names106    107    Returns:108        Formatted prediction dictionary109    """110    prediction = np.array(prediction)111    112    result = {113        'raw_prediction': prediction.tolist()114    }115    116    # Handle classification predictions117    if len(prediction.shape) == 2 and prediction.shape[1] > 1:118        predicted_class = np.argmax(prediction, axis=1)[0]119        confidence = np.max(prediction, axis=1)[0]120        121        result['predicted_class'] = int(predicted_class)122        result['confidence'] = float(confidence)123        124        if class_names and predicted_class < len(class_names):125            result['class_name'] = class_names[predicted_class]126        127        # Add probabilities for all classes128        result['probabilities'] = {129            f'class_{i}': float(prob) 130            for i, prob in enumerate(prediction[0])131        }132    133    # Handle regression predictions134    elif len(prediction.shape) == 1 or prediction.shape[1] == 1:135        result['value'] = float(prediction.flatten()[0])136    137    return result138