Spaces:
Sleeping
Sleeping
File size: 8,045 Bytes
8474315 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 |
# Class for preprocessing the data before the training phase
import os
import pandas as pd
import numpy as np
import logging
from sklearn.preprocessing import StandardScaler
from adtk.data import validate_series, to_events
from adtk.detector import SeasonalAD
from path_config import DATA_DIR
class Preprocessor:
def __init__(self):
self.scaler = StandardScaler()
self.raw_data = None
self.anomalies_events = None
self.logger = logging.getLogger(__name__)
self.window_size = None
def preprocess_data(self, file_path, val_split="2014-10-01", test_split="2014-10-16", window_size=48):
"""
Preprocess the raw data
:param window_size: The size of the sliding window, default is 48
:param file_path: Path to the CSV file
"""
# Load the raw data
self.logger.info("Loading raw data...")
self.load_raw_data(file_path)
# Detect anomalies and convert them to events
self.logger.info("Detecting anomalies...")
self.load_anomalies_events()
# Label the anomalies in the raw data
self.logger.info("Labeling anomalies...")
self._label_anomalies()
# Split the data chronologically
self.logger.info("Splitting the data into training, validation and testing set...")
train_data, val_data, test_data = self._chronological_split(val_split=val_split, test_split=test_split)
# Split the data into features and target
self.logger.info("Splitting the data into features and target...")
X_train, y_train = self._split_features_target(train_data)
X_val, y_val = self._split_features_target(val_data)
X_test, y_test = self._split_features_target(test_data)
# Scale the data
self.logger.info("Scaling the data...")
train_scaled, val_scaled, test_scaled = self._scale_data(X_train, X_val, X_test)
# Create a sliding window of data
self.logger.info(f"Creating sliding window with the length of {window_size} from the data...")
train_sequences = self._create_sliding_window(train_scaled, window_size=window_size)
val_sequences = self._create_sliding_window(val_scaled, window_size=window_size)
test_sequences = self._create_sliding_window(test_scaled, window_size=window_size)
self.window_size = window_size
# Save the preprocessed data
self.logger.info("Saving the preprocessed data...")
self.save_preprocessed_data(train_sequences, "train_features.npy")
self.save_preprocessed_data(val_sequences, "val_features.npy")
self.save_preprocessed_data(test_sequences, "test_features.npy")
self.save_preprocessed_data(y_train.values, "train_labels.npy")
self.save_preprocessed_data(y_val.values, "val_labels.npy")
self.save_preprocessed_data(y_test.values, "test_labels.npy")
print("Preprocessing completed!")
def load_raw_data(self, file_path):
"""
Load raw data from a CSV file
:param file_path: Path to the CSV file
"""
try:
df = pd.read_csv(
file_path,
usecols=["timestamp", "value"],
index_col="timestamp",
parse_dates=True,
)
df.sort_index(inplace=True)
# Rename the columns
df.rename(columns={"value": "Traffic"}, inplace=True)
df.index.rename("Timestamp", inplace=True)
# Validate the time series
self.raw_data = validate_series(df)
except FileNotFoundError as e:
print(f"File path does not exist: {file_path}")
def load_anomalies_events(self):
"""
Load the anomalies events
"""
if self.raw_data is None:
print("Raw data is not loaded")
return
events = [
('2014-07-04 00:00:00', '2014-07-06 23:59:59'), # Independence Day Celebration
('2014-09-01 00:00:00', '2014-09-01 23:59:59'), # Labour Day
('2014-11-02 00:00:00', '2014-11-02 11:59:59'), # NYC Marathon 2014
('2014-11-27 00:00:00', '2014-11-27 23:59:59'), # Thanksgiving Day
('2014-12-25 00:00:00', '2014-12-26 23:59:59'), # Christmas Holiday
('2015-01-01 00:00:00', '2015-01-01 23:59:59'), # New Year
('2015-01-26 12:00:00', '2015-01-28 11:59:59') # Snowstorm
]
# Store the events
self.anomalies_events = events
def _label_anomalies(self):
"""
Label the anomalies in the raw data
"""
if self.raw_data is None:
print("Raw data is not loaded")
return
if self.anomalies_events is None:
print("Anomalies are not detected")
return
# Label the anomalies as 1 and 0
self.raw_data["Anomaly"] = 0
for start, end in self.anomalies_events:
self.raw_data.loc[start:end, "Anomaly"] = 1
def _chronological_split(self, val_split="2014-10-01", test_split="2014-10-16"):
"""
Split the data chronologically into train, validation, and test sets
:param val_split: Validation split date
:param test_split: Test split date
"""
if self.raw_data is None:
print("Raw data is not loaded")
return
# Split the data
train_data = self.raw_data.loc[self.raw_data.index < val_split]
val_data = self.raw_data.loc[
(self.raw_data.index >= val_split) & (self.raw_data.index < test_split)
]
test_data = self.raw_data.loc[self.raw_data.index >= test_split]
return train_data, val_data, test_data
def _split_features_target(self, data, target_col="Anomaly"):
"""
Split the data into features and target
:param data: DataFrame containing the data
:param target_col: Column to predict
"""
# Split the data into features and target
X = data.drop(columns=[target_col])
y = data[target_col]
return X, y
def _scale_data(self, train_data, val_data, test_data):
"""
Scale the data using StandardScaler
:param train_data: Training data
:param val_data: Validation data
:param test_data: Test data
"""
if self.scaler is None:
self.scaler = StandardScaler()
# Fit and transform the training data
train_scaled = self.scaler.fit_transform(train_data)
val_scaled = self.scaler.transform(val_data)
test_scaled = self.scaler.transform(test_data)
return train_scaled, val_scaled, test_scaled
def _create_sliding_window(self, data, window_size=48, step_size=1):
"""
Create a sliding window of data
:param data: Scaled data
:param window_size: Size of the window
:param step_size: Step size for the window
"""
sequences = []
for i in range(0, len(data) - window_size + 1, step_size):
sequences.append(data[i : i + window_size])
return np.array(sequences)
def save_preprocessed_data(self, data, file_path):
"""
Save preprocessed data to a .npy file
:param data: Preprocessed data
:param file_path: Path to save the .npy file
"""
dir_path = os.path.join(DATA_DIR, "preprocessed_data")
if not os.path.exists(dir_path):
os.makedirs(dir_path)
file_path = os.path.join(dir_path, file_path)
np.save(file_path, data)
self.logger.info(f"{file_path} has been saved successfully!")
def get_seq_length(self):
"""
Get the length of the sequence
"""
return self.window_size |