An Efficient Communication Intrusion Detection Scheme in AMI Combining Feature Dimensionality Reduction and Improved LSTM

Algorithm 1

Algorithm description of data preprocessing stage.

	Input: original training dataset Original_train, testing dataset Original_test
Output: preprocessed training dataset Preprocessed_train, testing dataset Preprocessed_test
train = pd.read_csv (Original_train)
test = pd.read_csv (Original_train)
/ concat() complete data splicing /
Spliced_data = pd.concat([train, test])
/ get_dummies() complete one-hot encoding /
Encoded_data = get_dummies(Spl_data, [“Feature_1”, “Feature_2”, …, “Feature_n”])
Encoded_data.drop([“label”, “attack_cat”])
/MinMaxScaler() normalizes the data to [0, 1] /
Preprocessed_train = MinMaxScaler (Encoded_data, train, feature_range = (0, 1))
Preprocessed_test = MinMaxScaler (Encoded_data, test, feature_range = (0, 1))
End