from typing import Any, Callable, List, Dict, Union, Optional, Sequence, Tuple from numpy import ndarray from collections import OrderedDict from scipy import sparse import os import sklearn import numpy as np import typing import pandas as pd from keras.models import Sequential, load_model from keras.callbacks import History, EarlyStopping, Callback from keras.layers.recurrent import LSTM from keras.layers.core import Dense, Activation, Dropout from keras.layers import Flatten from d3m import container, utils from d3m.base import utils as base_ut from d3m.container.numpy import ndarray as d3m_ndarray from d3m.container import DataFrame as d3m_dataframe from d3m.metadata import hyperparams, params, base as metadata_base from d3m import utils from d3m.base import utils as base_utils from d3m.exceptions import PrimitiveNotFittedError from d3m.primitive_interfaces.base import CallResult, DockerContainer from d3m.primitive_interfaces.unsupervised_learning import UnsupervisedLearnerPrimitiveBase from d3m.primitive_interfaces.transformer import TransformerPrimitiveBase from detection_algorithm.UODBasePrimitive import Params_ODBase, Hyperparams_ODBase, UnsupervisedOutlierDetectorBase from detection_algorithm.core.CollectiveBase import CollectiveBaseDetector from sklearn.utils import check_array # from d3m.primitive_interfaces.base import ProbabilisticCompositionalityMixin, ContinueFitMixin from d3m import exceptions # from detection_algorithm.UODBasePrimitive import Params_ODBase, Hyperparams_ODBase, UnsupervisedOutlierDetectorBase from detection_algorithm.core.utils.errors import Errors from detection_algorithm.core.utils.channel import Channel from detection_algorithm.core.utils.modeling import Model # from pyod.models.base import BaseDetector __all__ = ('Telemanom',) Inputs = container.DataFrame Outputs = container.DataFrame class Params(Params_ODBase): ######## Add more Attributes ####### pass class Hyperparams(Hyperparams_ODBase): smoothing_perc = hyperparams.Hyperparameter[float]( default=0.05, semantic_types=['https://metadata.datadrivendiscovery.org/types/TuningParameter'], description="determines window size used in EWMA smoothing (percentage of total values for channel)" ) window_size_ = hyperparams.Hyperparameter[int]( default=100, semantic_types=['https://metadata.datadrivendiscovery.org/types/ControlParameter'], description="number of trailing batches to use in error calculation" ) error_buffer = hyperparams.Hyperparameter[int]( default=50, semantic_types=['https://metadata.datadrivendiscovery.org/types/ControlParameter'], description="number of values surrounding an error that are brought into the sequence (promotes grouping on nearby sequences" ) batch_size = hyperparams.Hyperparameter[int]( default=70, semantic_types=['https://metadata.datadrivendiscovery.org/types/ControlParameter'], description="Batch size while predicting" ) # LSTM Model Parameters dropout = hyperparams.Hyperparameter[float]( default=0.3, semantic_types=['https://metadata.datadrivendiscovery.org/types/TuningParameter'], description="Dropout rate" ) validation_split = hyperparams.Hyperparameter[float]( default=0.2, semantic_types=['https://metadata.datadrivendiscovery.org/types/ControlParameter'], description="Validation split" ) optimizer = hyperparams.Hyperparameter[typing.Union[str, None]]( default='Adam', semantic_types=['https://metadata.datadrivendiscovery.org/types/ControlParameter'], description="Optimizer" ) lstm_batch_size = hyperparams.Hyperparameter[int]( default=64, semantic_types=['https://metadata.datadrivendiscovery.org/types/ControlParameter'], description="lstm model training batch size" ) loss_metric = hyperparams.Hyperparameter[typing.Union[str, None]]( default='mean_squared_error', semantic_types=['https://metadata.datadrivendiscovery.org/types/ControlParameter'], description="loss function" ) layers = hyperparams.List( elements=hyperparams.Hyperparameter[int](1), default=[10,10], semantic_types=['https://metadata.datadrivendiscovery.org/types/TuningParameter'], description="No of units for the 2 lstm layers" ) # Training Parameters epochs = hyperparams.Hyperparameter[int]( default=1, semantic_types=['https://metadata.datadrivendiscovery.org/types/TuningParameter'], description="Epoch" ) patience = hyperparams.Hyperparameter[int]( default=10, semantic_types=['https://metadata.datadrivendiscovery.org/types/ControlParameter'], description="Number of consequetive training iterations to allow without decreasing the val_loss by at least min_delta" ) min_delta = hyperparams.Hyperparameter[float]( default=0.0003, semantic_types=['https://metadata.datadrivendiscovery.org/types/TuningParameter'], description="Number of consequetive training iterations to allow without decreasing the val_loss by at least min_delta" ) l_s = hyperparams.Hyperparameter[int]( default=100, semantic_types=['https://metadata.datadrivendiscovery.org/types/TuningParameter'], description="num previous timesteps provided to model to predict future values" ) n_predictions = hyperparams.Hyperparameter[int]( default=10, semantic_types=['https://metadata.datadrivendiscovery.org/types/ControlParameter'], description="number of steps ahead to predict" ) # Error thresholding parameters # ================================== p = hyperparams.Hyperparameter[float]( default=0.05, semantic_types=['https://metadata.datadrivendiscovery.org/types/TuningParameter'], description="minimum percent decrease between max errors in anomalous sequences (used for pruning)" ) # Contamination contamination = hyperparams.Uniform( lower=0., upper=0.5, default=0.1, description='the amount of contamination of the data set, i.e.the proportion of outliers in the data set. Used when fitting to define the threshold on the decision function', semantic_types=['https://metadata.datadrivendiscovery.org/types/TuningParameter'] ) class TelemanomPrimitive(UnsupervisedOutlierDetectorBase[Inputs, Outputs, Params, Hyperparams]): """ A primitive that uses telmanom for outlier detection Parameters ---------- """ __author__ = "Data Lab" metadata = metadata_base.PrimitiveMetadata( { '__author__' : "DATA Lab at Texas A&M University", 'name': "Telemanom", 'python_path': 'd3m.primitives.tods.detection_algorithm.telemanom', 'source': { 'name': 'DATA Lab at Texas A&M University', 'contact': 'mailto:khlai037@tamu.edu', 'uris': [ 'https://gitlab.com/lhenry15/tods.git', 'https://gitlab.com/lhenry15/tods/-/blob/purav/anomaly-primitives/anomaly_primitives/telemanom.py', ], }, 'algorithm_types': [ metadata_base.PrimitiveAlgorithmType.TELEMANOM, ], 'primitive_family': metadata_base.PrimitiveFamily.ANOMALY_DETECTION, 'id': 'c7259da6-7ce6-42ad-83c6-15238679f5fa', 'hyperparameters_to_tune':['layers','loss_metric','optimizer','epochs','p','l_s','patience','min_delta','dropout','smoothing_perc'], 'version': '0.0.1', }, ) def __init__(self, *, hyperparams: Hyperparams, # random_seed: int = 0, docker_containers: Dict[str, DockerContainer] = None) -> None: super().__init__(hyperparams=hyperparams, random_seed=random_seed, docker_containers=docker_containers) self._clf = Detector(smoothing_perc=self.hyperparams['smoothing_perc'], window_size=self.hyperparams['window_size_'], error_buffer=self.hyperparams['error_buffer'], batch_size = self.hyperparams['batch_size'], validation_split = self.hyperparams['validation_split'], optimizer = self.hyperparams['optimizer'], lstm_batch_size = self.hyperparams['lstm_batch_size'], loss_metric = self.hyperparams['loss_metric'], layers = self.hyperparams['layers'], epochs = self.hyperparams['epochs'], patience = self.hyperparams['patience'], min_delta = self.hyperparams['min_delta'], l_s = self.hyperparams['l_s'], n_predictions = self.hyperparams['n_predictions'], p = self.hyperparams['p'], contamination=hyperparams['contamination'] ) def set_training_data(self, *, inputs: Inputs) -> None: """ Set training data for outlier detection. Args: inputs: Container DataFrame Returns: None """ super().set_training_data(inputs=inputs) def fit(self, *, timeout: float = None, iterations: int = None) -> CallResult[None]: """ Fit model with training data. Args: *: Container DataFrame. Time series data up to fit. Returns: None """ return super().fit() def produce(self, *, inputs: Inputs, timeout: float = None, iterations: int = None) -> CallResult[Outputs]: """ Process the testing data. Args: inputs: Container DataFrame. Time series data up to outlier detection. Returns: Container DataFrame 1 marks Outliers, 0 marks normal. """ return super().produce(inputs=inputs, timeout=timeout, iterations=iterations) def produce_score(self, *, inputs: Inputs, timeout: float = None, iterations: int = None) -> CallResult[Outputs]: """ Process the testing data. Args: inputs: Container DataFrame. Time series data up to outlier detection. Returns: Container DataFrame Outlier score of input DataFrame. """ return super().produce_score(inputs=inputs, timeout=timeout, iterations=iterations) def get_params(self) -> Params: """ Return parameters. Args: None Returns: class Params """ return super().get_params() def set_params(self, *, params: Params) -> None: """ Set parameters for outlier detection. Args: params: class Params Returns: None """ super().set_params(params=params) class Detector(CollectiveBaseDetector): """Class to Implement Deep Log LSTM based on "https://www.cs.utah.edu/~lifeifei/papers/deeplog.pdf Only Parameter Value anomaly detection layer has been implemented for time series data""" def __init__(self,smoothing_perc=0.05,window_size = 10,error_buffer = 5,batch_size =30, \ dropout = 0.3, validation_split=0.2,optimizer='adam',lstm_batch_size=64,loss_metric='mean_squared_error', \ layers=[40,40],epochs = 1,patience =10,min_delta=0.0003,l_s=5,n_predictions=2,p = 0.05,contamination=0.1): # super(Detector, self).__init__(contamination=contamination) super(Detector, self).__init__(contamination=contamination, window_size=l_s, step_size=1, ) self._smoothin_perc = smoothing_perc self._window_size =window_size self._error_buffer = error_buffer self._batch_size = batch_size self._dropout = dropout self._validation_split = validation_split self._optimizer = optimizer self._lstm_batch_size = lstm_batch_size self._loss_metric = loss_metric self._layers = layers self._epochs = epochs self._patience = patience self._min_delta = min_delta self._l_s = l_s self._n_predictions = n_predictions self._p = p self.contamination = contamination # self.y_hat = None self.results = [] self.result_df = None self._model = None self._channel = None def fit(self,X,y=None): """ Fit data to LSTM model. Args: inputs : X , ndarray of size (number of sample,features) Returns: return : self object with trained model """ X = check_array(X).astype(np.float) self._set_n_classes(None) inputs = X self._channel = Channel(n_predictions = self._n_predictions,l_s = self._l_s) self._channel.shape_train_data(inputs) self._model = Model(self._channel,patience = self._patience, min_delta =self._min_delta, layers = self._layers, dropout = self._dropout, n_predictions = self._n_predictions, loss_metric = self._loss_metric, optimizer = self._optimizer, lstm_batch_size = self._lstm_batch_size, epochs = self._epochs, validation_split = self._validation_split, batch_size = self._batch_size, l_s = self._l_s ) self.decision_scores_, self.left_inds_, self.right_inds_ = self.decision_function(X) self._process_decision_scores() return self def decision_function(self, X: np.array): """Predict raw anomaly scores of X using the fitted detector. The anomaly score of an input sample is computed based on the fitted detector. For consistency, outliers are assigned with higher anomaly scores. Parameters ---------- X : numpy array of shape (n_samples, n_features) The input samples. Sparse matrices are accepted only if they are supported by the base estimator. Returns ------- anomaly_scores : numpy array of shape (n_samples,) The anomaly score of the input samples. """ X = check_array(X).astype(np.float) self._set_n_classes(None) inputs = X self._channel.shape_test_data(inputs) self._channel = self._model.batch_predict(channel = self._channel) errors = Errors(channel = self._channel, window_size = self._window_size, batch_size = self._batch_size, smoothing_perc = self._smoothin_perc, n_predictions = self._n_predictions, l_s = self._l_s, error_buffer = self._error_buffer, p = self._p ) # prediciton smoothed error prediction_errors = np.reshape(errors.e_s,(self._channel.X_train.shape[0],self._channel.X_train.shape[2])) prediction_errors = np.sum(prediction_errors,axis=1) left_indices = [] right_indices = [] scores = [] for i in range(len(prediction_errors)): left_indices.append(i) right_indices.append(i+self._l_s) scores.append(prediction_errors[i]) return np.asarray(scores),np.asarray(left_indices),np.asarray(right_indices) # if __name__ == "__main__": # csv = pd.read_csv("/home/purav/Downloads/yahoo_train.csv") # # X_train = np.asarray( # # [3., 4., 8., 16, 18, 13., 22., 36., 59., 128, 62, 67, 78, 100]).reshape(-1, 1) # # X_test = np.asarray( # # [3., 4., 8.6, 13.4, 22.5, 17, 19.2, 36.1, 127, -23, 59.2]).reshape(-1,1) # # print(X_train.shape, X_test.shape) # X_train = csv.iloc[:,[2,3,4,5,6]].values # clf = Detector(contamination=0.1) # clf.fit(X_train) # # pred_scores = clf.decision_function(X_test) # pred_labels = clf.predict(X_train) # print(clf.threshold_) # # print(np.percentile(pred_scores, 100 * 0.9)) # # print('pred_scores: ',pred_scores) # print('scores: ',pred_labels[0].shape) # print('left_indices: ',pred_labels[1].shape) # print('right_indices: ',pred_labels[2].shape)