Module PdmModule.PdMApi

Classes

class PdmModule (detector_class=PdmModule.Handlers.pairdetectionHandler.PairDectionHandler, parameters={'thresholdtype': 'selftunne', 'ProfileSize': 30, 'thresholdfactor': 15, 'ActualProfileSize': 0.25, 'sequencesize': 12, 'alarmsThreshold': 0, 'normz': False})

This class is used to provide an api for streaming usage of Handlers methods. The method collect_data is called to pass data and get prediction and the method collect_event to pass an event to trigger reset of the technique (In current implementation any event passed triger reset.).

detector_class: A tecnique handler wich follow the get_Datapoint, get_event interface from Handlers

parameters : The parameters of the Handlers object to initialize

Expand source code

class PdmModule():
    '''
    This class is used to provide an api for streaming usage of Handlers methods.
    The method collect_data is called to pass data and get prediction
    and the method collect_event to pass an event to trigger reset of the technique (In current implementation any event passed triger reset.).


    detector_class: A tecnique handler wich follow the get_Datapoint, get_event interface from Handlers

    parameters : The parameters of the Handlers object to initialize
    '''
    def __init__(self,detector_class=PairDectionHandler,parameters={
        "thresholdtype" : "selftunne",
        "ProfileSize" : 30,
        "thresholdfactor" : 15,
        "ActualProfileSize" : 0.25,
        "sequencesize" : 12,
        "alarmsThreshold" : 0,
        "normz" : False
    }):
        self.sources={}
        self.detector_class=detector_class
        self.parameters=parameters
        self.parameters["resetcodes"]=[('reset',id)]
    '''
    This method accepts a data point with timestamp from a source and returns a prediction. For not seen sources create a new instant of detector.
    
    timestamp: the timestamp of the data
    data: an numpy array contain features (real numbers)
    id: the source that these data related to
    '''
    def collect_data(self,timestamp,data,id):
        if id in self.sources.keys():
            prediction=self.sources[id].get_Datapoint(timestamp,data, id)[0]
        else:
            self.sources[id]=self.detector_class(source=id,**self.parameters)
            prediction=self.sources[id].get_Datapoint(timestamp, data, id)[0]
        #prediction = PredictionPoint(anomaly_score, 0.5, alarm,self.thresholdtype,point.timestamp,point.source,notes=description,ensemble_details=(pairthresholds,pair_anomaly_scores))
        if prediction.score is None:
            notes=""
            for qqq in range(len(data)):
                notes+=",0,0"
            prediction = PredictionPoint(0, 0, False, "", timestamp,
                                         id, notes=notes,
                                                 ensemble_details="<>")
        return prediction

    '''
        This method trigger rest on the specific source (id) passed
        
        timestamp: the timestamp of the event
        desc: description of the event
        id: the source in which reset is triggered.
        '''
    def collect_event(self,timestamp,desc,id):
        ev =Eventpoint(code=desc,source=id,timestamp=timestamp)
        if id in self.sources.keys():
            self.sources[id].get_event(ev)
        else:
            self.sources[id]=self.detector_class(source=id,**self.parameters)
            self.sources[id].get_event(ev)

Methods

def collect_data(self, timestamp, data, id)

Expand source code

def collect_data(self,timestamp,data,id):
    if id in self.sources.keys():
        prediction=self.sources[id].get_Datapoint(timestamp,data, id)[0]
    else:
        self.sources[id]=self.detector_class(source=id,**self.parameters)
        prediction=self.sources[id].get_Datapoint(timestamp, data, id)[0]
    #prediction = PredictionPoint(anomaly_score, 0.5, alarm,self.thresholdtype,point.timestamp,point.source,notes=description,ensemble_details=(pairthresholds,pair_anomaly_scores))
    if prediction.score is None:
        notes=""
        for qqq in range(len(data)):
            notes+=",0,0"
        prediction = PredictionPoint(0, 0, False, "", timestamp,
                                     id, notes=notes,
                                             ensemble_details="<>")
    return prediction

def collect_event(self, timestamp, desc, id)

Expand source code

def collect_event(self,timestamp,desc,id):
    ev =Eventpoint(code=desc,source=id,timestamp=timestamp)
    if id in self.sources.keys():
        self.sources[id].get_event(ev)
    else:
        self.sources[id]=self.detector_class(source=id,**self.parameters)
        self.sources[id].get_event(ev)

class PdmModuleStatic (detector_class=PdmModule.Handlers.pairdetectionHandler.PairDectionHandler, parameters={'thresholdtype': 'selftunne', 'ProfileSize': 30, 'thresholdfactor': 15, 'ActualProfileSize': 0.25, 'sequencesize': 12, 'alarmsThreshold': 0, 'normz': False})

This class is used to provide an api for simulating streaming application of a technique. So given all data (raw and event timestamps that triger rest to the tecnique, its simulates the stream of data and returns predictions

The method predict is called to pass data and events for specific source and get all predictions

detector_class: A tecnique handler wich follow the get_Datapoint, get_event interface from Handlers

parameters : The parameters of the Handlers object to initialize

Expand source code

class PdmModuleStatic():
    '''
        This class is used to provide an api for simulating streaming application of a technique.
        So given all data (raw and event timestamps that triger rest to the tecnique, its simulates the stream of data and returns predictions

        The method predict is called to pass data and events for specific source and get all predictions

        detector_class: A tecnique handler wich follow the get_Datapoint, get_event interface from Handlers

        parameters : The parameters of the Handlers object to initialize
        '''
    def __init__(self,detector_class=PairDectionHandler,parameters={
        "thresholdtype" : "selftunne",
        "ProfileSize" : 30,
        "thresholdfactor" : 15,
        "ActualProfileSize" : 0.25,
        "sequencesize" : 12,
        "alarmsThreshold" : 0,
        "normz" : False
    }):
        self.detector_class = detector_class
        self.parameters = parameters
        self.parameters["resetcodes"] = [('reset', id)]

    '''
        This method simulate the stream using the raw and event data

        data: a DataFrame of the feature data (each row is a data point) where the index contain timestamps
        events: a DataFrame with events data, it needs to have one column 'dt' with timestamps of events and one 'desc' with a relative description
        id: the source of simulated data.
    '''
    def predict(self,data,events,id):

        if 'dt' not in events.columns:
            assert False, " A column named dt is expected in event dataframe"

        if 'desc' not in events.columns:
            assert False, " A column named dt is expected in event dataframe"

        handler = self.detector_class(source=id, **self.parameters)
        predictions = []
        dtlist = events['dt'].values
        di = 0
        ei = 0
        for i in range(len(data.index) + len(events.index)):
            if di >= len(data.index):
                break
            elif ei >= len(dtlist):
                prediction = handler.get_Datapoint(data.index[di], data.iloc[di].values, id)[0]
                if prediction.score is None:
                    notes = ""
                    for qqq in range(len(data)):
                        notes += ",0,0"
                    prediction = PredictionPoint(0, 0, False, "", data.index[di], id, notes=notes,
                                                 ensemble_details="<>")
                predictions.append(prediction)
                di += 1
            elif data.index[di] < dtlist[ei]:
                prediction = handler.get_Datapoint(data.index[di], data.iloc[di].values, id)[0]
                if prediction.score is None:
                    notes = ""
                    for qqq in range(len(data)):
                        notes += ",0,0"
                    prediction = PredictionPoint(0, 0, False, "", data.index[di], id, notes=notes,
                                                 ensemble_details="<>")
                predictions.append(prediction)
                di += 1
            else:
                ev = Eventpoint(code=events.iloc[ei]["desc"], source=id, timestamp=dtlist[ei])
                handler.get_event(ev)
                ei += 1
        return predictions

Methods

def predict(self, data, events, id)

Expand source code

def predict(self,data,events,id):

    if 'dt' not in events.columns:
        assert False, " A column named dt is expected in event dataframe"

    if 'desc' not in events.columns:
        assert False, " A column named dt is expected in event dataframe"

    handler = self.detector_class(source=id, **self.parameters)
    predictions = []
    dtlist = events['dt'].values
    di = 0
    ei = 0
    for i in range(len(data.index) + len(events.index)):
        if di >= len(data.index):
            break
        elif ei >= len(dtlist):
            prediction = handler.get_Datapoint(data.index[di], data.iloc[di].values, id)[0]
            if prediction.score is None:
                notes = ""
                for qqq in range(len(data)):
                    notes += ",0,0"
                prediction = PredictionPoint(0, 0, False, "", data.index[di], id, notes=notes,
                                             ensemble_details="<>")
            predictions.append(prediction)
            di += 1
        elif data.index[di] < dtlist[ei]:
            prediction = handler.get_Datapoint(data.index[di], data.iloc[di].values, id)[0]
            if prediction.score is None:
                notes = ""
                for qqq in range(len(data)):
                    notes += ",0,0"
                prediction = PredictionPoint(0, 0, False, "", data.index[di], id, notes=notes,
                                             ensemble_details="<>")
            predictions.append(prediction)
            di += 1
        else:
            ev = Eventpoint(code=events.iloc[ei]["desc"], source=id, timestamp=dtlist[ei])
            handler.get_event(ev)
            ei += 1
    return predictions