def checkPredict(testX, testY):
currentDataPoint = 0
driftCount = 0
totalPredictions = 0
correctPredictions = 0
currentError = 0
drift = False
#Each model makes their own predicition
testPredict1 = model1.predict(testX)
testPredict2 = model2.predict(testX)
testPredict3 = model3.predict(testX)
testPredict4 = model4.predict(testX)
for i in range(len(testY)):
#final prediction is the average of all the models predictions
prediction = (testPredict1[i] + testPredict2[i] + testPredict3[i] +
testPredict4[i]) / 4
if (prediction >= (testY[currentDataPoint] * 0.98)):
if (prediction <= (testY[currentDataPoint] * 1.02)):
#if the prediction is correct tell the EDDM this
drift = EDDM.set_input(True)
correctPredictions = correctPredictions + 1
else:
#if it prediction is wrong tell the EDDM this
drift = EDDM.set_input(False)
else:
#if it prediction is wrong tell the EDDM this
drift = EDDM.set_input(False)
#If drift update counters
if (drift):
driftCount += 1
currentError += abs(
(prediction - testY[currentDataPoint]) / testY[currentDataPoint])
currentDataPoint += 1
totalPredictions += 1
return totalPredictions, correctPredictions, driftCount, currentError
def checkPredict(testX, testY): currentDataPoint= 0 driftCount = 0 totalPredictions = 0 correctPredictions = 0 currentError = 0 #create test predictions testPredict = model.predict(testX) #loop through all the test predicitions and compare to actual results for y in testPredict: if(y >= (testY[currentDataPoint]*0.98)): if(y <= (testY[currentDataPoint]*1.02)): #if the prediction is correct tell the EDDM this drift = EDDM.set_input(True) correctPredictions += 1 else: #if it prediction is wrong tell the EDDM this drift = EDDM.set_input(False) else: #if it prediction is wrong tell the EDDM this drift = EDDM.set_input(False) if(drift): driftCount += 1 currentError += abs((y-testY[currentDataPoint])/testY[currentDataPoint]) currentDataPoint += 1 totalPredictions += 1 return totalPredictions, correctPredictions, driftCount, currentError
def checkPredict(startPoint, crct):
currentDataPoint = startPoint
driftCount = 0
totalPredictions = 0
correctPredictions = 0
currentError = 0
#create test predictions
testPredict = model.predict(testX)
drift = False
driftPoint = 0
adjust = False
currentCorrect = crct
#loop through all the test predicitions and compare to actual results
for i in range(currentDataPoint, len(testPredict)):
if (testPredict[currentDataPoint] >= (testY[currentDataPoint] * 0.98)):
if (testPredict[currentDataPoint] <=
(testY[currentDataPoint] * 1.02)):
#if the prediction is correct tell the EDDM this
drift = EDDM.set_input(True)
correctPredictions += 1
currentCorrect += 1
else:
#if it prediction is wrong tell the EDDM this
drift = EDDM.set_input(False)
else:
#if the prediction is wrong tell the EDDM this
drift = EDDM.set_input(False)
#if Drift is detected and window size isn't 2 stop checking and change window size and retrain
if (drift):
driftCount += 1
driftPoint = currentDataPoint + 1
currentCorrect = 0
if (look_back > 2):
currentError += abs(
(testPredict[currentDataPoint] - testY[currentDataPoint]) /
testY[currentDataPoint])
totalPredictions += 1
return totalPredictions, correctPredictions, driftCount, currentCorrect, drift, adjust, currentError
#if 30 correct predictions sctop prediciting and change window size back to original
if (currentCorrect > 30):
currentCorrect = 0
if (look_back != 10):
currentError += abs(
(testPredict[currentDataPoint] - testY[currentDataPoint]) /
testY[currentDataPoint])
totalPredictions += 1
driftPoint = currentDataPoint + 1
adjust = True
return totalPredictions, correctPredictions, driftCount, currentCorrect, drift, adjust, currentError
currentError += abs(
(testPredict[currentDataPoint] - testY[currentDataPoint]) /
testY[currentDataPoint])
currentDataPoint += 1
totalPredictions += 1
#Or return total amount of predictions made, correct predicitions made, amount of drifts detected, how many are currently correct, if there is a drift, if the window needs to be adjusted, and the current percent error
return totalPredictions, correctPredictions, driftCount, currentCorrect, drift, adjust, currentError
global totalTrainRMSE
global totalTestRMSE
totalTestRMSE = 0
totalTrainRMSE = 0
os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"
os.environ['CUDA_VISIBLE_DEVICES'] = '-1'
# fix random seed for reproducibility
numpy.random.seed(7)
# load the dataset, Change DATASET.csv to the file you want to load, change usecols=[0] to correct coloumn for the dataset
dataframe = pandas.read_csv('DATASET.csv',
usecols=[0],
engine='python',
skipfooter=3)
dataset = dataframe.values
dataset = dataset.astype('float32')
EDDM = EDDM()
driftCount = 0
drift = False
# convert an array of values into a dataset matrix
def create_dataset(dataset, end, start):
dataX, dataY = [], []
for i in range(start, end - look_back):
dataX.append(dataset[i:(i + look_back), 0])
for i in range(start + look_back, end):
dataY.append(dataset[i, 0])
return numpy.array(dataX), numpy.array(dataY)
#create datasets