def score(self, X, y):
scores = outlier.SWLOF(**self.params).fit_predict(X[:,:-1])
if np.any(np.isnan(scores)):
print ('Warning: marking NaN as non-outlier')
scores[np.isnan(scores)] = 0
return compute_training_score(y, transform_scores(scores))
wrapper = wLODA()
best_param = parameter_tuning(param, np.hstack((data[0:training_len,:],timestamps[0:training_len,:])), labels[0:training_len], wrapper)
else:
print("ERROR: Invalid algorithm! \n")
quit()
print("--------------")
for idx in range(nrep):
if alg=='sdo':
detector = pysdo.SDO(k=best_param['k'], x=best_param['x'], qv=best_param['qv'], random_state=idx, chunksize=blocksize+1, return_scores = True)
elif alg=='sdostream':
detector = outlier.SDOstream(k=best_param['k'], T=timepam, x=best_param['x'], qv=best_param['qv'], seed=idx)
elif alg=='swknn':
detector = outlier.SWKNN(k=best_param['k'], window=timepam)
elif alg=='swlof':
detector = outlier.SWLOF(k=best_param['k'], window=timepam)
elif alg=='rrct':
detector = outlier.SWRRCT(n_estimators=best_param['n_estimators'], window=timepam, seed=idx, n_jobs=1)
elif alg=='xstream':
detector = xStream(k=best_param['k'], c=best_param['c'], d=best_param['d'], init_sample=timepam, seed=idx)
elif alg=='rshash':
detector = outlier.RSHash(n_estimators=best_param['n_estimators'], cms_w=best_param['cms_w'], cms_d=best_param['cms_d'], window=timepam, seed=idx, n_jobs=1)
elif alg=='loda':
detector = outlier.LODA(n_projections=best_param['n_projections'], n_bins=best_param['n_bins'], window=timepam)
print("Run ",idx , " of ", nrep)
scores = np.zeros(data.shape[0])
### TRAINING
start_tr = time.time()
scores[:training_len] = detector.fit_predict(data[:training_len,:])
end_tr = time.time()
