return json.dumps(result)
# params key checking
for key in ['tag', 'poi_label', 'seq', 'description']:
if key not in incoming_data:
logger.error("<%s>, [train gmm] [KeyError] params=%s, should have key: %s" % (x_request_id, incoming_data, key))
result['message'] = "Params content Error: can't find key=%s" % (key)
return json.dumps(result)
tag = incoming_data['tag']
label = incoming_data['poi_label']
seq = incoming_data['seq']
description = incoming_data['description']
algo_type = incoming_data.get('algo_type', 'gmm')
model = dao.get_model_by_tag_lable(algo_type, tag, label)
description += '\n last params: %s' % (model.get('params')) # store last params in description
_model = {
'nMix': model.get('nMix'),
'covarianceType': model.get('covarianceType'),
'nIter': model.get('nIter'),
'count': model.get('count'),
'params': model.get('params'),
}
my_trainer = Trainer(_model)
my_trainer.fit(seq)
dao.save_gmm('random_train', label, description, my_trainer.modelParams(), seq, model.get('count')+len(seq),
model.get('nIter'))
result = {'code': 0, 'message': 'success'}
logger.info('<%s>, [train gmm] success' % (x_request_id))
logger.info('<%s>, [train gmm data] last params: %s\t train data: %s\t current params: %s'
for key in ['tag', 'poi_label', 'seq', 'description']:
if key not in incoming_data:
logger.error(
"<%s>, [train gmm] [KeyError] params=%s, should have key: %s" %
(x_request_id, incoming_data, key))
result['message'] = "Params content Error: can't find key=%s" % (
key)
return json.dumps(result)
tag = incoming_data['tag']
label = incoming_data['poi_label']
seq = incoming_data['seq']
description = incoming_data['description']
algo_type = incoming_data.get('algo_type', 'gmm')
model = dao.get_model_by_tag_lable(algo_type, tag, label)
description += '\n last params: %s' % (
model.get('params')) # store last params in description
_model = {
'nMix': model.get('nMix'),
'covarianceType': model.get('covarianceType'),
'nIter': model.get('nIter'),
'count': model.get('count'),
'params': model.get('params'),
}
my_trainer = Trainer(_model)
my_trainer.fit(seq)
dao.save_gmm('random_train', label, description, my_trainer.modelParams(),
seq,
model.get('count') + len(seq), model.get('nIter'))
result = {'code': 0, 'message': 'success'}
def do_classify(incoming_data, x_request_id):
'''wrapper of classify
'''
result = {'code': 1, 'message': ''}
logger.debug('<%s>, [do_classify] enter, incoming_data=%s' % (x_request_id, incoming_data))
tag = incoming_data['tag']
seq = incoming_data['seq']
pois = incoming_data['pois']
algo_type = incoming_data.get('algo_type', 'gmm')
logger.info('<%s>, [classify gmm] params: tag=%s, seq=%s, algo_type=%s' %(x_request_id, tag, seq, algo_type))
# parse seq
seq = [util.parse_timestamp(e) for e in seq]
# classify
models = []
for poi in pois:
try:
model = dao.get_model_by_tag_lable(algo_type, tag, poi)
except IndexError:
logger.error('tag=%s, label=%s can not find model' % (tag, poi))
result['message'] = '<%s>, [do_classify] Can not find model whose tag=%s, label=%s ' % (x_request_id ,tag, poi)
raise ValueError(result['message'])
models.append(model)
_models = []
labels = []
for model in models:
labels.append(model.get('eventLabel'))
_model = {
'nMix': model.get('nMix'),
'covarianceType': model.get('covarianceType'),
'nIter': model.get('nIter'),
'count': model.get('count'),
'params': model.get('params'),
}
_models.append(_model)
my_predictor = Predictor(_models)
logger.debug('-------\n\n_models:\n %s\n' % (_models))
score_results = []
seq_scores = my_predictor.scores(seq)
logger.debug('seq_scores: %s' % (seq_scores))
for scores in seq_scores:
score_result = {}
for index, score in enumerate(scores):
if labels[index] in score_result:
score_result[labels[index]] += score
else:
score_result[labels[index]] = score
score_results.append(score_result)
logger.debug('!!!!!! score_result:%s' %(score_result))
# store seq in db
for index, timestamp in enumerate(seq):
event_label = max(score_results[index].iterkeys(), key=lambda key: score_results[index][key]) # max prob key is labelt
dao.save_train_data(timestamp, event_label)
logger.info('<%s> [classify] store timestamp=%s, label=%s to db success' % (x_request_id, timestamp, event_label))
logger.info('<%s> [classify gmm] success' % (x_request_id))
logger.debug('<%s> [classify gmm] result: %s' % (x_request_id, score_results))
result['code'] = 0
result['message'] = 'success'
result['result'] = score_results
return result
def do_classify(incoming_data, x_request_id):
'''wrapper of classify
'''
result = {'code': 1, 'message': ''}
logger.debug('<%s>, [do_classify] enter, incoming_data=%s' % (x_request_id, incoming_data))
tag = incoming_data['tag']
seq = incoming_data['seq']
pois = incoming_data['pois']
algo_type = incoming_data.get('algo_type', 'gmm')
logger.info('<%s>, [classify gmm] params: tag=%s, seq=%s, algo_type=%s' %(x_request_id, tag, seq, algo_type))
# parse seq
seq = [util.parse_timestamp(e) for e in seq]
# classify
models = []
for poi in pois:
try:
model = dao.get_model_by_tag_lable(algo_type, tag, poi)
except IndexError:
logger.error('tag=%s, label=%s can not find model' % (tag, poi))
result['message'] = '<%s>, [do_classify] Can not find model whose tag=%s, label=%s ' % (x_request_id ,tag, poi)
raise ValueError(result['message'])
models.append(model)
_models = []
labels = []
for model in models:
labels.append(model.get('eventLabel'))
_model = {
'nMix': model.get('nMix'),
'covarianceType': model.get('covarianceType'),
'nIter': model.get('nIter'),
'count': model.get('count'),
'params': model.get('params'),
}
_models.append(_model)
my_predictor = Predictor(_models)
logger.debug('-------\n\n_models:\n %s\n' % (_models))
score_results = []
seq_scores = my_predictor.scores(seq)
logger.debug('seq_scores: %s' % (seq_scores))
for scores in seq_scores:
score_result = {}
for index, score in enumerate(scores):
if labels[index] in score_result:
score_result[labels[index]] += score
else:
score_result[labels[index]] = score
score_results.append(score_result)
logger.debug('!!!!!! score_result:%s' %(score_result))
# store seq in db
for index, timestamp in enumerate(seq):
event_label = max(score_results[index].iterkeys(), key=lambda key: score_results[index][key]) # max prob key is labelt
dao.save_train_data(timestamp, event_label)
logger.info('<%s> [classify] store timestamp=%s, label=%s to db success' % (x_request_id, timestamp, event_label))
logger.info('<%s> [classify gmm] success' % (x_request_id))
logger.debug('<%s> [classify gmm] result: %s' % (x_request_id, score_results))
result['code'] = 0
result['message'] = 'success'
result['result'] = score_results
return result
