def __init__(self):
self._header = None
self._root = None
self._grace_period = 200
self._split_confidence = 0.0000001
self._hoeffding_tie_threshold = 0.05
self._min_frac_weight_for_two_branches_gain = 0.01
# Split metric stuff goes here
self.GINI_SPLIT = 0
self.INFO_GAIN_SPLIT = 1
self._selected_split_metric = self.INFO_GAIN_SPLIT
self._split_metric = InfoGainSplitMetric(
self._min_frac_weight_for_two_branches_gain)
# self._selected_split_metric = self.GINI_SPLIT
# self._split_metric = GiniSplitMetric()
# Leaf prediction strategy stuff goes here
# Only used when the leaf prediction strategy is baded on Naive Bayes, not useful right now
#self._nb_threshold = 0
self._active_leaf_count = 0
self._inactive_leaf_count = 0
self._decision_node_count = 0
# Print out leaf models in the case of naive Bayes or naive Bayes adaptive leaves
self._print_leaf_models = False
def __init__(self):
self._header = None
self._root = None
self._grace_period = 200
self._split_confidence = 0.0000001
self._hoeffding_tie_threshold = 0.05
self._min_frac_weight_for_two_branches_gain = 0.01
# Split metric stuff goes here
self.GINI_SPLIT = 0
self.INFO_GAIN_SPLIT = 1
self._selected_split_metric = self.INFO_GAIN_SPLIT
self._split_metric = InfoGainSplitMetric(
self._min_frac_weight_for_two_branches_gain)
#self._selected_split_metric = self.GINI_SPLIT
#self._split_metric = GiniSplitMetric()
# Leaf prediction strategy stuff goes here
# Only used when the leaf prediction strategy is baded on Naive Bayes, not useful right now
#self._nb_threshold = 0
self._active_leaf_count = 0
self._inactive_leaf_count = 0
self._decision_node_count = 0
# Print out leaf models in the case of naive Bayes or naive Bayes adaptive leaves
self._print_leaf_models = False
logging.basicConfig(
level=logging.DEBUG,
format=
'%(asctime)s %(filename)s[line:%(lineno)d] %(levelname)s %(message)s',
#filename='../log/tmp.log',
datefmt='%a, %d %b %Y %H:%M:%S',
filemode='w')
self.logger = logging.getLogger("vfdt") #self.name
# 定义一个FileHandler，将INFO级别或更高的日志信息记录到log文件，并将其添加到当前的日志处理对象#
# fh = logging.FileHandler('../log/' + "vfdt" + '.log', mode='w')# self.name
# formatter = logging.Formatter('%(asctime)s %(filename)s[line:%(lineno)d] %(levelname)s %(message)s')
# fh.setFormatter(formatter)
# fh.setLevel(logging.INFO)
# self.logger.addHandler(fh)
# 定义一个StreamHandler，将INFO级别或更高的日志信息打印到标准错误，并将其添加到当前的日志处理对象#
console = logging.StreamHandler()
console.setLevel(logging.INFO)
formatter = logging.Formatter(
'%(message)s') # ('%(name)-12s: %(levelname)-8s )
console.setFormatter(formatter)
self.logger.addHandler(console)
def build_classifier(self, dataset, testdataset=None):
"""Build the classifier.
Args:
dataset (Dataset): The data to start training the classifier.
"""
#日志部分
self.logger.info(
"Start build classifier--------------------------\n"
"grace_period = {},\n"
"split_confidence = {},\n"
"hoeffding_tie_threshold = {},\n"
"min_frac_weight_for_two_branches_gain = {}\n"
"--------------------------------------------------".format(
self._grace_period, self._split_confidence,
self._hoeffding_tie_threshold,
self._min_frac_weight_for_two_branches_gain))
#训练部分
self.reset()
self._header = dataset
if self._selected_split_metric is self.GINI_SPLIT:
self._split_metric = GiniSplitMetric()
else:
self._split_metric = InfoGainSplitMetric(
self._min_frac_weight_for_two_branches_gain)
count = 0
test_epoch = 100
test_acc_list = []
for i in range(dataset.num_instances()):
self.update_classifier(dataset.instance(i))
count += 1
if count % test_epoch == 0 and testdataset != None:
acc = self.valuate_acc(testdataset)
test_acc_list.append(acc)
#画图部分
if testdataset != None:
x_axis = range(len(test_acc_list))
plt.plot(
x_axis, test_acc_list,
label='test_acc') # Plot some data on the (implicit) axes.
plt.xlabel('iter')
plt.ylabel('acc')
figpath = './respic/' + 'acc.jpg'
plt.title(figpath.split('/')[-1])
plt.legend()
plt.savefig(figpath)
plt.close('all')
def build_classifier(self, dataset):
"""Build the classifier.
Args:
dataset (Dataset): The data to start training the classifier.
"""
self.reset()
self._header = dataset
if self._selected_split_metric is self.GINI_SPLIT:
self._split_metric = GiniSplitMetric()
else:
self._split_metric = InfoGainSplitMetric(
self._min_frac_weight_for_two_branches_gain)
for i in range(dataset.num_instances()):
self.update_classifier(dataset.instance(i))
