from ay_hw_3.util_data import load_data_and_label, get_all_datasets_path
from ay_hw_3.util_statistic import gen_statistic_result
from ay_hw_3._global import ROOT_PATH
import pandas as pd
import numpy as np
import pprint
if __name__ == "__main__":
allFilePaths = get_all_datasets_path(rootPath=ROOT_PATH)
statisticResult = pd.DataFrame()
for index, path in enumerate(allFilePaths):
fileItem, fileLabel = load_data_and_label(path)
staticResultItem = gen_statistic_result(fileItem, index + 1)
statisticResult = statisticResult.append(staticResultItem)
##----Same to the main_c_ii.py-------------------
confidence_interval = {}
for column in statisticResult.columns:
itemCIRange = []
for i in range(0, 999):
# Return a random sample of items from an axis of object.
ran_sample = statisticResult[column].sample(n=10, replace=True)
stat = ran_sample.std()
itemCIRange.append(stat)
itemCIRange.sort()
lowerValue = np.percentile(itemCIRange, 0.05)
upperValue = np.percentile(itemCIRange, 0.95)
confidence_interval[column] = [lowerValue, upperValue]
trainStaticResult = pd.DataFrame()
testStaticResult = pd.DataFrame()
gaussianTestErrorRateList = list()
for parts in range(1, 10):
for index, path in enumerate(allTrainFilePaths):
trainFileItem, trainFileLabel = load_data_and_label(path,
hasTime=False)
gluedTrainFile = split_DF_in_parts(trainFileItem,
parts=parts,
needConcat=True)
gluedTrainFile.columns = gen_multiple_column_name(parts=parts,
hasTime=False)
trainStaticResultItem = gen_statistic_result(gluedTrainFile,
index + 1,
hasTime=False)
trainStaticResultItem["label"] = convert_label_2_num(
trainFileLabel)
trainStaticResult = trainStaticResult.append(trainStaticResultItem,
sort=False)
for index, path in enumerate(allTestFilePaths):
testFileItem, testFileLabel = load_data_and_label(path,
hasTime=False)
gluedTestFile = split_DF_in_parts(testFileItem,
parts=parts,
needConcat=True)
gluedTestFile.columns = gen_multiple_column_name(parts=parts,
hasTime=False)
testStaticResultItem = gen_statistic_result(gluedTestFile,
if __name__ == "__main__":
if not sys.warnoptions:
warnings.simplefilter("ignore")
allTrainFilePaths = gen_train_data_file_paths()
# based on what the pdf said, we need to use all training data
trainStaticResult = pd.DataFrame()
for parts in range(1, 21):
for index, path in enumerate(allTrainFilePaths):
fileItem, fileLabel = load_data_and_label(path, hasTime=False)
splitedDFs = split_DF_in_parts(fileItem, parts=parts, needConcat=False)
statisticResultTemp = pd.DataFrame()
for DFItem in splitedDFs:
staticResultTempItem = gen_statistic_result(DFItem, index + 1, hasTime=False)
statisticResultTemp = statisticResultTemp.append(staticResultTempItem, sort=False)
statisticResultTemp["label"] = is_bending(fileLabel)
trainStaticResult = trainStaticResult.append(statisticResultTemp, sort=False)
logitModel = sm.Logit(trainStaticResult['label'],
trainStaticResult[gen_multiple_label(parts=1)])
logitModelResults = logitModel.fit(method="bfgs",disp=0)
# ['median(1)'] ['max(5)']
significantVars = \
[key for key, p_value in logitModelResults.pvalues.items() if p_value <= 0.05]
if len(significantVars) > 0:
print("When split all training data sets in {} times, "
"I got significant variables : ".format(parts), end=" ")
print(' '.join(significantVars))
if __name__ == "__main__":
simplefilter(action='ignore', category=FutureWarning)
# the best l I got is 3
bestL = 3
selectedFeatures = ['min(5)', '3rd quart(5)', '3rd quart(7)', 'max(18)']
allTrainFilePaths = gen_train_data_file_paths()
# based on what the pdf said, we need to use all training data
statisticResult = pd.DataFrame()
for index, path in enumerate(allTrainFilePaths):
fileItem, fileLabel = load_data_and_label(path, hasTime=False)
gluedFile = split_DF_in_parts(fileItem, parts=bestL, needConcat=True)
gluedFile.columns = gen_multiple_column_name(parts=bestL,
hasTime=False)
staticResultItem = gen_statistic_result(gluedFile,
index + 1,
hasTime=False)
staticResultItem["label"] = is_bending(fileLabel)
statisticResult = statisticResult.append(staticResultItem, sort=False)
X_trainData = statisticResult[selectedFeatures]
y_trainData = statisticResult['label']
skLogitModel = LogisticRegression()
skLogitModel.fit(X_trainData, y_trainData)
skYPredict = skLogitModel.predict(X_trainData)
falsePositiveRate, truePositiveRate, thresholds = roc_curve(
y_trainData, skYPredict)
# compute Area Under the Curve (AUC) using the trapezoidal rule
area = auc(falsePositiveRate, truePositiveRate)
#
__author__ = 'Aaron Yang'
__email__ = '*****@*****.**'
__date__ = '9/27/2019 11:13 PM'
from ay_hw_3.util_data import convert_label_2_num, load_data_and_label
from ay_hw_3.util_generate import gen_test_data_file_paths
from ay_hw_3.util_statistic import gen_statistic_result
from ay_hw_3._global import FULL_COLUMNS, ROOT_PATH
if __name__ == "__main__":
print(convert_label_2_num("bending1"))
print(convert_label_2_num("bending2"))
print(convert_label_2_num("cycling"))
print(convert_label_2_num("sitting"))
print(convert_label_2_num("walking"))
print(type(convert_label_2_num("lying")))
# print(gen_test_data_file_paths('.\\assets'))
dataframe, label = load_data_and_label('.\\assets\\cycling\\dataset2.csv')
staticResultItem = gen_statistic_result(dataframe, 1)
print(staticResultItem.to_string())
