def __init__(self, model: Model, clf):
self.model = model
self.get_accuracy = Get_Accuracy()
self.processing = Processing_DB_Files()
self.clf = clf
super().__init__()
#===INITIALIZATION===#
Debug.DEBUG = 0
arcma = ARCMA_Model()
processing = Processing_DB_Files()
project = Project()
#tuple from MPL
t_aux = []
for i in range(0,500):
t_aux.append(500)
t = tuple(t_aux)
####
classifiers = {"MPL": MLPClassifier(random_state=1, solver="adam", activation="relu", max_iter=100000, alpha=1e-5, hidden_layer_sizes=t), "Extratrees": ExtraTreesClassifier(n_estimators = 1000, random_state=1), "Knn":KNeighborsClassifier(n_neighbors=5), "Naive Bayes":GaussianNB(), "RandomForest":RandomForestClassifier(n_estimators = 1000, random_state=1), "Decision Tree":tree.DecisionTreeClassifier(random_state=1), "SVM":svm.SVC(probability=True, random_state=1)}
persons = [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]
get_accuracy = Get_Accuracy()
balance_data = BalanceData()
threshold_balance_data = 40
#Select the best classifier
accuracy_mean = pd.DataFrame(columns=["Classifier", "Accuracy"])
project.log("=====================ARCMA_SELECT_BEST_ALGORITHM=====================", file="arcma_best_algorithm.log")
for c in classifiers:
print(c)
person_accuracies = []
person_f_score = []
person_precision = []
person_recall = []
times_to_predict = []
for p in persons:
s = save()
try:
import pandas as pd
from pre_processing.processing_db_files import Processing_DB_Files
from utils.project import Project, slash
from scripts.save_workspace import save
from pre_processing.get_accuracy import Get_Accuracy
from sklearn.model_selection import StratifiedKFold
from pre_processing.balance_data import BalanceData
import numpy as np
#===INITIALIZATION===#
Debug.DEBUG = 0
hmp = HMP_Model()
processing = Processing_DB_Files()
project = Project()
extra_trees = ExtraTreesClassifier(n_estimators=1000, random_state=0)
get_accuracy = Get_Accuracy()
balance_data = BalanceData()
threshold_balance_data = 40
#===LOAD FEATURES===#
#Interate threshold to find de best value#
persons = [
"f1", "m1", "m2", "f2", "m3", "f3", "m4", "m5", "m6", "m7", "f4", "m8",
"m9", "f5", "m10", "m11"
]
accuracy_by_person = pd.DataFrame()
threshold = 0.65
project.log(
"=========== HMP StratifiedKFold Accuracy, Thresold = {}===========".
format(threshold),
hidden_layer_sizes=t),
"Extratrees":
ExtraTreesClassifier(n_estimators=1000, random_state=1),
"Knn":
KNeighborsClassifier(n_neighbors=5),
"Naive Bayes":
GaussianNB(),
"RandomForest":
RandomForestClassifier(n_estimators=1000, random_state=1),
"Decision Tree":
tree.DecisionTreeClassifier(random_state=1),
"SVM":
svm.SVC(probability=True, random_state=1)
}
get_accuracy = Get_Accuracy()
balance_data = BalanceData()
threshold_balance_data = 40
#Select the best classifier
for model in models:
print(model['model_name'])
persons = model['persons']
accuracy_mean = pd.DataFrame(columns=[
"Classifier", "Accuracy", "Precision", "Recall", "F-Score", "Time"
])
print("====================={}_SELECT_BEST_ALGORITHM=====================".
format(model['model_name']))
for c in classifiers:
print(c)
person_accuracies = []
person_f_score = []
from sklearn.ensemble import ExtraTreesClassifier # Extra Trees
from models.hmp_model import HMP_Model
import pandas as pd
from sklearn.model_selection import train_test_split
from pre_processing.processing_db_files import Processing_DB_Files
from utils.project import Project
from scripts.save_workspace import save
from pre_processing.get_accuracy import Get_Accuracy
#===INITIALIZATION===#
Debug.DEBUG = 0
hmp = HMP_Model()
processing = Processing_DB_Files()
project = Project()
extra_trees = ExtraTreesClassifier(n_estimators = 10000, random_state=0)
get_accuracy = Get_Accuracy()
#===LOAD FEATURES===#
#Interate threshold to find de best value#
s = save()
relevant_features = s.load_var("hmp_relevant_features\\relevant_features_f1.pkl")
y = s.load_var("hmp_relevant_features\\y_f1.pkl")
y = pd.DataFrame(y, columns=[hmp.label_tag])
X_train, X_test, y_train, y_test = train_test_split(relevant_features, y, test_size=0.2, random_state=42)
data = {}
data["f1"] = {}
data["f1"]["training"] = {}
data["f1"]["training"]["training_features"] = X_train
data["f1"]["training"]["training_labels"] = y_train
class Base_Classification(object):
def __init__(self, model: Model, clf):
self.model = model
self.get_accuracy = Get_Accuracy()
self.processing = Processing_DB_Files()
self.clf = clf
super().__init__()
#Finding the best window to improve the accuracy
def find_best_window(self, interval, list_people):
accuracies = []
for w in interval:
Debug.print_debug("WINDOWS SIZE: {}".format(w))
data_all_people = self.model.load_training_data_from_list_people(
w, list_people)
features_all_people = self.processing.calculating_features_to_each_person(
data_all_people, self.model)
accuracies_dict = self.get_accuracy.simple_accuracy_mean_to_each_person(
features_all_people, self.model, self.clf)
accuracy_mean = mean(accuracies_dict.values())
accuracies.append({'window': w, 'accuracy': accuracy_mean})
return accuracies
def predict_for_all_people_with_proba(self, window, threshold):
data_all_people = self.model.load_training_data_from_all_people(window)
features_all_people = self.processing.calculating_features_to_each_person(
data_all_people, self.model)
accuracies_proba = self.get_accuracy.simple_accuracy_mean_to_each_person_with_proba(
features_all_people, self.model, self.clf, threshold)
accuracies = self.get_accuracy.simple_accuracy_mean_to_each_person(
features_all_people, self.model, self.clf)
return accuracies, accuracies_proba
def predict_for_list_people_with_proba(self, window, list_people,
threshold):
data_list_people = self.model.load_training_data_from_list_people(
window, list_people)
features_all_people = self.processing.calculating_features_to_each_person(
data_list_people, self.model)
accuracies_proba = self.get_accuracy.simple_accuracy_mean_to_each_person_with_proba(
features_all_people, self.model, self.clf, threshold)
accuracies = self.get_accuracy.simple_accuracy_mean_to_each_person(
features_all_people, self.model, self.clf)
return accuracies, accuracies_proba
def predict_outliers_for_list_people_with_proba(self,
window,
list_people,
activity,
threshold,
remove_outliers=0):
data_list_people = self.model.load_training_data_from_list_people(
window, list_people, remove_outliers)
features_all_people = self.processing.calculating_features_to_each_person(
data_list_people, self.model)
#return self.get_accuracy.simple_accuracy_outlier_activity(features_all_people, self.model, self.clf, activity,threshold)
return self.get_accuracy.get_outliers_confused_with_activities(
features_all_people, self.model, self.clf, threshold)
training, training_labels, test, test_labels, outlier, outlier_labels = self.get_accuracy.simple_accuracy_outlier_activity(
features_all_people, self.model, self.clf, activity, threshold)
from sklearn.ensemble import ExtraTreesClassifier # Extra Trees
import pandas as pd
from sklearn.model_selection import train_test_split
from pre_processing.get_accuracy import Get_Accuracy
import numpy as np
from tsfresh import extract_relevant_features
import time
from pre_processing.balance_data import BalanceData
#===INITIALIZATION===#
Debug.DEBUG = 0
arcma = ARCMA_Model()
processing = Processing_DB_Files()
project = Project()
persons = [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]
get_accuracy = Get_Accuracy()
balance_data = BalanceData()
threshold_balance_data = 40
#Select de best windows
t = time.time()
best_model = ExtraTreesClassifier(n_estimators = 1000, random_state=0)
w_accuracies = pd.DataFrame(columns=["window", "accurary"])
p = 15 # pessoa com mais registros
project.log("=====================ARCMA_SELECT_BEST_WINDOWS=====================", file="arcma_log_best_window.log")
for w in range(20,110,10):
print("Load data with window len = {}".format(w))
data = arcma.load_training_data_by_people(p)
print("Slicing Window....")
data_tsfresh, y = arcma.slice_by_window_tsfresh(data, w)
y.index += 1
