for s in sigma:
for a in alpha:
# Sharpen test data with various sigma (for Gaussian filter) and alpha value combinations
X_test_sharpen = process_data.sharpen(X, s, a)
pred_dyna_sharpen = model.predict(np.expand_dims(X_test_sharpen,
axis=2),
batch_size=32)
print ">>> sigma={}, alpha={:.2f}".format(s, a)
print accuracy_score(y, np.argmax(pred_dyna_sharpen, axis=1))
print confusion_matrix(y, np.argmax(pred_dyna_sharpen, axis=1))
# Load all test data (* dynamic and static data are mixed.)
X_test = process_data.load_x("test")
y_test = process_data.load_y("test")
# Set seed to ensure reproducibility of the paper.
seed = 818
# Static (4-sitting, 5-standing, 6-laying) test data are selected and
# split it in two, first & second, in order to determine
# sigma & alpha values for test data sharpening.
random.seed(seed)
stat_1 = np.where(y_test == 4)[0]
stat_1_first = random.sample(stat_1, int(len(stat_1) * 0.5))
stat_1_second = list(set(stat_1) - set(stat_1_first))
random.seed(seed)
'''
See paper: Sensors 2018, 18(4), 1055; https://doi.org/10.3390/s18041055
"Divide and Conquer-Based 1D CNN Human Activity Recognition Using Test Data Sharpening"
by Heeryon Cho & Sang Min Yoon
This code learns dynamic HAR model given in Figure 10.
(Sensors 2018, 18(4), 1055, page 13 of 24)
'''
# Load all train and test data (* dynamic and static data are mixed.)
X_train_all = process_data.load_x(
"train"
) # at this stage, the data includes both dynamic and static HAR data
y_train_all = process_data.load_y("train")
X_test_all = process_data.load_x("test")
y_test_all = process_data.load_y("test")
# --------------------------------------
# Only dynamic HAR data are selected
# --------------------------------------
# Select dynamic HAR train data
dynamic_1 = np.where(y_train_all == 1)[0]
dynamic_2 = np.where(y_train_all == 2)[0]
dynamic_3 = np.where(y_train_all == 3)[0]
dynamic = np.concatenate([dynamic_1, dynamic_2, dynamic_3])
dynamic_list = dynamic.tolist()
import process_data
from sklearn.linear_model import LogisticRegression
from sklearn.svm import SVC
from sklearn.neural_network import MLPClassifier
from sklearn.tree import DecisionTreeClassifier
from sklearn.metrics import accuracy_score
from sklearn.metrics import confusion_matrix
# Load all train and test data (* dynamic and static data are mixed.)
X_train = process_data.load_x("train")
y_train = process_data.load_y("train")
X_test = process_data.load_x("test")
y_test = process_data.load_y("test")
print "=================================="
print " ACCURACY OF OTHER ML CLASSIFIERS"
print "=================================="
# Build a logistic regression classifier and predict
clf_lr = LogisticRegression(random_state=0)
clf_lr.fit(X_train, y_train)
pred_lr = clf_lr.predict(X_test)
print "\n--- Logistic Regression Classifier ---"
print accuracy_score(y_test, pred_lr)
print confusion_matrix(y_test, pred_lr)