In [1]:
import pandas as pd
import numpy as np
import os
from glob import glob
from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import train_test_split
In [29]:
from ngboost import NGBClassifier
from ngboost.distns import Bernoulli,k_categorical
from ngboost.learners import default_tree_learner
In [2]:
from sklearn.metrics import classification_report
from display_results import confusion_matrix
In [3]:
df_activity = pd.read_csv("activities.csv")
In [4]:
def segmentation(x_data,y,overlap_rate,time_window):
seg_data = []
overlap = int((1 - overlap_rate)*time_window)
y_segmented_list = []
for i in range(0,x_data.shape[0],overlap):
seg_data.append(x_data[i:i+time_window])
y_segmented_list.append(y)
return seg_data,y_segmented_list
In [5]:
def handle_missing_values(df):
df['x']=df['x'].replace(0, np.nan)
df['y']=df['y'].replace(0, np.nan)
df['z']=df['z'].replace(0, np.nan)
return df
In [47]:
change_act_id = {2:0, 3:1, 4:2, 6:3, 9:4, 12:5}
In [45]:
def get_act_id(seg_id):
seg = df_activity[df_activity["segment_id"]==seg_id]
activity_id = seg["activity_id"].values
return change_act_id[int(activity_id)]
In [46]:
def load_data(csv_file):
y_list = []
x_data_list = []
csv_df = pd.read_csv(csv_file)
csv_df = handle_missing_values(csv_df)
csv_df.dropna(inplace=True)
x_data = csv_df.values
act_id = get_act_id(int(os.path.splitext(os.path.basename(csv_file))[0].replace("segment","")))
return x_data,act_id
In [48]:
def get_features(x_data):
features = []
for i in range(x_data.shape[1]):
# std
features.append(x_data.T[i].std(ddof=0))
# avg
features.append(np.average(x_data.T[i]))
# max
features.append(np.max(x_data.T[i]))
# min
features.append(np.min(x_data.T[i]))
return features
In [42]:
csv_files = glob("train/*")
In [43]:
X_feature_data_list = []
y_list = []
for csv_file in csv_files:
x,y = load_data(csv_file)
X_feature_data_list.append(get_features(x))
y_list.append(y)
In [51]:
X_train,X_test,y_train,y_test = train_test_split(X_feature_data_list,y_list,test_size=0.3)
In [52]:
model_ml = RandomForestClassifier(n_estimators=500,n_jobs=-1)
In [53]:
model_ml.fit(X_train,y_train)
y_predict = model_ml.predict(X_test)
In [54]:
print(classification_report(y_test,y_predict))
precision recall f1-score support
0 0.42 0.83 0.56 6
1 1.00 0.67 0.80 15
2 0.67 0.29 0.40 7
3 0.50 0.33 0.40 6
4 0.43 0.60 0.50 5
5 0.67 0.80 0.73 15
accuracy 0.63 54
macro avg 0.61 0.59 0.56 54
weighted avg 0.69 0.63 0.63 54
In [55]:
print(confusion_matrix(y_test, y_predict))
None
In [58]:
ngb = NGBClassifier(Dist = k_categorical(6))
ngb.fit(X_train, y_train)
[iter 0] loss=1.7464 val_loss=0.0000 scale=0.5000 norm=3.6180 [iter 100] loss=0.5734 val_loss=0.0000 scale=2.0000 norm=4.4797 [iter 200] loss=0.2855 val_loss=0.0000 scale=4.0000 norm=6.6994 [iter 300] loss=0.2065 val_loss=0.0000 scale=2.0000 norm=3.0851 [iter 400] loss=0.1836 val_loss=0.0000 scale=1.0000 norm=1.5061
Out[58]:
NGBClassifier(Base=DecisionTreeRegressor(ccp_alpha=0.0,
criterion='friedman_mse', max_depth=3,
max_features=None, max_leaf_nodes=None,
min_impurity_decrease=0.0,
min_impurity_split=None,
min_samples_leaf=1,
min_samples_split=2,
min_weight_fraction_leaf=0.0,
presort='deprecated',
random_state=None, splitter='best'),
Dist=<class 'ngboost.distns.categorical.k_categorical.<locals>.Categorical'>,
Score=<class 'ngboost.scores.LogScore'>, col_sample=1.0,
learning_rate=0.01, minibatch_frac=1.0, n_estimators=500,
natural_gradient=True,
random_state=RandomState(MT19937) at 0x110FAFDB0, tol=0.0001,
verbose=True, verbose_eval=100)
In [59]:
y_predict_ngboost =ngb.predict(X_test)
In [60]:
print(classification_report(y_predict_ngboost,y_test))
precision recall f1-score support
0 0.50 0.75 0.60 4
1 0.67 0.62 0.65 16
2 0.29 0.29 0.29 7
3 0.17 0.33 0.22 3
4 0.40 0.40 0.40 5
5 0.73 0.58 0.65 19
accuracy 0.54 54
macro avg 0.46 0.50 0.47 54
weighted avg 0.58 0.54 0.55 54
In [61]:
confusion_matrix(y_predict_ngboost,y_test)
