def run(fold):
# reading the data
df = pd.read_csv(config.train_data_folds)
num_cols = [
'age', 'fnlwgt', 'capital.gain', 'capital.loss', 'hours.per.week'
]
target_map = {'<=50K': 0, '>50K': 1}
df.income = df.income.map(target_map)
# cat cols
cat_features = [
f for f in df.columns if f not in num_cols + ['income', 'kfold']
]
# feature engineering
df = utils.feature_engineering(df, cat_features)
# selecting the features
features = [f for f in df.columns if f not in ['income', 'kfold']]
# treating NANS
for col in features:
if col not in num_cols:
df.loc[:, col] = df[col].astype(str).fillna('NONE')
# label encoding
for feat in features:
if col not in num_cols:
lbl_enc = preprocessing.LabelEncoder()
lbl_enc.fit(df[feat])
df.loc[:, feat] = lbl_enc.transform(df[feat])
# splitting the data based on the folds created
df_train = df[df.kfold != fold].reset_index(drop=True)
df_valid = df[df.kfold == fold].reset_index(drop=True)
x_train = df_train[features].values
x_valid = df_valid[features].values
# xgb
model = xgb.XGBClassifier(n_jobs=-1, max_depth=7, n_estimators=200)
model.fit(x_train, df_train.income.values)
# AUC
# taking the probability of 1
valid_pred = model.predict_proba(x_valid)[:, 1]
auc = metrics.roc_auc_score(df_valid.income.values, valid_pred)
print('Fold: ', fold, 'Validation AUC: ', auc)
def NB_model():
train_data, test_data = utils.prepare_data()
print('训练集有{}条记录。'.format(len(train_data)))
print('测试集有{}条记录。'.format(len(test_data)))
X_train, X_test = utils.feature_engineering(train_data, test_data)
print('共有{}维特征。'.format(X_train.shape[1]))
y_train = train_data['label'].values
y_test = test_data['label'].values
#数据建模
nb_model = GaussianNB()
nb_model.fit(X_train, y_train)
y_pred = nb_model.predict(X_test)
print('准确率：', accuracy_score(y_test, y_pred))
def SVC_model():
train_data, test_data = utils.prepare_data()
print('训练集有{}条记录。'.format(len(train_data)))
print('测试集有{}条记录。'.format(len(test_data)))
X_train, X_test = utils.feature_engineering(train_data, test_data)
print('共有{}维特征。'.format(X_train.shape[1]))
y_train = train_data['label'].values
y_test = test_data['label'].values
#数据建模
c_values = [0.0001, 1, 10000]
for c_value in c_values:
svm_model = SVC(C=c_value)
svm_model.fit(X_train, y_train)
y_pred = svm_model.predict(X_test)
print('准确率：', accuracy_score(y_test, y_pred))
data_chunk = data_chunk[data_chunk.object_id != arr[len(arr) - 1]]
data_chunk = data_chunk.reset_index(drop=True)
meta_chunk = test_meta[test_meta['object_id'].isin(
data_chunk['object_id'].unique())]
meta_chunk = meta_chunk.reset_index(drop=True)
g_data, eg_data, g_meta, eg_meta = utils.gal_split_data(
data_chunk, meta_chunk, False)
g_features = None
eg_features = None
if g_meta.shape[0] > 0:
#make meta not drop object_id in the feature engineering function
g_features = utils.feature_engineering(g_data, g_meta, False)
if i_c == 0:
g_features.to_csv('test_g_features.csv',
header=True,
mode='a',
index=False)
else:
g_features.to_csv('test_g_features.csv',
header=False,
mode='a',
index=False)
if eg_meta.shape[0] > 0:
eg_features = utils.feature_engineering(eg_data, eg_meta, False)
if i_c == 0:
import utils
#path_to_data = '/courses/cs342/Assignment2/'
path_to_data = ''
train, train_meta = utils.load_train(path_to_data)
g_train, eg_train, g_meta, eg_meta, g_target, eg_target = utils.gal_split_data(
train, train_meta, True)
g_features = utils.feature_engineering(g_train, g_meta)
g_wtable, g_labels, g_classes, g_target_map = utils.preprocess_target(g_target)
g_features = utils.standardize_data(g_features)
utils.train_mlp(g_features, g_wtable, g_labels, g_classes, g_target_map, True)
eg_features = utils.feature_engineering(eg_train, eg_meta)
eg_wtable, eg_labels, eg_classes, eg_target_map = utils.preprocess_target(
eg_target)
eg_features = utils.standardize_data(eg_features)
utils.train_mlp(eg_features, eg_wtable, eg_labels, eg_classes, eg_target_map,
False)