def PrivCheck(df_train, df_test, df_test_rec_items, grid_area_dict,
area_grid_dict, grid_list, cluster_num, grid_area_number, deltaX,
pp):
df_train = funcs.update_grid_group(df_train, grid_area_dict)
model_rf = funcs.train_rf_model_check_in(df_train)
# xgboost
model_xgb = funcs.train_xgb_model_check_in(df_train)
pd.DataFrame(
funcs.cal_pgy_withoutGridGroup(df_test, cluster_num,
grid_list)).to_csv(
'tmp/pgy_check_in_privcheck.csv',
index=False,
header=None)
df_test = funcs.update_grid_group(df_test, grid_area_dict)
JSD_Mat_dict = np.zeros((cluster_num, cluster_num, grid_area_number))
group_user_size_dict = {}
for gg in range(grid_area_number):
df_test_gg = df_test.loc[df_test['grid_group'] == gg]
grid_list_gg = area_grid_dict[gg]
group_user_size_dict[gg] = df_test_gg.shape[0]
JSD_Mat_dict[:, :, gg] = funcs.cal_JSD_Matrix_withoutGridGroup(
df_test_gg, cluster_num, 4)
scipy.io.savemat('tmp/JSDM_girdGroup_privcheck.mat',
{"JSD_Mat_input_Yang_trueTrain": JSD_Mat_dict})
eng = matlab.engine.start_matlab()
eng.edit("../../matlab/checkin_tradeoff_scenario_I/PrivCheck", nargout=0)
eng.cd('../../matlab/checkin_tradeoff_scenario_I', nargout=0)
xpgg, distortion_budget = np.array(eng.PrivCheck(deltaX, nargout=2))
xpgg = np.array(xpgg)
df_test['grid_group'] = pd.Series(np.zeros(df_test.shape[0]),
index=df_test.index,
dtype='int32')
X_obf_dict = {}
for i in range(50):
X_obf_dict[i], _ = funcs.get_obf_X(df_test, xpgg, pp)
_, X_ori = funcs.get_obf_X(df_test, xpgg, pp)
for i in X_ori.keys():
user_grid = X_ori[i][-2]
X_ori[i][-3] = grid_area_dict[user_grid]
for j in range(50):
X_obf_dict[j][i][-1] = grid_area_dict[user_grid]
df_test_rec_items = funcs.update_grid_group(df_test_rec_items,
grid_area_dict)
return X_obf_dict, X_ori, model_rf, model_xgb
def PrivCheck(df_train, cluster_num, grid_list, grid_area_dict,
grid_area_number, area_grid_dict, deltaX, pp):
pd.DataFrame(
funcs.cal_pgy_withoutGridGroup(df_train, cluster_num,
grid_list)).to_csv(
'tmp/pgy_check_in_privcheck.csv',
index=False,
header=None)
df_train = funcs.update_grid_group(df_train, grid_area_dict)
JSD_Mat_dict = np.zeros((cluster_num, cluster_num, grid_area_number))
group_user_size_dict = {}
for gg in range(grid_area_number):
df_train_gg = df_train.loc[df_train['grid_group'] == gg]
grid_list_gg = area_grid_dict[gg]
group_user_size_dict[gg] = df_train_gg.shape[0]
JSD_Mat_dict[:, :, gg] = funcs.cal_JSD_Matrix_withoutGridGroup(
df_train_gg, cluster_num, 4)
scipy.io.savemat('tmp/JSDM_girdGroup_privcheck.mat',
{"JSD_Mat_input_Yang_allObf": JSD_Mat_dict})
eng = matlab.engine.start_matlab()
eng.edit("../../matlab/checkin_clusternum_scenario_II/PrivCheck",
nargout=0)
eng.cd('../../matlab/checkin_clusternum_scenario_II', nargout=0)
xpgg, distortion_budget = np.array(eng.PrivCheck(deltaX, nargout=2))
xpgg = np.array(xpgg)
df_train = funcs.update_grid_group(df_train, grid_area_dict)
X_obf_dict = {}
for i in range(25):
X_obf_dict[i], _ = funcs.get_obf_X_withAgeGroup(df_train, xpgg, pp)
_, X_ori = funcs.get_obf_X_withAgeGroup(df_train, xpgg, pp)
for i in X_ori.keys():
user_grid = X_ori[i][-2]
X_ori[i][-3] = grid_area_dict[user_grid]
for j in range(25):
X_obf_dict[j][i][-1] = grid_area_dict[user_grid]
return X_obf_dict, X_ori