def cal_gama_feature_table(XL_train,
YL_train,
XU_train,
YU_train,
gama_list,
gama_feature_table_path=None):
gama_label = []
fn = XL_train.shape[1]
xf = range(fn)
feature_order_list = []
for gama in gama_list:
XL, YL, XU, YU = XL_train.copy(), YL_train.copy(), XU_train.copy(
), YU_train.copy()
print(str(gama) + " : lsfs feature select start")
YL = read_data.label_n1_to_nc(YL)
YU = read_data.label_n1_to_nc(YU)
feature_order, time_dual = lsfs(XL, YL, XU, gama=gama)
feature_order_list.append(feature_order)
gama_label.append(str(gama))
print("finished feature select!")
# print(exc_fun_label)
gama_feature_table = pd.DataFrame(data=np.array(feature_order_list),
index=gama_label,
columns=xf)
if gama_feature_table_path != None:
gama_feature_table.to_csv(gama_feature_table_path)
return gama_feature_table
def compute_acc_diff_gama_fearture(XL_train, YL_train, XU_train, YU_train, gama_array, idx_array\
, output_file_name="feature_order"):
XL, YL, XU, YU = XL_train.copy(), YL_train.copy(), XU_train.copy(
), YU_train.copy()
YL = read_data.label_n1_to_nc(YL)
YU = read_data.label_n1_to_nc(YU)
data = []
for gama in gama_array:
feature_order, time_dual = lsfs(XL,
YL,
XU,
output_file_name=output_file_name,
gama=gama)
acc_array = evaluate.cal_many_acc_by_idx2(XL_train, YL_train, XU_train, YU_train,\
feature_order, idx_array)
data.append(acc_array)
def run_accuracy(fun,
XL_train,
YL_train,
XU_train,
YU_train,
sel_num=5,
output_file_name="feature_order"):
XL, YL, XU, YU = XL_train.copy(), YL_train.copy(), XU_train.copy(
), YU_train.copy()
if fun.__name__.lower() == "lsfs":
YL = read_data.label_n1_to_nc(YL)
YU = read_data.label_n1_to_nc(YU)
feature_order, time_dual = fun(XL,
YL,
XU,
output_file_name=output_file_name)
X,Y = evaluate.select_data(XL_train, YL_train, XU_train, YU_train,\
feature_order, sel_num=sel_num)
a = evaluate.run_acc(X, Y)
print("accuracy", ":", a)
return feature_order, time_dual, a
def cal_feature_order_table(XL_train, YL_train, XU_train, YU_train,fun_list, gama_list, feature_order_table_path = None):
exc_fun_label = []
fn = XL_train.shape[1]
xf = range(fn)
feature_order_list = []
for fun in fun_list:
XL, YL, XU, YU = XL_train.copy(), YL_train.copy(), XU_train.copy(), YU_train.copy()
fun_name = fun.__name__
if fun_name == "lsfs":
print("lsfs feature select start")
YL = read_data.label_n1_to_nc(YL)
YU = read_data.label_n1_to_nc(YU)
for gama in gama_list:
feature_order, time_dual = lsfs(XL, YL, XU, gama = gama)
exc_fun_label.append("LSFS :" + str(gama))
feature_order_list.append(feature_order)
with open("time.txt", "a+") as f:
print(fun_name + " " + str(gama) + " time : " + str(time_dual), file=f)
elif fun_name == "fisher_score":
print("fisher_score feature select start")
# feature_order, time_dual = fisher_score(XL_train, YL_train, output_file_name=output_file_name)
feature_order, time_dual = fun(XL, YL)
exc_fun_label.append("Fisher score")
with open("time.txt", "a+") as f:
print(fun_name + " time : " + str(time_dual), file=f)
elif fun_name == "laplacian_score_feature_order2":
print("laplacian_score_feature_order2 feature select start")
# feature_order, time_dual = laplacian_score_feature_order2(XL, k = 10, output_file_name=output_file_name)
feature_order, time_dual = fun(XL, k = 10)
exc_fun_label.append("Laplacian score")
with open("time.txt", "a+") as f:
print(fun_name + " time : " + str(time_dual), file=f)
elif fun_name == "lsdf":
print("lsdf feature select start")
# feature_order, time_dual = lsdf(XL, YL, XU, output_file_name=output_file_name)
feature_order, time_dual = fun(XL, YL, XU)
exc_fun_label.append("LSDF")
with open("time.txt", "a+") as f:
print(fun_name + " time : " + str(time_dual), file=f)
elif fun_name == "prpc":
print("prpc feature select start")
feature_order, time_dual = fun(XL, YL, XU)
exc_fun_label.append("PRPC")
with open("time.txt", "a+") as f:
print(fun_name + " time : " + str(time_dual), file=f)
elif fun_name == "sSelect":
print("sSelect feature select start")
feature_order, time_dual = fun(XL, YL, XU, k = 10, theta = 10, namuda = 0.1)
exc_fun_label.append("sSelect")
with open("time.txt", "a+") as f:
print(fun_name + " time : " + str(time_dual), file=f)
if fun_name != "lsfs":
feature_order_list.append(feature_order)
# print(feature_order)
with open("time.txt", "a+") as f:
print("======================================================================", file=f)
print("finished feature select!")
# print(exc_fun_label)
feature_order_table = pd.DataFrame(data=np.array(feature_order_list), index=exc_fun_label, columns=xf)
if feature_order_table_path != None:
feature_order_table.to_csv(feature_order_table_path)
return feature_order_table
output_file_name_nei = "..\\result\\iter_objFun\\" + "lsfs_iter_objFun_result_nei_" + file_path.split("\\")[-2] + "_" + \
str(example_rate) + "_" + str(feature_rate) + "_g" + str(gama)
output_file_name_wai = "..\\result\\iter_objFun\\" + "lsfs_iter_objFun_result_wai_" + file_path.split("\\")[-2] + "_" + \
str(example_rate) + "_" + str(feature_rate) + "_g" + str(gama)
XL_train, YL_train, XU_train, YU_train = get_data(file_path, selected_data_file_name, selected_cluster_name_file_name, \
unselected_data_file_name, unselected_cluster_name_file_name, example_rate, feature_rate)
# feature_order, time_dual, a = run_accuracy(lsfs, XL_train,YL_train,XU_train,YU_train, 10, output_file_name)
XL, YL, XU, YU = XL_train.copy(), YL_train.copy(), XU_train.copy(
), YU_train.copy()
YL = read_data.label_n1_to_nc(YL)
YU = read_data.label_n1_to_nc(YU)
iterations_wai = get_lsfs_obj_value(
XL,
YL,
XU,
gama=gama,
output_file_path=output_file_name_nei + ".png")
plot_iter_obj_funValue(iterations_wai,
xlabel_name=x_label,
ylabel_name=y_label,
output_file_path=output_file_name_wai + ".png")
print("gama : ", gama, " finish")