def RF(dataset, max_depth=3):
dataset = normalize(dataset)
mre_list = []
sa_list = []
rse_list = []
for train, test in KFold_df(dataset, 3):
train_input = train.iloc[:, :-1]
train_actual_effort = train.iloc[:, -1]
test_input = test.iloc[:, :-1]
test_actual_effort = test.iloc[:, -1]
model = RandomForestRegressor(max_depth)
model.fit(train_input, train_actual_effort)
test_predict_effort = model.predict(test_input)
test_predict_Y = test_predict_effort
test_actual_Y = test_actual_effort.values
mre_list.append(mre_calc(test_predict_Y,
test_actual_Y)) ######### for MRE
sa_list.append(
sa_calc(test_predict_Y, test_actual_Y,
train_actual_effort)) ######### for SA
# rse_list.append(rse_calc(test_predict_Y, test_actual_Y)) ######### for RSE
return mre_list, sa_list, rse_list
def CART(dataset, a=12, b=1, c=2):
dataset = normalize(dataset)
mre_list = []
sa_list = []
rse_list = []
for train, test in KFold_df(dataset, 3):
train_input = train.iloc[:, :-1]
train_actual_effort = train.iloc[:, -1]
test_input = test.iloc[:, :-1]
test_actual_effort = test.iloc[:, -1]
# max_depth: [1:12], min_samples_leaf: [1:12], min_samples_split: [2:21]
model = DecisionTreeRegressor(max_depth=a,
min_samples_leaf=b,
min_samples_split=c)
model.fit(train_input, train_actual_effort)
test_predict_effort = model.predict(test_input)
test_predict_Y = test_predict_effort
test_actual_Y = test_actual_effort.values
mre_list.append(mre_calc(test_predict_Y,
test_actual_Y)) ######### for MRE
sa_list.append(
sa_calc(test_predict_Y, test_actual_Y,
train_actual_effort)) ######### for SA
# rse_list.append(rse_calc(test_predict_Y, test_actual_Y)) ######### for RSE
return mre_list, sa_list, rse_list
def SVM(dataset):
dataset = normalize(dataset)
mre_list = []
sa_list = []
rse_list = []
for train, test in KFold_df(dataset, 3):
train_input = train.iloc[:, :-1]
train_actual_effort = train.iloc[:, -1]
test_input = test.iloc[:, :-1]
test_actual_effort = test.iloc[:, -1]
model = svm.SVR(gamma='scale')
model.fit(train_input, train_actual_effort)
test_predict_effort = model.predict(test_input)
test_predict_Y = test_predict_effort
test_actual_Y = test_actual_effort.values
mre_list.append(mre_calc(test_predict_Y,
test_actual_Y)) ######### for MRE
sa_list.append(
sa_calc(test_predict_Y, test_actual_Y,
train_actual_effort)) ######### for SA
# rse_list.append(rse_calc(test_predict_Y, test_actual_Y)) ######### for RSE
return mre_list, sa_list, rse_list
def KNN(dataset, n_neighbors=3):
dataset = normalize(dataset)
mre_list = []
sa_list = []
rse_list = []
for train, test in KFold_df(dataset, 3):
train_input = train.iloc[:, :-1]
train_actual_effort = train.iloc[:, -1]
test_input = test.iloc[:, :-1]
test_actual_effort = test.iloc[:, -1]
model = neighbors.KNeighborsRegressor(n_neighbors)
model.fit(train_input, train_actual_effort)
test_predict_effort = model.predict(test_input)
test_predict_Y = test_predict_effort
test_actual_Y = test_actual_effort.values
mre_list.append(mre_calc(test_predict_Y,
test_actual_Y)) ######### for MRE
sa_list.append(
sa_calc(test_predict_Y, test_actual_Y,
train_actual_effort)) ######### for SA
# rse_list.append(rse_calc(test_predict_Y, test_actual_Y)) ######### for RSE
return mre_list, sa_list, rse_list
def cart_builder_future(a, b, c):
model = DecisionTreeRegressor(
max_depth=a,
min_samples_leaf=b,
min_samples_split=c
)
model.fit(validate_train_input, validate_train_output)
validate_test_predict = model.predict(validate_test_input)
validate_test_actual = validate_test_output.values
if metrics == 0:
return mre_calc(validate_test_predict, validate_test_actual)
if metrics == 1:
return sa_calc(validate_test_predict, validate_test_actual, validate_train_output)
def flash(train_input, train_actual_effort, test_input, test_actual_effort,
metrics, pop_size):
def convert(index): # 12 12 20
a = int(index / 240 + 1)
b = int(index % 240 / 20 + 1)
c = int(index % 20 + 2)
return a, b, c
def convert_lr(index): # 30 2 2 100
a = int(index / 400 + 1)
b = int(index % 400 / 200 + 1)
c = int(index % 200 / 100 + 1)
d = int(index % 100 + 1)
return a, b, c, d
all_case = set(range(0, 2880))
modeling_pool = random.sample(all_case, pop_size)
List_X = []
List_Y = []
for i in range(len(modeling_pool)):
temp = convert(modeling_pool[i])
List_X.append(temp)
model = DecisionTreeRegressor(max_depth=temp[0],
min_samples_leaf=temp[1],
min_samples_split=temp[2])
model.fit(train_input, train_actual_effort)
test_predict_effort = model.predict(test_input)
test_predict_Y = test_predict_effort
test_actual_Y = test_actual_effort.values
if metrics == 0:
List_Y.append(mre_calc(test_predict_Y,
test_actual_Y)) ######### for MRE
if metrics == 1:
List_Y.append(
sa_calc(test_predict_Y, test_actual_Y,
train_actual_effort)) ######### for SA
remain_pool = all_case - set(modeling_pool)
test_list = []
for i in list(remain_pool):
test_list.append(convert(i))
upper_model = DecisionTreeRegressor()
life = 20
while len(List_X) < 201 and life > 0: # eval_number
upper_model.fit(List_X, List_Y)
candidate = random.sample(test_list, 1)
test_list.remove(candidate[0])
candi_pred_value = upper_model.predict(candidate)
if metrics == 0:
if candi_pred_value < np.median(List_Y): ######### for MRE
List_X.append(candidate[0])
candi_config = candidate[0]
candi_model = DecisionTreeRegressor(
max_depth=candi_config[0],
min_samples_leaf=candi_config[1],
min_samples_split=candi_config[2])
candi_model.fit(train_input, train_actual_effort)
candi_pred_Y = candi_model.predict(test_input)
candi_actual_Y = test_actual_effort.values
List_Y.append(mre_calc(candi_pred_Y,
candi_actual_Y)) ######### for MRE
else:
life -= 1
if metrics == 1:
if candi_pred_value > np.median(List_Y): ######### for SA
List_X.append(candidate[0])
candi_config = candidate[0]
candi_model = DecisionTreeRegressor(
max_depth=candi_config[0],
min_samples_leaf=candi_config[1],
min_samples_split=candi_config[2])
candi_model.fit(train_input, train_actual_effort)
candi_pred_Y = candi_model.predict(test_input)
candi_actual_Y = test_actual_effort.values
List_Y.append(
sa_calc(candi_pred_Y, candi_actual_Y,
train_actual_effort)) ######### for SA
else:
life -= 1
# temp_tree = candi_model
# tree.plot_tree(temp_tree, feature_names=list(train_input.columns.values))
# plt.show()
if metrics == 0:
return np.median(List_Y) ########## MRE
if metrics == 1:
return np.median(List_Y) ########## SA
def CART_DE(dataset, metrics, month):
dataset = normalize(dataset)
for trainset, testset in df_split(dataset, month):
train_input = trainset.iloc[:, :-1]
train_output = trainset.iloc[:, -1]
test_input = testset.iloc[:, :-1]
test_output = testset.iloc[:, -1]
for validate_trainset, validate_testset in df_split(trainset, 1):
validate_train_input = validate_trainset.iloc[:, :-1]
validate_train_output = validate_trainset.iloc[:, -1]
validate_test_input = validate_testset.iloc[:, :-1]
validate_test_output = validate_testset.iloc[:, -1]
def cart_builder(a, b, c):
model = DecisionTreeRegressor(
max_depth=a,
min_samples_leaf=b,
min_samples_split=c
)
model.fit(train_input, train_output)
test_predict = model.predict(test_input)
test_actual = test_output.values
if metrics == 0:
return mre_calc(test_predict, test_actual)
if metrics == 1:
return sa_calc(test_predict, test_actual, train_output)
def cart_builder_future(a, b, c):
model = DecisionTreeRegressor(
max_depth=a,
min_samples_leaf=b,
min_samples_split=c
)
model.fit(validate_train_input, validate_train_output)
validate_test_predict = model.predict(validate_test_input)
validate_test_actual = validate_test_output.values
if metrics == 0:
return mre_calc(validate_test_predict, validate_test_actual)
if metrics == 1:
return sa_calc(validate_test_predict, validate_test_actual, validate_train_output)
# config_optimized = de(cart_builder, metrics, bounds=[(10,20), (1,10), (2,12)])[1]
config_optimized = de(cart_builder_future, metrics, bounds=[(10, 20), (1, 10), (2, 12)])[1]
# print(config_optimized[0], config_optimized[1], config_optimized[2])
model_touse = DecisionTreeRegressor(
max_depth=config_optimized[0],
min_samples_leaf=config_optimized[1],
min_samples_split=config_optimized[2]
)
model_touse.fit(train_input, train_output)
test_predict = np.rint(model_touse.predict(test_input))
test_actual = test_output.values
result_list_mre = []
result_list_sa = []
result_list_mre.append(mre_calc(test_predict, test_actual))
result_list_sa.append(sa_calc(test_predict, test_actual, train_output))
# print("pre", test_predict, "act", test_actual)
if metrics == 0:
return result_list_mre
if metrics == 1:
return result_list_sa