def workload_characterization():
'''path = r'/Users/rachananarayanacharya/MLbasedDBMStuning/Data/' # use your path
all_files = glob.glob(path + "/*.csv")
li = []
for filename in all_files:
df = pd.read_csv(filename, index_col=None, header=0)
li.append(df)
df = pd.concat(li, axis=0, ignore_index=True)
train_df, test_df = train_test_split(df, test_size=0.2)'''
#train_df=pd.read_pickle("../../Data/WorkloadFiles/workload_11.pkl",compression=None)
train_df3=pd.read_csv("../../Data/CSVFiles/Online_workloadB_preprocessed.csv")
#train_df1=pd.read_csv("../../Data/CSVFiles/Online_workloadC_preprocessed.csv")
train_df2=pd.read_csv("../../Data/CSVFiles/Offline_workload_preprocessed.csv")
frames = [train_df2, train_df3]
train_df = pd.concat(frames)
train_df=train_df.fillna(0)
model = FA()
X=train_df.to_numpy()
print("Shape of the data:",X.shape)
n_rows, n_cols = X.shape
model=model._fit(X, 50, 10000)
components = model.components_.T.copy()
print("After Components: ",components.shape)
c_algo="kmeans" # kmeans or em
clustering_model, det=select_clustering_model(c_algo)
clustering_model.fit(components, min_cluster=1,
max_cluster=min(n_cols - 1, 20),
estimator_params={'n_init': 50})
det.fit(components, clustering_model.cluster_map_)
print("Optimal no of clusters:",det.optimal_num_clusters_)
#print(clustering_model.cluster_map_)
pruned_metrics = clustering_model.cluster_map_[det.optimal_num_clusters_].get_closest_samples()
print("pruned metrics:",pruned_metrics)
idx_set = columnsToPrune(pruned_metrics)
cols_rem = []
for i in range(14,373):
if i not in idx_set:
cols_rem.append(i)
print(len(idx_set))
print_pruned_metrics(idx_set, train_df, c_algo)
loadWorkLoadRemCols(cols_rem)
writeCSV("pruned_metric.csv",pruned_metrics)
def make_nfa(self):
"""make NFA from the state graph"""
nfa = FA()
for current in self.all_states(self.start):
for edge in current.move:
for next_node in current.move[edge]:
nfa.connect(current.name, next_node.name, edge)
nfa.set_start(self.start.name)
nfa.add_final(self.final.name)
return nfa
def output_plain(this):
FA.output_plain(this, "NFA", "Thompson NFA")
def output_dot(this):
FA.output_dot(this, "NFA")
def to_fa(otatable, n):
"""Given an ota table, build a finite automaton.
"""
### First, need to transform the resettimedwords of the elements in S_U_R
### to clock valuation timedwords with reset informations.
#S_U_R = [s for s in otatable.S] + [r for r in otatable.R]
#table_elements = [Element(dRTWs_to_lRTWs(e.tws), e.value) for e in S_U_R]
table_elements = [s for s in otatable.S] + [r for r in otatable.R]
### build a finite automaton
## FAStates
rtw_alphabet = []
states = []
initstate_name = ""
accept_names = []
value_name_dict = {}
sink_name = ""
for s, i in zip(otatable.S, range(1, len(otatable.S) + 1)):
name = str(i)
value_name_dict[s.whichstate()] = name
init = False
accept = False
if s.tws == []:
init = True
initstate_name = name
if s.value[0] == 1:
accept = True
accept_names.append(name)
if s.value[0] == -1:
sink_name = name
temp_state = FAState(name, init, accept)
states.append(temp_state)
## FATrans
trans_number = 0
trans = []
for r in table_elements:
if r.tws == []:
continue
resettimedwords = [tw for tw in r.tws]
w = resettimedwords[:-1]
a = resettimedwords[len(resettimedwords) - 1]
if a not in rtw_alphabet:
rtw_alphabet.append(a)
source = ""
target = ""
#label = [a]
for element in table_elements:
if w == element.tws:
source = value_name_dict[element.whichstate()]
if resettimedwords == element.tws:
target = value_name_dict[element.whichstate()]
need_newtran = True
for tran in trans:
if source == tran.source and target == tran.target:
if a.action == tran.label[0].action and a.reset == tran.label[
0].reset:
need_newtran = False
if a not in tran.label:
tran.label.append(a)
break
if need_newtran == True:
temp_tran = FATran(trans_number, source, target, [a])
trans.append(temp_tran)
trans_number = trans_number + 1
fa = FA("FA_" + str(n), rtw_alphabet, states, trans, initstate_name,
accept_names)
return fa, sink_name
