class NeuralNetworkWifiTest(NeuralNetworkTest):
def __init__(self):
NeuralNetworkTest.__init__(self)
self.load_in = LoadWifiData()
pass
def load_wifi_data_old(self, dir_path, save_path="wifi_floor.npz"):
file_list = os.listdir(dir_path)
floor_data = {}
for file_name in file_list:
if file_name.endswith("wp"):
name, ext = os.path.splitext(file_name)
print name
wifi_file = name + ".wifi"
wifi_file = os.path.join(dir_path, wifi_file)
wp_file = os.path.join(dir_path, file_name)
floor_data[name] = self.load_in.extract(wp_file, wifi_file)
# temp = load_in.extract_raw(wp_file, wifi_file)
# print temp
# Get All WiFi Listed
t_bind = []
for x in floor_data:
t_bind += floor_data[x].wifi_list
wifi_bld_list = list(set(t_bind))
TD = 0
for x in floor_data:
D = floor_data[x].wifi_matrix.shape[0]
TD += D
L = len(wifi_bld_list)
t_mat = zeros((D, L))
for i, j in enumerate(floor_data[x].wifi_list):
ind = wifi_bld_list.index(j)
t_mat[:, ind] = floor_data[x].wifi_matrix[:, i]
floor_data[x].wifi_bld_mat = t_mat
savez(save_path, floor=floor_data, wifilist=wifi_bld_list, entry_num=TD)
return floor_data, wifi_bld_list, TD
def load_wifi_data_new(self, dir_path, save_path="wifi_floor_temp.npz", ref_list=[]):
load_in = LoadWifiData()
file_list = os.listdir(dir_path)
floor_data = {}
for floor in file_list:
print floor
fl_path = os.path.join(dir_path, floor)
if os.path.isfile(fl_path):
continue
floor_file_list = os.listdir(fl_path)
for file_name in floor_file_list:
if file_name.endswith("wifi"):
name, ext = os.path.splitext(file_name)
print name
wifi_file = name + ".wifi"
wifi_file = os.path.join(fl_path, wifi_file)
floor_data[floor] = load_in.extract_wifi(wifi_file, ref_list)
t_bind = []
for x in floor_data:
t_bind += floor_data[x].wifi_list
wifi_bld_list = list(set(t_bind))
TD = 0
for x in floor_data:
D = floor_data[x].wifi_matrix.shape[0]
TD += D
L = len(wifi_bld_list)
t_mat = zeros((D, L))
for i, j in enumerate(floor_data[x].wifi_list):
ind = wifi_bld_list.index(j)
t_mat[:, ind] = floor_data[x].wifi_matrix[:, i]
floor_data[x].wifi_bld_mat = t_mat
savez(save_path, floor=floor_data, wifilist=wifi_bld_list, entry_num=TD)
return floor_data, wifi_bld_list, TD
def fann_wifi_test_recovered(self, class_num, test_data, normalize=True, savepath="./temp_save.conf"):
if not os.path.exists(savepath):
print "No File Included"
return []
net = libfann.neural_net()
net.create_from_file(savepath)
res = zeros((test_data.shape[0], class_num))
for i, test_array in enumerate(test_data):
if normalize:
# test_array = -test_array / 100.0
test_array = self.normalize(test_array, self.m_min, self.m_max)
res[i, :] = net.run(test_array)
self.net = net
return res
def fann_wifi_test_recovered_accum(self, class_num, test_data, normalize=True, savepath="./temp_save.conf",\
accum_depth=3):
if not os.path.exists(savepath):
print "No File Included"
return []
net = libfann.neural_net()
net.create_from_file(savepath)
res = zeros((test_data.shape[0], class_num))
for i, test_array in enumerate(test_data):
test_array_raw = test_data[max(0, i - accum_depth):i + 1, :]
test_array = self.stack_data(test_array_raw)[0]
if normalize:
# test_array = -test_array / 100.0
test_array = self.normalize(test_array, self.m_min, self.m_max)
# print test_array
res[i, :] = net.run(test_array.round())
self.net = net
return res
def fann_wifi_train(self, floor_data, wifi_bld_list, td, save_path="wifi_floor.conf", normalize=True):
all_data_mat = zeros((td, len(wifi_bld_list)))
all_data_cls = zeros((td, len(floor_data.keys())))
pos_pt = 0
cls_dict = {}
for i, xx in enumerate(floor_data):
d = floor_data[xx].wifi_bld_mat.shape[0]
all_data_mat[pos_pt:pos_pt + d, :] = floor_data[xx].wifi_bld_mat
all_data_cls[pos_pt:pos_pt + d, i] = 1
pos_pt += d
cls_dict[i] = xx
net = self.fann_learn_save_norm(all_data_mat, all_data_cls, save_path, normalize)
self.net = net
return net
def fann_wifi_test(self, floor_data, rec_path="wifi_floor.conf", normalize=True):
cls_dict = {}
for i, xx in enumerate(floor_data):
cls_dict[i] = xx
class_num = len(floor_data.keys())
for x in cls_dict:
floor = cls_dict[x]
st = time.time()
res = self.fann_wifi_test_recovered(class_num, floor_data[floor].wifi_bld_mat, normalize, rec_path)
ed = time.time()
print res
print "Time Comsumption: %s , Ave Time Consumption: %s" % (ed-st, (ed-st) / floor_data['F1'].wifi_bld_mat.shape[0])
res_cls = zeros((res.shape[0],1))
cor_count = 0;
for i, arr in enumerate(res):
res_cls[i] = arr.argmax()
if res_cls[i] == x:
cor_count += 1
print "Res_Cor: %s,Total Amount : %s, Correct Rate: %s" % (cor_count, len(res_cls), cor_count / float(len(res_cls)))
class AeonUtility(SDKUtility): """ Including Utility Functions for Aeon Usage DataStorage Content: self.wp_filepath = wp_filepath self.wifi_filepath = wifi_filepath self.wifi_list = wifi_list self.wp_pos = wp_pos self.wp_ind = wp_ind self.wifi_matrix = wifi_matrix """ def __init__(self): SDKUtility.__init__(self) self.loader = LoadWifiData() self.map_ratio = 10 def load_wifi(self, wp_path, wifi_path, ref_list=[]): if ref_list: wifi = self.loader.extract_with_ref(wp_path, wifi_path, ref_list) else: wifi = self.loader.extract(wp_path, wifi_path) return wifi def extract_aeon(self, wifi_path, ref_list): file_list = os.listdir(wifi_path) route = [] for fn in file_list: if 'journal' in fn: continue data = self.extract_wifi(os.path.join(wifi_path,fn), ref_list) route += data return route def accuracy(self, res, test_pos, train_pos): RMSE = [] for i,n in enumerate(res): pos_res = array(train_pos[n]); pos_tar = array(test_pos[i]); pos_diff = pos_res se = sqrt(((pos_tar - pos_diff) ** 2).sum()) / self.map_ratio RMSE.append(se) return RMSE def pos_weighted(self, w_arr, cls_pos): t_pos = zeros(array(cls_pos[0]).shape) for i, w in enumerate(w_arr): t_pos += array(cls_pos[i]) * w t_pos = t_pos / w_arr.sum() return t_pos def accuracy_proba(self, res, test_pos, train_pos): RMSE = [] for n in range(res.shape[0]): w_arr = res[n,:] pos_res = self.pos_weighted(w_arr, train_pos) pos_tar = test_pos[n] se = sqrt(((pos_tar - pos_res) ** 2).sum()) / self.map_ratio RMSE.append(se) return RMSE def res_to_coord(self, res, train_pos, proba=False): # print res pos = [] for x_res in res: if proba: pos_res = self.pos_weighted(x_res, train_pos) else: pos_res = array(train_pos[x_res]) pos.append(pos_res) return pos
class AeonUtility(SDKUtility): """ Including Utility Functions for Aeon Usage DataStorage Content: self.wp_filepath = wp_filepath self.wifi_filepath = wifi_filepath self.wifi_list = wifi_list self.wp_pos = wp_pos self.wp_ind = wp_ind self.wifi_matrix = wifi_matrix """ def __init__(self): SDKUtility.__init__(self) self.loader = LoadWifiData() self.map_ratio = 1 def load_wifi(self, wp_path, wifi_path, ref_list=[]): if ref_list: wifi = self.loader.extract_with_ref(wp_path, wifi_path, ref_list) else: wifi = self.loader.extract(wp_path, wifi_path) return wifi def extract_aeon(self, wifi_path, ref_list): file_list = os.listdir(wifi_path) route = [] for fn in file_list: if 'journal' in fn: continue data = self.extract_wifi(os.path.join(wifi_path,fn), ref_list) route += data return route def accuracy(self, res, test_pos, train_pos): RMSE = [] for i,n in enumerate(res): pos_res = array(train_pos[n]); pos_tar = array(test_pos[i]); pos_diff = pos_res se = sqrt(((pos_tar - pos_diff) ** 2).sum()) / self.map_ratio RMSE.append(se) return RMSE def pos_weighted(self, w_arr, cls_pos): t_pos = zeros(array(cls_pos[0]).shape) for i, w in enumerate(w_arr): t_pos += array(cls_pos[i]) * w t_pos = t_pos / w_arr.sum() return t_pos def accuracy_proba(self, res, test_pos, train_pos): RMSE = [] for n in range(res.shape[0]): w_arr = res[n,:] pos_res = self.pos_weighted(w_arr, train_pos) pos_tar = test_pos[n] se = sqrt(((pos_tar - pos_res) ** 2).sum()) / self.map_ratio RMSE.append(se) return RMSE def res_to_coord(self, res, train_pos, proba=False): # print res pos = [] for x_res in res: if proba: pos_res = self.pos_weighted(x_res, train_pos) else: pos_res = array(train_pos[x_res]) pos.append(pos_res) return pos