dmodel1 = {"acc": [], "dt": [], "fsize": []}
dmodel = {
"data": copy.deepcopy(dmodel1),
"count": [],
"files": [],
"avg": copy.deepcopy(dmodel1),
"top": copy.deepcopy(dmodel1)
}
# create separate models for each data file
for filename in filenames:
data_file = root_data_folder + "/" + filename + ".csv"
x, y, _, _ = loader.load_dataset(data_file)
acc_train_vect[filename] = copy.deepcopy(dmodel)
acc_test_vect[filename] = copy.deepcopy(dmodel)
# print(y)
# binarize the outputs
y = loader.binarize(y)
if config["one_hot_encoding"]:
# use integer encoding
y = prep.encode(prep.adapt_input(y))
y = prep.decode_int_onehot(y)
# print(y)
# quit()
# y = prep.encode(prep.adapt_input(y))
input_file = "./data/exp_39.csv"
if use_matching_random_model:
model_file = root_crt_model_folder + "/" + "exp_179_1_top.h5"
# model_file = root_crt_model_folder + "/" + "exp_217_2_top.h5"
else:
model_file = root_crt_model_folder + "/" + "exp_39_3_multi_top.skl"
else:
input_file = "./data/exp_39.csv"
model_file = root_crt_model_folder + "/" + "exp_39_5_top.h5"
nvalves = config["n_valves"]
nrowskip = 0
# X1, y1 = loader.load_dataset_raw_buffer(input_file)
X1, y1, _, _ = loader.load_dataset(input_file)
# X1 = X1[120:1700]
# y1 = y1[120:1700]
# binarize the outputs
y1 = loader.binarize(y1)
s = np.shape(X1)
print(s)
nrows = s[0]
ncols = s[1]
n_bins = 20
rowskip = int(nrows / n_bins)
save_best_model = True
if n_reps > 1:
use_saved_model = False
append_timestamp = True
save_best_model = True
else:
save_best_model = False
# bookmarks = [bookmarks[-1]]
from_bookmark_index = len(bookmarks) - 1
# create separate models for each data file
for filename in filenames:
data_file = root_data_folder + "/" + filename + ".csv"
x, y = loader.load_dataset(data_file)
x_train = x
y_train = y
x_eval = x
y_eval = y
sizex = np.shape(x_train)
for bookmark_index in range(len(bookmarks)):
if bookmark_index < from_bookmark_index:
continue
x_train = x[0:bookmarks[bookmark_index], :]
y_train = y[0:bookmarks[bookmark_index], :]
x_eval = x[0:bookmarks[len(bookmarks) - 1], :]
y_eval = y[0:bookmarks[len(bookmarks) - 1], :]
# x = remove_outliers(x)
# tss = create_timeseries(x, xheader)
# fig, _ = graph.plot_timeseries_multi(tss, "sensor output", "samples [x0.1s]", "flow [L/h]", False)
# graph.save_figure(fig, "./figs/sensor_output")
# graph.plot_timeseries(ts, "title", "x", "y")
# quit()
# create separate models for each data file
for filename in filenames:
data_file = root_data_folder + "/" + filename + ".csv"
x, y, xheader, yheader, times = loader.load_dataset(data_file)
# tss = create_timeseries(x, xheader)
# TODO: sort by chan number 0 - 10
# TODO: show as subplot
print(xheader)
print(yheader)
print(len(xheader))
order = [0, 1, 3, 4, 5, 6, 7, 8, 9, 10, 2]
xheader = reorder(xheader, order)