def get_prediction(ANN, timed_Xtest, timed_ytest):
# timed_ypred = mdl.get_pred_timed(ANN, timed_Xtest, data.drop(['label','shock_ind'],axis=1))
timed_ypred = mdl.get_pred_timed(ANN, timed_Xtest,
data.drop(['label'], axis=1))
# raw_proba = mdl.get_prob_timed(ANN, timed_Xtest, data.drop(['label','shock_ind'],axis=1))
raw_proba = mdl.get_prob_timed(ANN, timed_Xtest,
data.drop(['label'], axis=1))
# timed_ypred = dat.append_data_to_timed(timed_ypred, data, ['x', 'y', 'z', 'rho'])
# raw_proba = dat.append_data_to_timed(raw_proba, data, ['x', 'y', 'z', 'rho'])
#variations
pred_variations = dat.get_var(timed_ypred)
true_variations = dat.get_var(timed_ytest)
true_variations = dat.get_category(true_variations)
pred_variations = dat.get_closest_var_by_cat(
true_variations, dat.get_category(pred_variations))
#
# true_variations = dat.append_data_to_timed(true_variations, data, ['x', 'y', 'z', 'rho'])
# pred_variations = dat.append_data_to_timed(pred_variations, data, ['x', 'y', 'z', 'rho'])
#crossings reference
true_crossings = dat.crossings_from_var(true_variations)
return timed_ypred, raw_proba, true_variations, pred_variations, true_crossings
def pred_from_unseen(model, unseen_data, scale_data, dt_corr, dt_density):
unseen_data = unseen_data.fillna(scale_data.median())
init_pred = mdl.get_pred_timed(model, unseen_data, scale_data)
proba = mdl.get_prob_timed(model, unseen_data, scale_data)
init_var = dat.get_var(init_pred)
init_var = dat.get_category(init_var)
corr_pred = dat.get_corrected_pred(init_var, init_pred, proba, dt_corr)
vcorr = dat.get_category(dat.get_var(corr_pred))
vcorr = dat.corrected_var(vcorr, 15) #deletes variations faster than 15s
corr_crossings = dat.crossings_from_var(vcorr)
corr_pred = dat.crossings_density(corr_pred, corr_crossings, dt_density)
final_crossings = dat.final_list(corr_pred)
return corr_pred, vcorr, final_crossings
def run_prediction():
global timed_ypred
global raw_proba
timed_ypred = mdl.get_pred_timed(ANN, timed_Xtest, data.drop('label',axis=1))
raw_proba = mdl.get_prob_timed(ANN, timed_Xtest, data.drop('label',axis=1))
timed_ypred = dat.append_data_to_timed(timed_ypred, data, ['x', 'y', 'z', 'rho'])
raw_proba = dat.append_data_to_timed(raw_proba, data, ['x', 'y', 'z', 'rho'])
global true_variations
global pred_variations
pred_variations = dat.get_var(timed_ypred)
true_variations = dat.get_var(timed_ytest)
true_variations = dat.get_category(true_variations)
pred_variations = dat.get_closest_var_by_cat(true_variations, dat.get_category(pred_variations))
true_variations = dat.append_data_to_timed(true_variations, data, ['x', 'y', 'z', 'rho'])
pred_variations = dat.append_data_to_timed(pred_variations, data, ['x', 'y', 'z', 'rho'])
global true_crossings
true_crossings = dat.crossings_from_var(true_variations)
def get_corrected_prediction(timed_ypred, raw_proba, true_variations,
pred_variations):
timed_ycorr = dat.get_corrected_pred(pred_variations, timed_ypred,
raw_proba, dt_corr_pred)
corr_variations = dat.get_var(timed_ycorr)
corr_variations = dat.corrected_var(
corr_variations, 15) #deletes variations faster than 15s
corr_variations = dat.get_closest_var_by_cat(
true_variations, dat.get_category(corr_variations))
# timed_ycorr = dat.append_data_to_timed(timed_ycorr, data, ['x', 'y', 'z', 'rho'])
# corr_variations = dat.append_data_to_timed(corr_variations, data, ['x', 'y', 'z', 'rho'])
corr_crossings = dat.crossings_from_var(corr_variations)
return timed_ycorr, corr_variations, corr_crossings
def run_correction():
global timed_ycorr
global corr_variations
global Dt_corr_pred
global corr_crossings
Dt_corr_pred = int(Dt_var_selection.get())
timed_ycorr = dat.get_corrected_pred(pred_variations,timed_ypred, raw_proba, Dt_corr_pred)
corr_variations = dat.get_var(timed_ycorr)
corr_variations = dat.get_closest_var_by_cat(true_variations, dat.get_category(corr_variations))
timed_ycorr = dat.append_data_to_timed(timed_ycorr, data, ['x', 'y', 'z', 'rho'])
corr_variations = dat.append_data_to_timed(corr_variations, data, ['x', 'y', 'z', 'rho'])
corr_crossings = dat.crossings_from_var(corr_variations)
corr_crossings = dat.get_closest_cross(true_crossings, corr_crossings)