def LASSO_implementation(alpha, train_frame, relevant_vec, stat_vec):
curr_df = train_frame.loc[:, ["id",
"x_Al"]] # from coreallation scatter graphs
for c in stat_vec["NAME"]:
if c == "x_Al": continue
temp_df = pd.merge(curr_df, train_frame.loc[:, ["id", c]], how='outer')
curr_var, model1 = dm_tools.lasso(curr_df.drop(["id"], axis=1),
train_frame[predict_E],
alpha=alpha)
test_var, model2 = dm_tools.lasso(temp_df.drop(["id"], axis=1),
train_frame[predict_E],
alpha=alpha)
curr_df, model = statistics.z_test(curr_var, test_var, defines.alpha_f,
relevant_vec, c, model1, model2,
temp_df, curr_df)
predictions = model.predict(test_frame.loc[:, relevant_vec])
return predictions, model
def POLY_implementation(deg, train_frame, relevant_vec, stat_vec):
# f test on linear regression
curr_df = train_frame.loc[:, ["id",
"x_Al"]] #from coreallation scatter graphs
for c in stat_vec["NAME"]:
if c == "x_Al": continue
temp_df = pd.merge(curr_df, train_frame.loc[:, ["id", c]], how='outer')
curr_var, model1 = dm_tools.poly(curr_df.drop(["id"], axis=1),
train_frame[predict_E],
deg=deg)
test_var, model2 = dm_tools.poly(temp_df.drop(["id"], axis=1),
train_frame[predict_E],
deg=deg)
curr_df, model = statistics.z_test(curr_var, test_var, defines.alpha_f,
relevant_vec, c, model1, model2,
temp_df, curr_df)
poly = PolynomialFeatures(deg)
X_test_transform = poly.fit_transform(test_frame.loc[:, relevant_vec])
predictions = model.predict(X_test_transform)
return predictions, model
curr_df = train_frame.loc[:,
["id", "x_Al"
]] # from coreallation scatter graphs
for c in stat_vec["NAME"]:
if c == "x_Al": continue
temp_df = pd.merge(curr_df,
train_frame.loc[:, ["id", c]],
how='outer')
curr_var, model1 = dm_tools.poly(curr_df.drop(["id"], axis=1),
train_frame[predict_E],
deg=deg)
test_var, model2 = dm_tools.poly(temp_df.drop(["id"], axis=1),
train_frame[predict_E],
deg=deg)
curr_df, model = statistics.z_test(curr_var, test_var,
defines.alpha_f,
relevant_vec, c, model1,
model2, temp_df, curr_df)
poly = PolynomialFeatures(deg)
X_test_transform = poly.fit_transform(test_frame.loc[:,
relevant_vec])
predictions.append(model.predict(X_test_transform))
ex_dic = {
'id': range(1, 601),
'formation_energy_ev_natom': predictions[0],
'bandgap_energy_ev': predictions[1]
}
col = ['id', 'formation_energy_ev_natom', 'bandgap_energy_ev']
df = pd.DataFrame(ex_dic, columns=col)