# import joblib # import eli5 feature_long_names = pd.read_json('/work/demonstration/datasets/Features_Metadata.json') feature_names = get_features_name() pi = pd.read_csv('../4_data_modelling/permutation_importance_train_f1.csv') from IPython.display import HTML, display def display_table(data: pd.DataFrame): html = "" html += "" for i, col in enumerate(data.columns): html+="".format('right' if i%2==0 else 'left', '16px' if i%2==0 else '0px', col) html += '' for i, row in data.iterrows(): style = 'background:transparent !important; border:0px;' if i==6: style += 'border:2px solid red; font-size:13px;font-weight:bold !important;' html += "".format(style, 'bold' if i==6 else 'normal') for j, field in enumerate(row): html += "".format( '0 16px 0 4px' if j%2 == 0 else '0 4px 0 0', field) html += "" html += "

{}
{}

" display(HTML(html)) pi['Weight'] = pi['weight'].apply(lambda v: format(v, '.4f')) + ' ± ' + pi['std'].apply(lambda v: format(v, '.4f')) pi['index'] = pi['feature'].str.replace('X','').astype(int) pi = pi.sort_values('index') pi['Feature'] = pi['feature'] + ': ' + feature_names pi.sort_values('weight',ascending=False, inplace=True) display_table(pi[['Weight', 'Feature']]) pi = pd.read_csv('../4_data_modelling/permutation_importance_train_f1.csv') sns.barplot(x='weight', y='feature', data=pi) plt.title('Permutation importance') sns.despine()