Automated_classification_of_stroke_deficit/scripts/feature_importance.py

import matplotlib.pyplot as plt
from sklearn.datasets import make_classification
from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestClassifier
from sklearn.inspection import permutation_importance
from sklearn.preprocessing import StandardScaler
from sklearn.metrics import accuracy_score
import numpy as np
import pandas as pd
from master_funcs import *
from collections import Counter
from imblearn.over_sampling import SMOTE
from imblearn.under_sampling import RandomUnderSampler
from imblearn.pipeline import Pipeline
from imblearn.over_sampling import SMOTE
import dill
import warnings
warnings.filterwarnings("ignore")
%matplotlib qt




patients=['PB_T2_3_1', 'PB_T2_3_2', 'PB_T2_3_3', 'PB_T2_4_2', 'PB_T2_5_1', 'PB_T2_6_1', 'PB_T2_6_2', 'PB_T3_23_1', 'PB_T3_23_2', 'PB_T3_23_3', 'PB_T3_24_1', 'PB_T3_24_2', 'PB_T3_24_3']

for z in patients:  
    dill.load_session('/home/user/Documents/Master/GW_'+z+'_data.pkl') 
    names=df_vel_acc.columns.tolist()
    names=[sub.replace('acceleration', 'acc') for sub in names]
    names=[sub.replace('velocity', 'vel') for sub in names]
    names=[sub.replace('angle', 'ang') for sub in names]
    names=[sub.replace('right', 'r') for sub in names]
    names=[sub.replace('left', 'l') for sub in names]
    names=[sub.replace('top', 't') for sub in names]
    names=[sub.replace('bottom', 'b') for sub in names]
    importances = best_model.feature_importances_
    std = np.std([tree.feature_importances_ for tree in best_model.estimators_], axis=0)
    ranked=np.argsort(importances)[::-1]
    std=pd.Series(std)
    
    std=std.reindex(ranked)
    forest_importances = pd.Series(importances, index=names)
    forest_importances=forest_importances.sort_values(ascending=False)
    forest_importances=forest_importances[:20]
    fig, ax= plt.subplots()
    ax.figure.set_size_inches(5, 8)
    forest_importances.plot.bar(yerr=np.array(std[:20]/10), ax=ax)
    ax.set_ylabel('Mean decrease in impurity')
    plt.tight_layout()
    #plt.savefig('/home/user/owncloud/thesis_figures/GW_FI_'+z+'_3d.png', dpi=200)
    plt.show()
    
    
    forest_importances = pd.Series(result.importances_mean, index=names)
    forest_importances=forest_importances.sort_values(ascending=False)
    forest_importances=forest_importances[:20]
    fig, ax = plt.subplots()
    ax.figure.set_size_inches(5, 8)
    forest_importances.plot.bar(yerr=result.importances_std[:20], ax=ax)
    ax.set_ylabel('Mean accuracy decrease')
    fig.tight_layout()
    #fig.savefig('/home/user/owncloud/thesis_figures/GW_FI_PM_'+z+'_3d.png', dpi=200)
    plt.show()

patients=['PB_T2_3_1', 'PB_T2_3_2', 'PB_T2_3_3', 'PB_T2_4_2', 'PB_T2_5_1', 'PB_T2_6_1', 'PB_T2_6_2', 'PB_T3_23_1', 'PB_T3_23_2', 'PB_T3_23_3', 'PB_T3_24_1', 'PB_T3_24_2', 'PB_T3_24_3']

for z in patients:  
    dill.load_session('/home/user/Documents/Master/GW_'+z+'_data.pkl')
    importances = best_model.feature_importances_
    names=df_vel_acc.columns.tolist()
    names=[sub.replace('acceleration', 'acc') for sub in names]
    names=[sub.replace('velocity', 'vel') for sub in names]
    names=[sub.replace('angle', 'ang') for sub in names]
    names=[sub.replace('right', 'r') for sub in names]
    names=[sub.replace('left', 'l') for sub in names]
    names=[sub.replace('top', 't') for sub in names]
    names=[sub.replace('bottom', 'b') for sub in names]
    if z=='PB_T2_3_1':
        total_imp=pd.Series(np.zeros((df_vel_acc.shape[1])), index=names)
        total_imp_perm=pd.Series(np.zeros((df_vel_acc.shape[1])), index=names)
    forest_importances = pd.Series(importances, index=names)
    total_imp+=forest_importances*acc_score
    forest_importances_perm = pd.Series(result.importances_mean, index=names)
    total_imp_perm+=forest_importances_perm*acc_score

total_imp_sort=total_imp.sort_values(ascending=False)
total_imp_sort=total_imp_sort[:20]
fig, ax = plt.subplots()
ax.figure.set_size_inches(5, 8)
total_imp_sort.plot.bar(ax=ax)
ax.set_ylabel('Mean decrease in impurity')
fig.tight_layout()
#fig.savefig('/home/user/owncloud/thesis_figures/GW_FI_total_3d.png', dpi=200)
plt.show()

total_imp_perm_sort=total_imp_perm.sort_values(ascending=False)
total_imp_perm_sort=total_imp_perm_sort[:20]
fig, ax = plt.subplots()
ax.figure.set_size_inches(5, 8)
total_imp_perm_sort.plot.bar(ax=ax)
ax.set_ylabel('Mean decrease in impurity')
fig.tight_layout()
#fig.savefig('/home/user/owncloud/thesis_figures/GW_FI_PM_total_3d.png', dpi=200)
plt.show()

####________________________________

patients=['PB_T2_3_1', 'PB_T2_3_2', 'PB_T2_3_3', 'PB_T2_4_2', 'PB_T2_5_1', 'PB_T2_6_1', 'PB_T2_6_2', 'PB_T3_23_1', 'PB_T3_23_2', 'PB_T3_23_3', 'PB_T3_24_1', 'PB_T3_24_2', 'PB_T3_24_3']


for z in patients:  
    dill.load_session('/home/user/Documents/Master/CY_'+z+'_data.pkl') 
    importances = best_model_touch.feature_importances_
    std = np.std([tree.feature_importances_ for tree in best_model_touch.estimators_], axis=0)
    ranked=np.argsort(importances)[::-1]
    std=pd.Series(std)
    
    std=std.reindex(ranked)
    names=df_vel_acc.columns.tolist()
    names=[sub.replace('acceleration', 'acc') for sub in names]
    names=[sub.replace('velocity', 'vel') for sub in names]
    names=[sub.replace('angle', 'ang') for sub in names]
    names=[sub.replace('right', 'r') for sub in names]
    names=[sub.replace('left', 'l') for sub in names]
    names=[sub.replace('top', 't') for sub in names]
    names=[sub.replace('bottom', 'b') for sub in names]
    forest_importances = pd.Series(importances, index=names)
    forest_importances=forest_importances.sort_values(ascending=False)
    forest_importances=forest_importances[:20]
    fig, ax = plt.subplots()
    ax.figure.set_size_inches(5, 8)
    forest_importances.plot.bar(yerr=np.array(std[:20]/10), ax=ax)
    ax.set_ylabel('Mean decrease in impurity')
    fig.tight_layout()
    #plt.savefig('/home/user/owncloud/thesis_figures/CY_touch_FI_'+z+'_3d.png', dpi=200)
    plt.show()
    
    
    forest_importances = pd.Series(result_touch.importances_mean, index=names)
    forest_importances=forest_importances.sort_values(ascending=False)
    forest_importances=forest_importances[:20]
    fig, ax = plt.subplots()
    ax.figure.set_size_inches(5, 8)
    forest_importances.plot.bar(yerr=result_touch.importances_std[:20], ax=ax)
    ax.set_ylabel('Mean accuracy decrease')
    fig.tight_layout()
    #fig.savefig('/home/user/owncloud/thesis_figures/CY_touch_FI_PM_'+z+'_3d.png', dpi=200)
    plt.show()
     ###_____
     
     
    importances = best_model_drag.feature_importances_
    std = np.std([tree.feature_importances_ for tree in best_model_drag.estimators_], axis=0)
    ranked=np.argsort(importances)[::-1]
    std=pd.Series(std)
    
    std=std.reindex(ranked)
    forest_importances = pd.Series(importances, index=names)
    forest_importances=forest_importances.sort_values(ascending=False)
    forest_importances=forest_importances[:20]
    fig, ax = plt.subplots()
    ax.figure.set_size_inches(5, 8)
    forest_importances.plot.bar(yerr=np.array(std[:20]/10), ax=ax)
    ax.set_ylabel('Mean decrease in impurity')
    fig.tight_layout()
    plt.savefig('/home/user/owncloud/thesis_figures/CY_drag_FI_'+z+'_3d.png', dpi=200)
    plt.show()
    
    
    forest_importances = pd.Series(result_drag.importances_mean, index=names)
    forest_importances=forest_importances.sort_values(ascending=False)
    forest_importances=forest_importances[:20]
    fig, ax = plt.subplots()
    ax.figure.set_size_inches(5, 8)
    forest_importances.plot.bar(yerr=result_drag.importances_std[:20], ax=ax)
    ax.set_ylabel('Mean accuracy decrease')
    fig.tight_layout()
    fig.savefig('/home/user/owncloud/thesis_figures/CY_drag_FI_PM_'+z+'_3d.png', dpi=200)
    plt.show()



patients=['PB_T2_3_1', 'PB_T2_3_2', 'PB_T2_3_3', 'PB_T2_4_2', 'PB_T2_5_1', 'PB_T2_6_1', 'PB_T2_6_2', 'PB_T3_23_1', 'PB_T3_23_2', 'PB_T3_23_3', 'PB_T3_24_1', 'PB_T3_24_2', 'PB_T3_24_3']

for z in patients:  
    dill.load_session('/home/user/Documents/Master/CY_'+z+'_data.pkl')
    names=df_vel_acc.columns.tolist()
    names=[sub.replace('acceleration', 'acc') for sub in names]
    names=[sub.replace('velocity', 'vel') for sub in names]
    names=[sub.replace('angle', 'ang') for sub in names]
    names=[sub.replace('right', 'r') for sub in names]
    names=[sub.replace('left', 'l') for sub in names]
    names=[sub.replace('top', 't') for sub in names]
    names=[sub.replace('bottom', 'b') for sub in names]
    if z=='PB_T2_3_1':
        total_imp_touch=pd.Series(np.zeros((df_vel_acc.shape[1])), index=names)
        total_imp_perm_touch=pd.Series(np.zeros((df_vel_acc.shape[1])), index=names)
        total_imp_drag=pd.Series(np.zeros((df_vel_acc.shape[1])), index=names)
        total_imp_perm_drag=pd.Series(np.zeros((df_vel_acc.shape[1])), index=names)
    importances = best_model_touch.feature_importances_
    forest_importances_touch = pd.Series(importances, index=names)
    total_imp_touch+=forest_importances_touch*acc_score_touch
    forest_importances_perm_touch = pd.Series(result_touch.importances_mean, index=names)
    total_imp_perm_touch+=forest_importances_perm_touch*acc_score_touch
    
    importances = best_model_drag.feature_importances_
    forest_importances_drag = pd.Series(importances, index=names)
    total_imp_drag+=forest_importances_drag*acc_score_drag
    forest_importances_perm_drag = pd.Series(result_drag.importances_mean, index=names)
    total_imp_perm_drag+=forest_importances_perm_drag*acc_score_drag

total_imp_touch_sort=total_imp_touch.sort_values(ascending=False)
total_imp_touch_sort_20=total_imp_touch_sort[:20]
fig, ax = plt.subplots()
ax.figure.set_size_inches(5, 8)
total_imp_touch_sort_20.plot.bar(ax=ax)
ax.set_ylabel('Mean decrease in impurity')
fig.tight_layout()
fig.savefig('/home/user/owncloud/thesis_figures/CY_touch_FI_total_3d.png', dpi=200)
plt.show()

total_imp_perm_touch_sort=total_imp_perm_touch.sort_values(ascending=False)
total_imp_perm_touch_sort_20=total_imp_perm_touch_sort[:20]
fig, ax = plt.subplots()
ax.figure.set_size_inches(5, 8)
total_imp_perm_touch_sort_20.plot.bar(ax=ax)
ax.set_ylabel('Mean decrease in impurity')
fig.tight_layout()
fig.savefig('/home/user/owncloud/thesis_figures/CY_touch_FI_PM_total_3d.png', dpi=200)
plt.show()

total_imp_drag_sort=total_imp_drag.sort_values(ascending=False)
total_imp_drag_sort_20=total_imp_drag_sort[:20]
fig, ax = plt.subplots()
ax.figure.set_size_inches(5, 8)
total_imp_drag_sort_20.plot.bar(ax=ax)
ax.set_ylabel('Mean decrease in impurity')
fig.tight_layout()
fig.savefig('/home/user/owncloud/thesis_figures/CY_drag_FI_total_3d.png', dpi=200)
plt.show()

total_imp_perm_drag_sort=total_imp_perm_drag.sort_values(ascending=False)
total_imp_perm_drag_sort_20=total_imp_perm_drag_sort[:20]
fig, ax = plt.subplots()
ax.figure.set_size_inches(5, 8)
total_imp_perm_drag_sort_20.plot.bar(ax=ax)
ax.set_ylabel('Mean decrease in impurity')
fig.tight_layout()
fig.savefig('/home/user/owncloud/thesis_figures/CY_drag_FI_PM_total_3d.png', dpi=200)
plt.show()


patients=['PB_T2_3_1', 'PB_T2_3_2', 'PB_T2_3_3', 'PB_T2_4_2', 'PB_T2_5_1', 'PB_T2_6_1', 'PB_T2_6_2', 'PB_T3_23_1', 'PB_T3_23_2', 'PB_T3_23_3', 'PB_T3_24_1', 'PB_T3_24_2', 'PB_T3_24_3']
for z in patients:  
    dill.load_session('/home/user/Documents/Master/GW_2D_'+z+'_data.pkl') 
    names=df_big.columns.tolist()
    names=[sub.replace('acceleration', 'acc') for sub in names]
    names=[sub.replace('velocity', 'vel') for sub in names]
    names=[sub.replace('angle', 'ang') for sub in names]
    names=[sub.replace('right', 'r') for sub in names]
    names=[sub.replace('left', 'l') for sub in names]
    names=[sub.replace('top', 't') for sub in names]
    names=[sub.replace('bottom', 'b') for sub in names]
    importances = best_model.feature_importances_
    std = np.std([tree.feature_importances_ for tree in best_model.estimators_], axis=0)
    ranked=np.argsort(importances)[::-1]
    std=pd.Series(std)
    
    std=std.reindex(ranked)
    forest_importances = pd.Series(importances, index=names)
    forest_importances=forest_importances.sort_values(ascending=False)
    forest_importances=forest_importances[:20]
    fig, ax= plt.subplots()
    ax.figure.set_size_inches(5, 8)
    forest_importances.plot.bar(yerr=np.array(std[:20]/10), ax=ax)
    ax.set_ylabel('Mean decrease in impurity')
    plt.tight_layout()
    #plt.savefig('/home/user/owncloud/thesis_figures/GW_FI_'+z+'_2d.png', dpi=200)
    plt.show()
    
    
    forest_importances = pd.Series(result.importances_mean, index=names)
    forest_importances=forest_importances.sort_values(ascending=False)
    forest_importances=forest_importances[:20]
    fig, ax = plt.subplots()
    ax.figure.set_size_inches(5, 8)
    forest_importances.plot.bar(yerr=result.importances_std[:20], ax=ax)
    ax.set_ylabel('Mean accuracy decrease')
    fig.tight_layout()
    #fig.savefig('/home/user/owncloud/thesis_figures/GW_FI_PM_'+z+'_2d.png', dpi=200)
    plt.show()


patients=['PB_T2_3_1', 'PB_T2_3_2', 'PB_T2_3_3', 'PB_T2_4_2', 'PB_T2_5_1', 'PB_T2_6_1', 'PB_T2_6_2', 'PB_T3_23_1', 'PB_T3_23_2', 'PB_T3_23_3', 'PB_T3_24_1', 'PB_T3_24_2', 'PB_T3_24_3']

for z in patients:  
    dill.load_session('/home/user/Documents/Master/GW_2D_'+z+'_data.pkl')
    importances = best_model.feature_importances_
    names=df_big.columns.tolist()
    names=[sub.replace('acceleration', 'acc') for sub in names]
    names=[sub.replace('velocity', 'vel') for sub in names]
    names=[sub.replace('angle', 'ang') for sub in names]
    names=[sub.replace('right', 'r') for sub in names]
    names=[sub.replace('left', 'l') for sub in names]
    names=[sub.replace('top', 't') for sub in names]
    names=[sub.replace('bottom', 'b') for sub in names]
    if z=='PB_T2_3_1':
        total_imp=pd.Series(np.zeros((df_vel_acc.shape[1])), index=names)
        total_imp_perm=pd.Series(np.zeros((df_vel_acc.shape[1])), index=names)
    forest_importances = pd.Series(importances, index=names)
    total_imp+=forest_importances*acc_score
    forest_importances_perm = pd.Series(result.importances_mean, index=names)
    total_imp_perm+=forest_importances_perm*acc_score

total_imp_sort=total_imp.sort_values(ascending=False)
total_imp_sort_20=total_imp_sort[:20]
fig, ax = plt.subplots()
ax.figure.set_size_inches(5, 8)
total_imp_sort_20.plot.bar(ax=ax)
ax.set_ylabel('Mean decrease in impurity')
fig.tight_layout()
#fig.savefig('/home/user/owncloud/thesis_figures/GW_FI_total_2d.png', dpi=200)
plt.show()

total_imp_perm_sort=total_imp_perm.sort_values(ascending=False)
total_imp_perm_sort_20=total_imp_perm_sort[:20]
fig, ax = plt.subplots()
ax.figure.set_size_inches(5, 8)
total_imp_perm_sort_20.plot.bar(ax=ax)
ax.set_ylabel('Mean decrease in impurity')
fig.tight_layout()
#fig.savefig('/home/user/owncloud/thesis_figures/GW_FI_PM_total_2d.png', dpi=200)
plt.show()


patients=['PB_T2_3_1', 'PB_T2_3_2', 'PB_T2_3_3', 'PB_T2_4_2', 'PB_T2_5_1', 'PB_T2_6_1', 'PB_T2_6_2', 'PB_T3_23_1', 'PB_T3_23_2', 'PB_T3_23_3', 'PB_T3_24_1', 'PB_T3_24_2', 'PB_T3_24_3']

z='PB_T3_23_2'
for z in patients:  
    dill.load_session('/home/user/Documents/Master/CY_2D_'+z+'_data.pkl') 
    importances = best_model_touch.feature_importances_
    std = np.std([tree.feature_importances_ for tree in best_model_touch.estimators_], axis=0)
    ranked=np.argsort(importances)[::-1]
    std=pd.Series(std)
    
    std=std.reindex(ranked)
    names=df_big.columns.tolist()
    names=[sub.replace('acceleration', 'acc') for sub in names]
    names=[sub.replace('velocity', 'vel') for sub in names]
    names=[sub.replace('angle', 'ang') for sub in names]
    names=[sub.replace('right', 'r') for sub in names]
    names=[sub.replace('left', 'l') for sub in names]
    names=[sub.replace('top', 't') for sub in names]
    names=[sub.replace('bottom', 'b') for sub in names]
    forest_importances = pd.Series(importances, index=names)
    forest_importances=forest_importances.sort_values(ascending=False)
    forest_importances=forest_importances[:20]
    fig, ax = plt.subplots()
    ax.figure.set_size_inches(5, 8)
    forest_importances.plot.bar(yerr=np.array(std[:20]/10), ax=ax)
    ax.set_ylabel('Mean decrease in impurity')
    fig.tight_layout()
    #plt.savefig('/home/user/owncloud/thesis_figures/CY_touch_FI_'+z+'_2d.png', dpi=200)
    plt.show()
    
    
    forest_importances = pd.Series(result_touch.importances_mean, index=names)
    forest_importances=forest_importances.sort_values(ascending=False)
    forest_importances=forest_importances[:20]
    fig, ax = plt.subplots()
    ax.figure.set_size_inches(5, 8)
    forest_importances.plot.bar(yerr=result_touch.importances_std[:20], ax=ax)
    ax.set_ylabel('Mean accuracy decrease')
    fig.tight_layout()
    #fig.savefig('/home/user/owncloud/thesis_figures/CY_touch_FI_PM_'+z+'_2d.png', dpi=200)
    plt.show()
     ###_____
     
     
    importances = best_model_drag.feature_importances_
    std = np.std([tree.feature_importances_ for tree in best_model_drag.estimators_], axis=0)
    ranked=np.argsort(importances)[::-1]
    std=pd.Series(std)
    
    std=std.reindex(ranked)
    forest_importances = pd.Series(importances, index=names)
    forest_importances=forest_importances.sort_values(ascending=False)
    forest_importances=forest_importances[:20]
    fig, ax = plt.subplots()
    ax.figure.set_size_inches(5, 8)
    forest_importances.plot.bar(yerr=np.array(std[:20]/10), ax=ax)
    ax.set_ylabel('Mean decrease in impurity')
    fig.tight_layout()
    #plt.savefig('/home/user/owncloud/thesis_figures/CY_drag_FI_'+z+'_2d.png', dpi=200)
    plt.show()
    
    
    forest_importances = pd.Series(result_drag.importances_mean, index=names)
    forest_importances=forest_importances.sort_values(ascending=False)
    forest_importances=forest_importances[:20]
    fig, ax = plt.subplots()
    ax.figure.set_size_inches(5, 8)
    forest_importances.plot.bar(yerr=result_drag.importances_std[:20], ax=ax)
    ax.set_ylabel('Mean accuracy decrease')
    fig.tight_layout()
    #fig.savefig('/home/user/owncloud/thesis_figures/CY_drag_FI_PM_'+z+'_2d.png', dpi=200)
    plt.show()



patients=['PB_T2_3_1', 'PB_T2_3_2', 'PB_T2_3_3', 'PB_T2_4_2', 'PB_T2_5_1', 'PB_T2_6_1', 'PB_T2_6_2', 'PB_T3_23_1', 'PB_T3_23_2', 'PB_T3_23_3', 'PB_T3_24_1', 'PB_T3_24_2', 'PB_T3_24_3']

for z in patients:  
    dill.load_session('/home/user/Documents/Master/CY_2D_'+z+'_data.pkl')
    names=df_big.columns.tolist()
    names=[sub.replace('acceleration', 'acc') for sub in names]
    names=[sub.replace('velocity', 'vel') for sub in names]
    names=[sub.replace('angle', 'ang') for sub in names]
    names=[sub.replace('right', 'r') for sub in names]
    names=[sub.replace('left', 'l') for sub in names]
    names=[sub.replace('top', 't') for sub in names]
    names=[sub.replace('bottom', 'b') for sub in names]
    if z=='PB_T2_3_1':
        total_imp_touch=pd.Series(np.zeros((df_vel_acc.shape[1])), index=names)
        total_imp_perm_touch=pd.Series(np.zeros((df_vel_acc.shape[1])), index=names)
        total_imp_drag=pd.Series(np.zeros((df_vel_acc.shape[1])), index=names)
        total_imp_perm_drag=pd.Series(np.zeros((df_vel_acc.shape[1])), index=names)
    importances = best_model_touch.feature_importances_
    forest_importances_touch = pd.Series(importances, index=names)
    total_imp_touch+=forest_importances_touch*acc_score_touch
    forest_importances_perm_touch = pd.Series(result_touch.importances_mean, index=names)
    total_imp_perm_touch+=forest_importances_perm_touch*acc_score_touch
    
    importances = best_model_drag.feature_importances_
    forest_importances_drag = pd.Series(importances, index=names)
    total_imp_drag+=forest_importances_drag*acc_score_drag
    forest_importances_perm_drag = pd.Series(result_drag.importances_mean, index=names)
    total_imp_perm_drag+=forest_importances_perm_drag*acc_score_drag

total_imp_touch_sort=total_imp_touch.sort_values(ascending=False)
total_imp_touch_sort_20=total_imp_touch_sort[:20]
fig, ax = plt.subplots()
ax.figure.set_size_inches(5, 8)
total_imp_touch_sort_20.plot.bar(ax=ax)
ax.set_ylabel('Mean decrease in impurity')
fig.tight_layout()
#fig.savefig('/home/user/owncloud/thesis_figures/CY_touch_FI_total_2d.png', dpi=200)
plt.show()

total_imp_perm_touch_sort=total_imp_perm_touch.sort_values(ascending=False)
total_imp_perm_touch_sort_20=total_imp_perm_touch_sort[:20]
fig, ax = plt.subplots()
ax.figure.set_size_inches(5, 8)
total_imp_perm_touch_sort_20.plot.bar(ax=ax)
ax.set_ylabel('Mean decrease in impurity')
fig.tight_layout()
#fig.savefig('/home/user/owncloud/thesis_figures/CY_touch_FI_PM_total_2d.png', dpi=200)
plt.show()

total_imp_drag_sort=total_imp_drag.sort_values(ascending=False)
total_imp_drag_sort_20=total_imp_drag_sort[:20]
fig, ax = plt.subplots()
ax.figure.set_size_inches(5, 8)
total_imp_drag_sort_20.plot.bar(ax=ax)
ax.set_ylabel('Mean decrease in impurity')
fig.tight_layout()
#fig.savefig('/home/user/owncloud/thesis_figures/CY_drag_FI_total_2d.png', dpi=200)
plt.show()

total_imp_perm_drag_sort=total_imp_perm_drag.sort_values(ascending=False)
total_imp_perm_drag_sort_20=total_imp_perm_drag_sort[:20]
fig, ax = plt.subplots()
ax.figure.set_size_inches(5, 8)
total_imp_perm_drag_sort_20.plot.bar(ax=ax)
ax.set_ylabel('Mean decrease in impurity')
fig.tight_layout()
#fig.savefig('/home/user/owncloud/thesis_figures/CY_drag_FI_PM_total_2d.png', dpi=200)
plt.show()