Heart Disease Classification using Tabnet

 This research aims to develop an artificial intelligence-based system that identifies patients who are more likely to develop heart disease based on their medical history. The heart disease dataset from the UCI Machine Learning Repository was used for training and validation.

In [1]:

import numpy as np # linear algebra
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
import os
import warnings
warnings.filterwarnings("ignore")
from sklearn.preprocessing import MinMaxScaler
from sklearn import metrics
import random

Install Required Libraries¶

In [2]:

%%capture
!pip install -q hvplot
!pip install pytorch-tabnet

Import Libraries¶

In [3]:

import matplotlib.pyplot as plt
import seaborn as sns

from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score, roc_auc_score
import scikitplot as skplt
from sklearn.model_selection import train_test_split
from sklearn.svm import SVC
from xgboost import XGBClassifier
from sklearn.ensemble import RandomForestClassifier,StackingClassifier
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import accuracy_score
from sklearn.ensemble import GradientBoostingClassifier
from sklearn.ensemble import ExtraTreesClassifier
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import AdaBoostClassifier
from sklearn.naive_bayes import GaussianNB
from sklearn.neighbors import KNeighborsClassifier
from sklearn.metrics import confusion_matrix, ConfusionMatrixDisplay
from scipy import stats
from numpy import isnan
from sklearn.impute import KNNImputer


from sklearn.model_selection import GridSearchCV, cross_val_score, StratifiedKFold, learning_curve
import pytorch_tabnet
from pytorch_tabnet.tab_model import TabNetClassifier

import torch
from sklearn.preprocessing import LabelEncoder
from sklearn.metrics import roc_auc_score, accuracy_score
from sklearn.model_selection import KFold
from matplotlib.pyplot import figure

Get Cleveland Data from UCI Repository¶

In [4]:

data=pd.read_csv("https://archive.ics.uci.edu/ml/machine-learning-databases/heart-disease/processed.cleveland.data",header=None)
data = data.replace("?",np.nan)

data = data.dropna().reset_index(drop=True)
data.columns = ['age', 'sex', 'chest pain type', 'resting bp s', 'cholesterol',
       'fasting blood sugar', 'resting ecg', 'max heart rate',
       'exercise angina', 'oldpeak', 'ST slope','ca', 'thal', 'target']

k=['age', 'sex', 'chest pain type', 'resting bp s', 'cholesterol',
       'fasting blood sugar', 'resting ecg', 'max heart rate',
       'exercise angina', 'ST slope','ca', 'thal', 'target']

for j in k:
    data[j] = data[j].astype('float').astype('int') 

data['oldpeak'] = data['oldpeak'].astype('float')
data['target'] = np.where(data.target>0,1,0)
dataTab = data.copy()
data.head()

Out[4]:

	age	sex	chest pain type	resting bp s	cholesterol	fasting blood sugar	resting ecg	max heart rate	exercise angina	oldpeak	ST slope	ca	thal	target
0	63	1	1	145	233	1	2	150	0	2.3	3	0	6	0
1	67	1	4	160	286	0	2	108	1	1.5	2	3	3	1
2	67	1	4	120	229	0	2	129	1	2.6	2	2	7	1
3	37	1	3	130	250	0	0	187	0	3.5	3	0	3	0
4	41	0	2	130	204	0	2	172	0	1.4	1	0	3	0

Convert Nominal Variables - chest pain type, resting ecg and thal¶

In [5]:

CP_Dict = {1:'typical angina',2:'atypical angina',3:'non-anginal',4:'asymptomatic'}
ECG_Dict = {0:'normal',1:'ST-T wave abnormality',2:'left ventricular hypertrophy'}
thal_Dict = {3:'normal',6:'fixed defect',7:'reversable defect'}

data.replace({"chest pain type": CP_Dict},inplace=True)
data.replace({"resting ecg": ECG_Dict},inplace=True)
data.replace({"thal": thal_Dict},inplace=True)

data.head()

Out[5]:

	age	sex	chest pain type	resting bp s	cholesterol	fasting blood sugar	resting ecg	max heart rate	exercise angina	oldpeak	ST slope	ca	thal	target
0	63	1	typical angina	145	233	1	left ventricular hypertrophy	150	0	2.3	3	0	fixed defect	0
1	67	1	asymptomatic	160	286	0	left ventricular hypertrophy	108	1	1.5	2	3	normal	1
2	67	1	asymptomatic	120	229	0	left ventricular hypertrophy	129	1	2.6	2	2	reversable defect	1
3	37	1	non-anginal	130	250	0	normal	187	0	3.5	3	0	normal	0
4	41	0	atypical angina	130	204	0	left ventricular hypertrophy	172	0	1.4	1	0	normal	0

In [6]:

Sex_Dict = {1:'male',0:'female'}
FS_Dict = {0:'under 120mgdl',1:'over 120mgdl'}
exang_Dict = {0:'not induced',1:'induced'}
slope_Dict = {1:'upsloping',2:'flat',3:'downsloping'}

data.replace({"sex": Sex_Dict},inplace=True)
data.replace({"fasting blood sugar": FS_Dict},inplace=True)
data.replace({"exercise angina": exang_Dict},inplace=True)
data.replace({"ST slope": slope_Dict},inplace=True)

dataset = data.copy()
data.head()

Out[6]:

	age	sex	chest pain type	resting bp s	cholesterol	fasting blood sugar	resting ecg	max heart rate	exercise angina	oldpeak	ST slope	ca	thal	target
0	63	male	typical angina	145	233	over 120mgdl	left ventricular hypertrophy	150	not induced	2.3	downsloping	0	fixed defect	0
1	67	male	asymptomatic	160	286	under 120mgdl	left ventricular hypertrophy	108	induced	1.5	flat	3	normal	1
2	67	male	asymptomatic	120	229	under 120mgdl	left ventricular hypertrophy	129	induced	2.6	flat	2	reversable defect	1
3	37	male	non-anginal	130	250	under 120mgdl	normal	187	not induced	3.5	downsloping	0	normal	0
4	41	female	atypical angina	130	204	under 120mgdl	left ventricular hypertrophy	172	not induced	1.4	upsloping	0	normal	0

In [7]:

data['target'].value_counts(dropna=False)

Out[7]:

0    160
1    137
Name: target, dtype: int64

Exploratory Data Analysis¶

Target Variables¶

In [8]:

f, axes = plt.subplots(1, 1, figsize=(4, 6))
sns.countplot(ax=axes,x='target', data=data, palette=['green','orange'])
axes.set_title("Target Distribution", fontsize=20)

Out[8]:

Text(0.5, 1.0, 'Target Distribution')

Categorical Binaries - Sex, FBS and Exang¶

In [9]:

f, axes = plt.subplots(1, 3, figsize=(15, 5))

sns.countplot(ax=axes[0],x='sex', data=data, palette=['green','orange'],hue="target")
axes[0].set_title("sex", fontsize=20)

sns.countplot(ax=axes[1],x='fasting blood sugar', data=data, palette=['green','orange'],hue="target")
axes[1].set_title("fasting blood sugar", fontsize=20)

sns.countplot(ax=axes[2],x='exercise angina', data=data, palette=['green','orange'],hue="target")
plt.title("exercise angina", fontsize=20)

Out[9]:

Text(0.5, 1.0, 'exercise angina')

Categorical Variables - CP, ECG and thal¶

In [10]:

plt.figure(figsize=(12,5))
sns.countplot(x='chest pain type', data=data, palette=['green','orange'],hue="target")
plt.title("chest pain type", fontsize=20)

Out[10]:

Text(0.5, 1.0, 'chest pain type')

In [11]:

plt.figure(figsize=(12,5))
sns.countplot(x='ST slope', data=data, palette=['green','orange'],hue="target")
plt.title("ST slope", fontsize=20)

Out[11]:

Text(0.5, 1.0, 'ST slope')

In [12]:

plt.figure(figsize=(12,5))
ax=sns.countplot(x='resting ecg', data=data, palette=['green','orange'],hue="target")


plt.title("resting ecg", fontsize=20)

Out[12]:

Text(0.5, 1.0, 'resting ecg')

In [13]:

plt.figure(figsize=(12,5))
sns.countplot(x='ca', data=data, palette=['green','orange'],hue="target")
plt.title("ca", fontsize=20)

Out[13]:

Text(0.5, 1.0, 'ca')

In [14]:

plt.figure(figsize=(12,5))
sns.countplot(x='thal', data=data, palette=['green','orange'],hue="target")
plt.title("thal", fontsize=20)

Out[14]:

Text(0.5, 1.0, 'thal')

Numeric Variables - Age, Cholestrol,Resting BP and Max heart rate¶

In [15]:

data_disease = data[data["target"] == 1]
data_normal = data[data["target"] == 0]

Age¶

In [16]:

plt.figure(figsize=(8,5))
sns.distplot(data_normal["age"], bins=24, color='g')
sns.distplot(data_disease["age"], bins=24, color='r')
plt.title("Distribuition and density by Age",fontsize=20)
plt.xlabel("Age",fontsize=15)
plt.show()

Cholestrol¶

In [17]:

#figure size
plt.figure(figsize=(8,5))
sns.distplot(data_normal["cholesterol"], bins=24, color='g')
sns.distplot(data_disease["cholesterol"], bins=24, color='r')
plt.title("Distribuition and density by cholesterol",fontsize=20)
plt.xlabel("cholesterol",fontsize=15)
plt.show()

Resting BP¶

In [18]:

plt.figure(figsize=(8,5))
sns.distplot(data_normal["resting bp s"], bins=24, color='g')
sns.distplot(data_disease["resting bp s"], bins=24, color='r')
plt.title("Distribuition and density by resting bp",fontsize=20)
plt.xlabel("resting bp",fontsize=15)
plt.show()

Max Heart Rate¶

In [19]:

plt.figure(figsize=(8,5))
sns.distplot(data_normal["max heart rate"], bins=24, color='g')
sns.distplot(data_disease["max heart rate"], bins=24, color='r')
plt.title("Distribuition and density by max heart rate",fontsize=20)
plt.xlabel("max heart rate",fontsize=15)
plt.show()

Oldpeak¶

In [20]:

plt.figure(figsize=(8,5))
sns.distplot(data_normal["oldpeak"], bins=24, color='g')
sns.distplot(data_disease["oldpeak"], bins=24, color='r')
plt.title("Distribuition and density by old peak",fontsize=20)
plt.xlabel("oldpeak",fontsize=15)
plt.show()

Age vs Max HR¶

In [21]:

plt.figure(figsize=(9, 7))
plt.scatter(data_disease["age"],
            data_disease["max heart rate"],
            c="salmon")
plt.scatter(data_normal["age"],
            data_normal["max heart rate"],
            c="lightblue")
plt.title("Heart Disease in function of Age and Max Heart Rate")
plt.xlabel("Age")
plt.ylabel("Max Heart Rate")
plt.legend(["Disease", "No Disease"]);

Correlations¶

In [22]:

import hvplot.pandas
data.drop('target', axis=1).corrwith(data.target).hvplot.barh(
    width=600, height=400, 
    title="Correlation between Heart Disease and Numeric Features", 
    ylabel='Correlation', xlabel='Numerical Features'
)

Out[22]:

In [23]:

sns.set(rc = {'figure.figsize':(12,12)})
sns.heatmap(data.corr(), annot = True, fmt='.2g',cmap= 'coolwarm')

Out[23]:

<AxesSubplot:>

Test Train Split¶

In [24]:

from sklearn.preprocessing import LabelEncoder
X = dataTab.drop('target', axis = 1)
feats = X.columns

categorical_columns = ['sex','resting ecg', 'chest pain type','fasting blood sugar' ,
              'exercise angina','ST slope','thal']

categorical_dims =  {}
for col in categorical_columns:
    print(col, X[col].nunique())
    l_enc = LabelEncoder()
    X[col] = l_enc.fit_transform(X[col].values)
    categorical_dims[col] = len(l_enc.classes_)
    
cat_idxs = [ i for i, f in enumerate(feats) if f in categorical_columns]
cat_dims = [ categorical_dims[f] for i, f in enumerate(feats) if f in categorical_columns]
    
X = np.array(X) 
target = np.array(dataTab['target'])
    
x_train, x_val, y_train, y_val = train_test_split(X, target, test_size=0.3, random_state=42)
x_val, x_test, y_val, y_test = train_test_split(x_val, y_val, test_size=0.4, random_state=42)

sex 2
resting ecg 3
chest pain type 4
fasting blood sugar 2
exercise angina 2
ST slope 3
thal 3

Modeling¶

Base Models¶

• LogisticRegression
• RandomForestClassifier
• XGBClassifier
• GradientBoostingClassifier

In [25]:

LR = LogisticRegression(solver='liblinear',random_state=42)
RF = RandomForestClassifier(criterion = 'entropy',random_state=42)
XGB = XGBClassifier(max_depth=5,random_state=42)
GBM = GradientBoostingClassifier(learning_rate = 0.01,random_state=42)

In [26]:

models = [LR,RF,XGB,GBM]
metric_list = []
for m in models:
    mName = type(m).__name__
    m.fit(x_train,y_train)
    
    y_pred = m.predict(x_test)
    
    auroc = np.round(roc_auc_score(y_test, y_pred),4)
    accuracy = np.round(accuracy_score(y_test, y_pred),4)
    precision = np.round(precision_score(y_test, y_pred),4)
    recall = np.round(recall_score(y_test, y_pred),4)
    f1 = np.round(f1_score(y_test, y_pred),4)
    globals()[f"y_pred_{mName}"] = y_pred
    
    
    l = [mName,auroc,accuracy,precision,recall,f1]
    metric_list.append(l)
    print(mName,":",auroc,accuracy,precision,recall,f1)

LogisticRegression : 0.9 0.9167 1.0 0.8 0.8889
RandomForestClassifier : 0.8762 0.8889 0.9231 0.8 0.8571
XGBClassifier : 0.8429 0.8611 0.9167 0.7333 0.8148
GradientBoostingClassifier : 0.7857 0.8056 0.8333 0.6667 0.7407

In [27]:

df_metric_list = pd.DataFrame(metric_list)
df_metric_list.columns = ['modelName','auroc','accuracy','precision','recall','f1_score']
df_metric_list = df_metric_list.sort_values(["accuracy","auroc"],ascending=False).reset_index(drop=True)
df_metric_list

Out[27]:

	modelName	auroc	accuracy	precision	recall	f1_score
0	LogisticRegression	0.9000	0.9167	1.0000	0.8000	0.8889
1	RandomForestClassifier	0.8762	0.8889	0.9231	0.8000	0.8571
2	XGBClassifier	0.8429	0.8611	0.9167	0.7333	0.8148
3	GradientBoostingClassifier	0.7857	0.8056	0.8333	0.6667	0.7407

Confusion Matrix for Base Models¶

In [28]:

font = {'family' : 'normal',
    'weight' : 'bold',
    'size'   : 18}
plt.rc('font', **font)

f, axes = plt.subplots(2, 2, figsize=(8,9))
i=0
axes = axes.ravel()
f.suptitle("Confusion Matrix Base Models", fontsize=20, fontweight='bold')
for m in models:
    mName = type(m).__name__
    disp = ConfusionMatrixDisplay(confusion_matrix(y_test,globals()[f"y_pred_{mName}"]))
    disp.plot(ax=axes[i], values_format='.20g')
    axes[i].grid(False)
    disp.ax_.set_title(mName,fontweight='bold',fontsize=12)
    disp.im_.colorbar.remove()
    i=i+1
    
plt.show()

Tabular Net (TABNET)¶

In [29]:

def set_seed(seed: int = 42):
    np.random.seed(seed)
    random.seed(seed)
    torch.manual_seed(seed)
    torch.cuda.manual_seed(seed)
    torch.backends.cudnn.deterministic = True
    torch.backends.cudnn.benchmark = False
    os.environ["PYTHONHASHSEED"] = str(seed)
set_seed()

In [30]:

tabnet = TabNetClassifier(optimizer_fn=torch.optim.Adam,
                           optimizer_params=dict(lr=0.01),
                           scheduler_params={"step_size":100,
                                             "gamma":0.95},
                           scheduler_fn=torch.optim.lr_scheduler.StepLR,
                           verbose=0,
                           mask_type='sparsemax',
                           cat_dims=cat_dims,cat_idxs=cat_idxs
                          )

tabnet.fit(
             x_train,y_train,
            eval_set=[(x_train, y_train), (x_val, y_val)],
            eval_name=['train', 'valid'],
            eval_metric=['auc','accuracy'],
            max_epochs=1000 , patience=50,
                batch_size=16,virtual_batch_size=16,
            num_workers=0,
            drop_last=False
        )


y_pred = tabnet.predict(x_test)
test_acc = accuracy_score(y_pred, y_test)


preds_valid = tabnet.predict(x_val)
valid_acc = accuracy_score(preds_valid, y_val)

print("valid_acc:",valid_acc,", test_acc:",test_acc)

Early stopping occurred at epoch 126 with best_epoch = 76 and best_valid_accuracy = 0.85185
valid_acc: 0.8518518518518519 , test_acc: 0.9444444444444444

Tabular Nets Plots¶

In [31]:

fig = plt.figure(figsize=(10, 5))
fig.suptitle("Tabnet Training Loss",fontsize=20)
plt.plot(tabnet.history['loss'])

Out[31]:

[<matplotlib.lines.Line2D at 0x7f56006777d0>]

In [32]:

fig = plt.figure(figsize=(10, 5))
fig.suptitle("Tabnet Train and Valid Accuracy",fontsize=20)
plt.plot(tabnet.history['train_accuracy'])
plt.plot(tabnet.history['valid_accuracy'])

Out[32]:

[<matplotlib.lines.Line2D at 0x7f56005f5190>]

Tabnet Metrics¶

In [33]:

metric_list=[]
auroc = np.round(roc_auc_score(y_test, y_pred),4)
accuracy = np.round(accuracy_score(y_test, y_pred),4)
precision = np.round(precision_score(y_test, y_pred),4)
recall = np.round(recall_score(y_test, y_pred),4)
f1 = np.round(f1_score(y_test, y_pred),4)

l = ['Tabular Net',auroc,accuracy,precision,recall,f1]
metric_list.append(l)

df_metric_list = pd.DataFrame(metric_list)
df_metric_list.columns = ['modelName','auroc','accuracy','precision','recall','f1_score']
df_metric_list = df_metric_list.sort_values(["accuracy","auroc"],ascending=False).reset_index(drop=True)
df_metric_list

Out[33]:

	modelName	auroc	accuracy	precision	recall	f1_score
0	Tabular Net	0.9429	0.9444	0.9333	0.9333	0.9333

Confusion Matrix for Tabular Net¶

In [34]:

font = {'family' : 'normal',
    'weight' : 'bold',
    'size'   : 18}
plt.rc('font', **font)

f, axes = plt.subplots(1, 1, figsize=(5, 5))
i=0
f.suptitle("Confusion Matrix For Tab Net", fontsize=20, fontweight='bold')
for m in models:
    disp = ConfusionMatrixDisplay(confusion_matrix(y_test,y_pred))
    disp.plot(ax=axes, values_format='.20g')
    axes.grid(False)
    disp.im_.colorbar.remove()
    i=i+1
    
plt.show()

Feature Importances¶

In [35]:

feat_importances = pd.Series(tabnet.feature_importances_, index=feats)
feat_importances.nlargest(15).plot(kind='barh')

Out[35]:

<AxesSubplot:>

In [36]:

fig = plt.figure(figsize=(5, 7))
importances = tabnet.feature_importances_
indices = np.argsort(importances)[4:]
plt.title('Feature Importances',fontweight='bold',fontsize=22)
plt.barh(range(len(indices)), importances[indices], color='g', align='center')
plt.yticks(range(len(indices)), [feats[i] for i in indices],
           fontweight='bold',fontsize=12)
plt.xlabel('Relative Importance')
plt.show()

Feature Explainability¶

By using the masks, we can understand which features are being used at a prediction level

In [37]:

font = {'family' : 'normal',
    'weight' : 'bold',
    'size'   : 18}
plt.rc('font', **font)

explain_matrix, masks = tabnet.explain(x_test)
f, axes = plt.subplots(1, 3, figsize=(12,6))
axes = axes.ravel()
f.suptitle("Masks", fontsize=20, fontweight='bold')
for i in range(3):
    axes[i].imshow(masks[i])
    axes[i].set_title(f"mask {i}")
    
plt.show()

In [38]:

masksum = masks[0]+masks[1]+masks[2]
masksumdf = pd.DataFrame(masksum)
masksumdf.columns = feats
masksumdf

Out[38]:

	age	sex	chest pain type	resting bp s	cholesterol	fasting blood sugar	resting ecg	max heart rate	exercise angina	oldpeak	ST slope	ca	thal
0	0.421180	0.000000	0.000000	0.000000	0.0	0.239193	0.254920	0.084707	0.000000	1.000000	0.000000	1.000000	0.000000
1	0.000000	0.032162	0.000000	1.000000	0.0	0.000000	0.000000	0.410161	0.000000	0.557677	0.000000	1.000000	0.000000
2	0.000000	0.000000	0.000000	0.977114	0.0	0.000000	0.000000	0.468335	0.000000	0.531665	0.000000	0.022886	1.000000
3	0.272394	0.031191	0.000000	0.000000	0.0	0.288377	0.000000	0.054172	0.000000	1.020705	0.000000	0.893932	0.439229
4	0.000000	0.866500	0.000000	0.000000	0.0	0.000000	0.564887	0.000000	0.042170	0.000000	0.000000	1.000000	0.526442
5	0.000000	0.000059	0.000000	0.999941	0.0	0.000000	0.000000	0.430287	0.000000	0.569713	0.000000	0.000000	1.000000
6	0.000000	0.000000	0.000000	0.000000	0.0	0.925404	0.000000	0.000000	0.000000	1.000000	0.000000	1.000000	0.074596
7	0.669275	0.000000	0.000000	0.000000	0.0	0.176432	0.063401	0.090891	0.000000	0.815062	0.000000	1.184938	0.000000
8	0.000000	0.277727	0.000000	0.467442	0.0	0.254831	0.000000	0.477747	0.000000	0.522253	0.000000	1.000000	0.000000
9	0.091034	0.893204	0.000000	0.000000	0.0	0.000000	0.486976	0.000000	0.000000	0.000000	0.000000	1.000000	0.528785
10	0.000000	0.052177	0.000000	0.656777	0.0	0.000000	0.000000	0.540997	0.000000	0.459003	0.000000	0.291047	1.000000
11	0.000000	0.000000	0.001158	0.000000	0.0	0.884686	0.000000	0.000000	0.000000	0.000000	0.000000	1.000000	1.114155
12	0.180344	0.469384	0.000000	0.000000	0.0	0.255227	0.244604	0.155433	0.000000	0.000000	0.000000	1.419348	0.275661
13	0.274975	0.000000	0.000000	0.000000	0.0	0.405568	0.030806	0.000000	0.000000	0.000000	0.000000	1.016979	1.271672
14	0.181538	0.301859	0.000000	0.000000	0.0	0.109740	0.505199	0.147025	0.000000	0.610345	0.000000	1.000000	0.144294
15	0.251564	0.867052	0.000000	0.000000	0.0	0.029788	0.326736	0.000000	0.000000	0.000000	0.000000	1.000000	0.524859
16	0.000000	0.515716	0.000000	0.541488	0.0	0.000000	0.000000	0.282020	0.000000	0.613046	0.000000	1.047730	0.000000
17	0.225985	0.000000	0.000000	0.000000	0.0	0.547413	0.000000	0.000000	0.000000	0.000000	0.000000	1.000000	1.226603
18	0.000000	0.000000	0.000000	0.000000	0.0	0.316066	0.287177	0.304699	0.084341	0.000000	0.000000	1.000000	1.007718
19	0.000000	0.707483	0.000000	0.000000	0.0	0.463950	0.000000	0.000000	0.000000	0.059420	0.000000	1.233096	0.536050
20	0.515521	0.028114	0.000000	0.000000	0.0	0.279003	0.131515	0.000000	0.000000	0.420060	0.000000	0.697103	0.928685
21	0.000000	0.753129	0.000000	0.136883	0.0	0.000000	0.229975	0.000000	0.000000	0.374388	0.000000	1.453410	0.052215
22	0.329486	0.006196	0.000000	0.000000	0.0	0.000000	0.114499	0.010178	0.000000	0.914524	0.000000	0.606188	1.018930
23	0.000000	0.000000	0.000000	0.000000	0.0	1.000000	0.000000	0.000000	0.000000	0.000000	0.000000	1.000000	1.000000
24	0.000000	0.000000	0.000000	0.000000	0.0	0.874469	0.000000	0.125531	0.000000	0.000000	0.000000	1.000000	1.000000
25	0.000000	0.000000	0.000000	0.000000	0.0	0.700572	0.011820	0.287608	0.000000	0.000000	0.000000	1.000000	1.000000
26	0.000000	0.000000	0.000000	1.000000	0.0	0.000000	0.000000	0.505667	0.000000	0.494333	0.000000	1.000000	0.000000
27	0.000000	0.000000	0.000000	0.000000	0.0	0.637512	0.000000	0.362488	0.000000	0.000000	0.000000	1.000000	1.000000
28	0.138939	0.597025	0.000000	0.000000	0.0	0.070083	0.552135	0.000000	0.000000	0.402975	0.000000	1.000000	0.238843
29	0.000000	0.000000	0.000000	0.000000	0.0	0.955776	0.000000	0.000000	0.000000	0.000000	0.044224	1.000000	1.000000
30	0.000000	0.624539	0.000000	0.100942	0.0	0.000000	0.000000	0.000000	0.098922	0.000000	0.000000	1.632383	0.543213
31	0.000000	0.000000	0.000000	1.000000	0.0	0.000000	0.000000	0.669707	0.000000	0.330293	0.000000	0.000000	1.000000
32	0.000000	0.000000	0.013528	0.000000	0.0	0.986472	0.000000	0.000000	0.000000	1.000000	0.000000	1.000000	0.000000
33	0.000000	0.385101	0.000000	0.645728	0.0	0.000000	0.000000	0.416254	0.000000	0.480495	0.000000	0.072423	1.000000
34	0.000000	0.000000	0.000000	1.000000	0.0	0.000000	0.000000	0.669707	0.000000	0.330293	0.000000	0.000000	1.000000
35	0.261785	0.000000	0.000000	0.000000	0.0	0.385700	0.000000	0.000000	0.000000	0.000000	0.000000	1.062956	1.289559

In [39]:

font = {'family' : 'normal',
    'weight' : 'bold',
    'size'   : 18}

plt.figure(figsize=(8,6))
plt.rc('font', **font)
sns.heatmap(masksumdf,cbar=True)
plt.rc('font', **font)
plt.title('Features used for prediction',fontweight='bold',fontsize=16)

Out[39]:

Text(0.5, 1.0, 'Features used for prediction')