In [1]:
import pandas as pd 
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
In [3]:
heart_data = pd.read_csv('heart_disease_uci.csv')
In [4]:
heart_data.head()
Out[4]:
id age sex dataset cp trestbps chol fbs restecg thalch exang oldpeak slope ca thal num
0 1 63 Male Cleveland typical angina 145.0 233.0 True lv hypertrophy 150.0 False 2.3 downsloping 0.0 fixed defect 0
1 2 67 Male Cleveland asymptomatic 160.0 286.0 False lv hypertrophy 108.0 True 1.5 flat 3.0 normal 2
2 3 67 Male Cleveland asymptomatic 120.0 229.0 False lv hypertrophy 129.0 True 2.6 flat 2.0 reversable defect 1
3 4 37 Male Cleveland non-anginal 130.0 250.0 False normal 187.0 False 3.5 downsloping 0.0 normal 0
4 5 41 Female Cleveland atypical angina 130.0 204.0 False lv hypertrophy 172.0 False 1.4 upsloping 0.0 normal 0

Data Analysis¶

In [5]:
heart_data.drop(['id','dataset'], axis=1, inplace=True)
heart_data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 920 entries, 0 to 919
Data columns (total 14 columns):
 #   Column    Non-Null Count  Dtype  
---  ------    --------------  -----  
 0   age       920 non-null    int64  
 1   sex       920 non-null    object 
 2   cp        920 non-null    object 
 3   trestbps  861 non-null    float64
 4   chol      890 non-null    float64
 5   fbs       830 non-null    object 
 6   restecg   918 non-null    object 
 7   thalch    865 non-null    float64
 8   exang     865 non-null    object 
 9   oldpeak   858 non-null    float64
 10  slope     611 non-null    object 
 11  ca        309 non-null    float64
 12  thal      434 non-null    object 
 13  num       920 non-null    int64  
dtypes: float64(5), int64(2), object(7)
memory usage: 100.8+ KB
In [6]:
heart_data.describe()
Out[6]:
age trestbps chol thalch oldpeak ca num
count 920.000000 861.000000 890.000000 865.000000 858.000000 309.000000 920.000000
mean 53.510870 132.132404 199.130337 137.545665 0.878788 0.676375 0.995652
std 9.424685 19.066070 110.780810 25.926276 1.091226 0.935653 1.142693
min 28.000000 0.000000 0.000000 60.000000 -2.600000 0.000000 0.000000
25% 47.000000 120.000000 175.000000 120.000000 0.000000 0.000000 0.000000
50% 54.000000 130.000000 223.000000 140.000000 0.500000 0.000000 1.000000
75% 60.000000 140.000000 268.000000 157.000000 1.500000 1.000000 2.000000
max 77.000000 200.000000 603.000000 202.000000 6.200000 3.000000 4.000000
In [7]:
# separateing numeric and categoriecal variables for visualization purpose 
CATEGORICAL_COLS = ['sex', 'cp', 'fbs', 'restecg', 'exang', 'slope', 'thal', 'ca']
NUMERICAL_COLS = ['age', 'trestbps', 'chol', 'thalch', 'oldpeak']

heart_cat = heart_data[CATEGORICAL_COLS]
heart_num = heart_data[NUMERICAL_COLS]

heart_cat.nunique()
Out[7]:
sex        2
cp         4
fbs        2
restecg    3
exang      2
slope      3
thal       3
ca         4
dtype: int64
In [8]:
fig, axes = plt.subplots(2, 4, figsize=(20,10))

sns.countplot(x='sex', data=heart_cat, ax=axes[0,0])
axes[0,0].set_title('Gender Distribution')

sns.countplot(x='cp', data=heart_cat, ax=axes[0,1])
axes[0,1].tick_params(axis='x', rotation=45)
axes[0,1].set_title('Chest Pain Types')

sns.countplot(x='fbs', data=heart_cat, ax=axes[0,2])
axes[0,2].set_title('Fasting Blood Sugar > 120 mg/dl')

sns.countplot(x='restecg', data=heart_cat, ax=axes[0,3])
axes[0,3].set_title('Resting Electrocardiographic Results')

sns.countplot(x='exang', data=heart_cat, ax=axes[1,0])
axes[1,0].set_title('Exercise Induced Angina')

sns.countplot(x='slope', data=heart_cat, ax=axes[1,1])
axes[1,1].set_title('Slope of the Peak Exercise ST Segment')

sns.countplot(x='thal', data=heart_cat, ax=axes[1,2])
axes[1,2].set_title('Defects')

sns.countplot(x='ca', data=heart_cat, ax=axes[1,3])
axes[1,3].set_title('Number of Major Vessels colored by Fluoroscopy')
plt.tight_layout()
plt.show()
In [9]:
# use scatterplots to visualize key relationships in numerical data

fig, axes = plt.subplots(2, 2, figsize=(10,10))

heart_num.plot('age', 'chol', kind='scatter', ax=axes[0,0])
axes[0,0].set_title('Age Against Cholesterol Levels')

heart_num.plot('age', 'trestbps', kind='scatter', ax=axes[0,1])
axes[0,1].set_title('Age Against Resting Blood Pressure')

heart_num.plot('age', 'thalch', kind='scatter', ax=axes[1,0])
axes[1,0].set_title('Age Against Maximum Heart Rate Achieved')

heart_num.plot('age', 'oldpeak', kind='scatter', ax=axes[1,1])
axes[1,1].set_title('Age Against ST Depression')

plt.tight_layout()
plt.show()
In [10]:
fig, axes = plt.subplots(3, figsize=(7,10))

sns.scatterplot(x='chol', y='thalch', hue='num', data=heart_data, ax=axes[0])
axes[0].set_title('Affect of Cholesterol on Maximum Heart Rate')

sns.scatterplot(x='chol', y='thalch', hue='sex', data=heart_data, ax=axes[1])

sns.scatterplot(x='chol', y='thalch', hue='restecg', data=heart_data, ax=axes[2])
plt.show()
In [11]:
sns.scatterplot(x='trestbps', y='thalch', hue='restecg', data=heart_data)
plt.show()
In [12]:
fig, axes = plt.subplots(3, figsize=(7,10))

axes[0].set_title('Affect of Cholesterol on Resting Blood Pressure')
sns.scatterplot(x='chol', y='trestbps', hue='num', data=heart_data, ax=axes[0])
sns.scatterplot(x='chol', y='trestbps', hue='sex', data=heart_data, ax=axes[1])
sns.scatterplot(x='chol', y='trestbps', hue='restecg', data=heart_data, ax=axes[2])

plt.tight_layout()
plt.show()
In [13]:
heart_data.groupby('num').mean()
Out[13]:
age trestbps chol thalch oldpeak ca
num
0 50.547445 129.913043 227.905612 148.800512 0.418205 0.278788
1 53.528302 132.861111 195.255814 131.035714 1.001200 0.741379
2 57.577982 133.613861 143.859813 128.666667 1.353465 1.222222
3 59.214953 136.152174 159.716981 120.500000 1.581319 1.459459
4 59.214286 138.720000 192.148148 127.846154 2.307692 1.692308
In [14]:
print('Average Cholesterol Level Based on Target Variable and Chest Pain Type')
print(pd.crosstab(index=heart_data.num, columns=heart_data.cp, values=heart_data.chol, aggfunc=np.mean))
print('\n')

print('Average Cholesterol Level Based on Target Variable and Patient Gender')
print(pd.crosstab(index=heart_data.num, columns=heart_data.sex, values=heart_data.chol, aggfunc=np.mean))
print('\n')

print('Average Cholesterol Level Based on Target Variable and Cardiographic Results')
print(pd.crosstab(index=heart_data.num, columns=heart_data.restecg, values=heart_data.chol, aggfunc=np.mean))

Average Cholesterol Level Based on Target Variable and Chest Pain Type
cp   asymptomatic  atypical angina  non-anginal  typical angina
num                                                            
0      227.843137       233.957143   222.209677      222.730769
1      193.273684       250.157895   170.756757      215.250000
2      152.321839       123.000000   118.642857       58.500000
3      157.219512       200.000000   152.888889      228.666667
4      196.478261              NaN   146.000000      231.000000


Average Cholesterol Level Based on Target Variable and Patient Gender
sex      Female        Male
num                        
0    248.102190  217.054902
1    221.366667  191.820175
2    216.400000  136.381443
3    216.250000  155.102041
4    316.000000  182.240000


Average Cholesterol Level Based on Target Variable and Cardiographic Results
restecg  lv hypertrophy      normal  st-t abnormality
num                                                  
0            251.768293  227.797619        194.637931
1            216.318182  194.243902        181.395833
2            231.666667  116.629630        132.187500
3            241.230769  130.408163        137.677419
4            238.166667  137.875000        175.285714
In [15]:
# Display correlation matrix and heatmap
corr = heart_data.corr()
print(corr)

sns.heatmap(corr)
plt.show()

               age  trestbps      chol    thalch   oldpeak        ca       num
age       1.000000  0.244253 -0.086234 -0.365778  0.258243  0.370416  0.339596
trestbps  0.244253  1.000000  0.092853 -0.104899  0.161908  0.093705  0.122291
chol     -0.086234  0.092853  1.000000  0.236121  0.047734  0.051606 -0.231547
thalch   -0.365778 -0.104899  0.236121  1.000000 -0.151174 -0.264094 -0.366265
oldpeak   0.258243  0.161908  0.047734 -0.151174  1.000000  0.281817  0.443084
ca        0.370416  0.093705  0.051606 -0.264094  0.281817  1.000000  0.516216
num       0.339596  0.122291 -0.231547 -0.366265  0.443084  0.516216  1.000000
In [16]:
# Display boxplot to visualize outliers in the data

heart_data.boxplot()
plt.show()
In [17]:
heart_data.loc[heart_data['chol']==0,:]
Out[17]:
age sex cp trestbps chol fbs restecg thalch exang oldpeak slope ca thal num
597 32 Male typical angina 95.0 0.0 NaN normal 127.0 False 0.7 upsloping NaN NaN 1
598 34 Male asymptomatic 115.0 0.0 NaN NaN 154.0 False 0.2 upsloping NaN NaN 1
599 35 Male asymptomatic NaN 0.0 NaN normal 130.0 True NaN NaN NaN reversable defect 3
600 36 Male asymptomatic 110.0 0.0 NaN normal 125.0 True 1.0 flat NaN fixed defect 1
601 38 Female asymptomatic 105.0 0.0 NaN normal 166.0 False 2.8 upsloping NaN NaN 2
... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
818 43 Male asymptomatic 122.0 0.0 False normal 120.0 False 0.5 upsloping NaN NaN 1
819 63 Male non-anginal 130.0 0.0 True st-t abnormality 160.0 False 3.0 flat NaN NaN 0
822 48 Male non-anginal 102.0 0.0 NaN st-t abnormality 110.0 True 1.0 downsloping NaN NaN 1
839 56 Male asymptomatic NaN 0.0 False lv hypertrophy NaN NaN NaN NaN NaN NaN 1
840 62 Male non-anginal NaN 0.0 True st-t abnormality NaN NaN NaN NaN NaN NaN 2

172 rows × 14 columns

Data Cleaning¶

In [19]:
heart_data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 920 entries, 0 to 919
Data columns (total 14 columns):
 #   Column    Non-Null Count  Dtype  
---  ------    --------------  -----  
 0   age       920 non-null    int64  
 1   sex       920 non-null    object 
 2   cp        920 non-null    object 
 3   trestbps  861 non-null    float64
 4   chol      890 non-null    float64
 5   fbs       830 non-null    object 
 6   restecg   918 non-null    object 
 7   thalch    865 non-null    float64
 8   exang     865 non-null    object 
 9   oldpeak   858 non-null    float64
 10  slope     611 non-null    object 
 11  ca        309 non-null    float64
 12  thal      434 non-null    object 
 13  num       920 non-null    int64  
dtypes: float64(5), int64(2), object(7)
memory usage: 100.8+ KB
In [20]:
# Cholesterol Levels

median_chol = heart_data.loc[heart_data['chol']!=0, 'chol'].median()
heart_df = heart_data.fillna(value={'chol': median_chol})
heart_df.loc[heart_df['chol']==0, 'chol'] = median_chol
In [21]:
# Resting Blood Pressure

mean_bp = heart_df.loc[heart_df['trestbps']!=0,'trestbps'].mean()
heart_df = heart_df.fillna(value={'trestbps': mean_bp})
heart_df.loc[heart_df['trestbps']==0, 'trestbps'] = mean_bp
In [22]:
# Maximum Heart Rate

mean_hr = heart_df.loc[heart_df['thalch']!=0,'thalch'].mean()
heart_df = heart_df.fillna(value={'thalch': mean_hr})
heart_df.loc[heart_df['thalch']==0, 'thalch'] = mean_hr
In [23]:
# Old Peak

mean_peak = heart_df.oldpeak.mean()
heart_df = heart_df.fillna(value={'oldpeak': mean_peak})
heart_df.loc[heart_df['oldpeak']==0, 'oldpeak'] = mean_peak
In [24]:
# Drop columns with a great number of missing values and reassign datatypes

heart_df.drop(labels=['ca','thal','slope'], axis=1, inplace=True)
heart_df = heart_df.astype({'sex':'category', 'cp':'category', 'fbs':'bool', 'restecg':'category', 'exang':'bool'})

# Drop remaining rows with missing values and display distribution for target variables

heart_df.dropna(inplace=True)
sns.countplot('num', data=heart_df)
plt.show()

C:\Users\Koushik Reddy\anaconda3\lib\site-packages\seaborn\_decorators.py:36: FutureWarning: Pass the following variable as a keyword arg: x. From version 0.12, the only valid positional argument will be `data`, and passing other arguments without an explicit keyword will result in an error or misinterpretation.
  warnings.warn(

Preparing the Data for MOdel Training¶

In [25]:
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import LabelEncoder
In [26]:
# One hot encode the categorical variables and split the target and independent variables
heart_onehot = pd.get_dummies(heart_df, columns=['sex','cp', 'fbs', 'restecg', 'exang'])

X = heart_onehot.drop('num', axis=1)
y = heart_onehot.num

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

y_train.value_counts()
Out[26]:
0    340
1    204
3     86
2     82
4     22
Name: num, dtype: int64
In [27]:
heart_onehot.info()

<class 'pandas.core.frame.DataFrame'>
Int64Index: 918 entries, 0 to 919
Data columns (total 19 columns):
 #   Column                    Non-Null Count  Dtype  
---  ------                    --------------  -----  
 0   age                       918 non-null    int64  
 1   trestbps                  918 non-null    float64
 2   chol                      918 non-null    float64
 3   thalch                    918 non-null    float64
 4   oldpeak                   918 non-null    float64
 5   num                       918 non-null    int64  
 6   sex_Female                918 non-null    uint8  
 7   sex_Male                  918 non-null    uint8  
 8   cp_asymptomatic           918 non-null    uint8  
 9   cp_atypical angina        918 non-null    uint8  
 10  cp_non-anginal            918 non-null    uint8  
 11  cp_typical angina         918 non-null    uint8  
 12  fbs_False                 918 non-null    uint8  
 13  fbs_True                  918 non-null    uint8  
 14  restecg_lv hypertrophy    918 non-null    uint8  
 15  restecg_normal            918 non-null    uint8  
 16  restecg_st-t abnormality  918 non-null    uint8  
 17  exang_False               918 non-null    uint8  
 18  exang_True                918 non-null    uint8  
dtypes: float64(4), int64(2), uint8(13)
memory usage: 61.9 KB

Decision Tree Classifier¶

In [28]:
from sklearn.ensemble import GradientBoostingClassifier, RandomForestClassifier
from sklearn.metrics import classification_report
from sklearn.tree import DecisionTreeClassifier
In [29]:
weights = {0:1, 1:0.5, 2:0.5, 3:0.5, 4:0.5}

clf = DecisionTreeClassifier(criterion='entropy', max_depth=5)
clf = clf.fit(X_train, y_train)
y_pred = clf.predict(X_test)

print(classification_report(y_test, y_pred))

              precision    recall  f1-score   support

           0       0.62      0.92      0.74        71
           1       0.40      0.32      0.36        59
           2       0.40      0.15      0.22        27
           3       0.31      0.19      0.24        21
           4       0.22      0.33      0.27         6

    accuracy                           0.51       184
   macro avg       0.39      0.38      0.36       184
weighted avg       0.47      0.51      0.47       184
In [30]:
# Perform Decision Tree model with class weighting
weights = {0:1, 1:0.5, 2:0.5, 3:0.5, 4:0.5}

clf = DecisionTreeClassifier(criterion='entropy', max_depth=5, class_weight='balanced')
clf = clf.fit(X_train, y_train)
y_pred = clf.predict(X_test)

print(classification_report(y_test, y_pred))

              precision    recall  f1-score   support

           0       0.75      0.68      0.71        71
           1       0.45      0.37      0.41        59
           2       0.29      0.37      0.33        27
           3       0.30      0.33      0.32        21
           4       0.14      0.33      0.20         6

    accuracy                           0.48       184
   macro avg       0.39      0.42      0.39       184
weighted avg       0.52      0.48      0.50       184

Gradient Boosting¶

In [31]:
gradient_booster = GradientBoostingClassifier(learning_rate=0.02, max_depth=3, n_estimators=150)
gradient_booster.fit(X_train, y_train)
y_pred = gradient_booster.predict(X_test)

print(classification_report(y_test, y_pred))

              precision    recall  f1-score   support

           0       0.62      0.89      0.73        71
           1       0.44      0.46      0.45        59
           2       0.33      0.07      0.12        27
           3       0.45      0.24      0.31        21
           4       0.50      0.33      0.40         6

    accuracy                           0.54       184
   macro avg       0.47      0.40      0.40       184
weighted avg       0.50      0.54      0.49       184

Random Forest Classifier¶

In [32]:
clf = RandomForestClassifier(n_estimators=150)
clf.fit(X_train, y_train)
y_pred = clf.predict(X_test)

print(classification_report(y_test, y_pred))

              precision    recall  f1-score   support

           0       0.61      0.87      0.72        71
           1       0.48      0.42      0.45        59
           2       0.23      0.11      0.15        27
           3       0.31      0.24      0.27        21
           4       0.00      0.00      0.00         6

    accuracy                           0.52       184
   macro avg       0.33      0.33      0.32       184
weighted avg       0.46      0.52      0.47       184
In [33]:
clf = RandomForestClassifier(n_estimators=150, class_weight='balanced_subsample')
clf.fit(X_train, y_train)
y_pred = clf.predict(X_test)

print(classification_report(y_test, y_pred))

              precision    recall  f1-score   support

           0       0.59      0.86      0.70        71
           1       0.47      0.37      0.42        59
           2       0.29      0.15      0.20        27
           3       0.30      0.29      0.29        21
           4       0.00      0.00      0.00         6

    accuracy                           0.51       184
   macro avg       0.33      0.33      0.32       184
weighted avg       0.45      0.51      0.47       184


C:\Users\Koushik Reddy\anaconda3\lib\site-packages\sklearn\metrics\_classification.py:1318: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))
C:\Users\Koushik Reddy\anaconda3\lib\site-packages\sklearn\metrics\_classification.py:1318: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))
C:\Users\Koushik Reddy\anaconda3\lib\site-packages\sklearn\metrics\_classification.py:1318: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))

Decision Tree With SMOTE¶

In [34]:
from imblearn.over_sampling import SMOTE
In [35]:
smt = SMOTE(sampling_strategy='not majority')

print('Before', y_train.value_counts())

X_train_SM, y_train_SM = smt.fit_resample(X_train, y_train)

val, counter = np.unique(y_train_SM, return_counts=True)
print('After', (val, counter))

Before 0    340
1    204
3     86
2     82
4     22
Name: num, dtype: int64
After (array([0, 1, 2, 3, 4], dtype=int64), array([340, 340, 340, 340, 340], dtype=int64))
In [36]:
clf = DecisionTreeClassifier(criterion='entropy', max_depth=6)
clf.fit(X_train_SM, y_train_SM)
y_pred = clf.predict(X_test)

print(classification_report(y_test, y_pred))

              precision    recall  f1-score   support

           0       0.72      0.68      0.70        71
           1       0.42      0.31      0.35        59
           2       0.14      0.11      0.12        27
           3       0.20      0.33      0.25        21
           4       0.18      0.50      0.26         6

    accuracy                           0.43       184
   macro avg       0.33      0.39      0.34       184
weighted avg       0.46      0.43      0.44       184
In [ ]: