Heart Attack Dataset Analysis

This is notebook is for analysing the Heart Attack dataset in kaggle (https://www.kaggle.com/rashikrahmanpritom/heart-attack-analysis-prediction-dataset)

This notebook goes through each and every available variable that might contribute to the probability of a heart attack and gives an estimation of the values that might lead to a heart attack for every variable.

#Import libraries import pandas as pd import matplotlib.pyplot as plt import numpy as np import seaborn as sns import sklearn

#reading the csv files o2_filepath='/work/o2Saturation.csv' heart_filepath='/work/heart.csv' o2_data=pd.read_csv(o2_filepath) heart_data=pd.read_csv(heart_filepath)

O2 saturation Data

o2_data.head()

o2_data.columns

o2_data["98.6"].unique()

len(o2_data["98.6"].unique())

We can notice here that there isn't much variation in the levels of O2 saturation, and that there are only 17 unique values in our dataset.

o2_data.mean()

fig, ax1=plt.subplots() plt.figure(figsize=(20,10)) ax1.set_title('box plot') ax1.boxplot(o2_data)

o2_data.value_counts()

o2_data.describe()

Through these previous blocks of code, we can conclude that the standard level of O2 saturation ranges from 96.5 to 98.6

n=0 s=o2_data.squeeze() for i in range (0,len(o2_data['98.6'])): if 96.5<= s[i] <=98.6: n+=1 print(n)

o2_data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 3585 entries, 0 to 3584
Data columns (total 1 columns):
 #   Column  Non-Null Count  Dtype  
---  ------  --------------  -----  
 0   98.6    3585 non-null   float64
dtypes: float64(1)
memory usage: 28.1 KB

100*(3184/3585)

Since the values ranging from 96.5 to 98.6 represent 88.81% of the values then it's safe to assume that usual values are these values, adding that the most common one out of them is 96.5

plt.figure(figsize=(16,7)) sns.histplot(data=o2_data['98.6'], bins=40)

Heart Attack Dataset

Data preparation

#spliting the data into training and test data from sklearn.model_selection import train_test_split train_set, test_set = train_test_split(heart_data, test_size=0.2, random_state=42)

#resetting the indexes train_set=train_set.reset_index(drop=True) test_set=test_set.reset_index(drop=True)

train_set.head()

test_set.head()

train_set.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 242 entries, 0 to 241
Data columns (total 14 columns):
 #   Column    Non-Null Count  Dtype  
---  ------    --------------  -----  
 0   age       242 non-null    int64  
 1   sex       242 non-null    int64  
 2   cp        242 non-null    int64  
 3   trtbps    242 non-null    int64  
 4   chol      242 non-null    int64  
 5   fbs       242 non-null    int64  
 6   restecg   242 non-null    int64  
 7   thalachh  242 non-null    int64  
 8   exng      242 non-null    int64  
 9   oldpeak   242 non-null    float64
 10  slp       242 non-null    int64  
 11  caa       242 non-null    int64  
 12  thall     242 non-null    int64  
 13  output    242 non-null    int64  
dtypes: float64(1), int64(13)
memory usage: 26.6 KB

train_set.describe()

There output reprsents whether the person had a heart-attack (output=1) or not (output=0)

heart_attack=train_set.loc[train_set.output==1] n_heart_attack=train_set.loc[train_set.output==0]

len(heart_attack), len(n_heart_attack)

Data analysis

Sex variable

heart_attack.sex.value_counts(normalize=True)

heart_attack.hist(figsize=(30,20))

heart_data.columns

Age variable

heart_attack.age.hist(figsize=(16,9), bins=30)

def age_category_values(df): m1=round(df.max()/5)*5 m2=round(df.min()/5)*5 L=[i for i in range(m2,m1,5)] dicts={} M=[] for a in range(len(L)): dicts[L[a]]=0 for j in df: for k in L: if j<k: dicts[k]+=1 break for b in dicts: M.append(([b-5,b],dicts[b])) return M

age_category_values(heart_attack.age)

s=0 for i in heart_attack.age: if 40<= i: s+=1 m=len(heart_attack) print(100*s/m)

91.72932330827068

CP variable

The cp column represents the chest pain type

Value 1: typical angina which is defined as substernal chest pain precipitated by physical exertion or emotional stress and relieved with rest or nitroglycerin

Value 2: atypical angina (Unlike typical chest pain, which is usually a dull pain or pressure sensation, atypical chest pain may be sharp, stabbing, or tearing. Atypical chest pain may get worse when breathing in, may get better with leaning forward, and may be worse when you push on the chest)

Value 3: non-anginal pain

Value 4: asymptomatic

heart_attack.cp.value_counts()

heart_attack.cp.hist()

heart_attack.cp.value_counts(normalize=True)

Trtbps variable

heart_attack.trtbps.hist(figsize=(20,8))

def category_values(df, step): m1=round(df.max()/step)*step m2=round(df.min()/step)*step L=[i for i in range(m2,m1+(2*step),step)] dicts={} M=[] for a in range(len(L)): dicts[L[a]]=0 for j in df: for k in L: if j<k: dicts[k]+=1 break for b in dicts: M.append(([b-step,b],dicts[b])) return M

category_values(heart_attack.trtbps,10)

sns.histplot(data=heart_attack,x="trtbps", bins=(80,100,110,120,130,140,150,160,190))

a=0 for i in heart_attack.trtbps: if 110 <= i <=150: a+=1 print(100*(a/m))

82.70676691729322

Since the most of the values of the trtbps index range from 110 to 150(around 82.70%) it is safe to assume that most of heart attacks are associated with a blood pressure between 110 and 150

chol variable

heart_attack.chol.hist(figsize=(16,8), bins=30)

category_values(heart_attack.chol,20)

a=0 for i,j in category_values(heart_attack.chol,20): if i[0]>=180 and i[1]<=320: a+=j print((a/len(heart_attack.chol))*100,"%")

85.71428571428571 %

fbs variable

heart_attack.fbs.hist(figsize=(16,8))

sns.histplot(data=heart_attack, x="fbs", stat="probability")

REST ECG variable

heart_attack.restecg.hist()

heart_attack.restecg.value_counts()

sns.histplot(data=heart_attack, x="restecg", stat="probability")

Thalach variable

heart_attack.thalachh.hist(figsize=(16,8))

category_values(heart_attack.thalachh, 20)

heart_attack.thalachh.max(), heart_attack.thalachh.min()

a=0 for i,j in category_values(heart_attack.thalachh,20): if i[0]>=120 and i[1]<=180: a+=j print((a/len(heart_attack.thalachh))*100,"%")

87.21804511278195 %

exng variable

heart_attack.exng.hist()

sns.histplot(data=heart_attack, x="exng", stat="probability")

slp variable

heart_attack.slp.hist()

sns.histplot(data=heart_attack, x="slp", stat="probability")

caa variable

caa - number of major vessels (0-4) colored by flourosopy

heart_attack.caa.hist()

sns.histplot(data=heart_attack, x="caa", stat="probability")

thall variable

thal - 2 = normal; 1 = fixed defect; 3 = reversable defect

heart_attack.thall.hist()

sns.histplot(data=heart_attack, x="thall", stat="probability")

Most of the individuals who had heart attack (around 80% of them) had a value of 2 for the thall followed by 11% that had a value of 3

Correlation Matrix

corr_matrix = train_set.corr() corr_matrix

plt.figure(figsize=(20,10)) sns.heatmap(corr_matrix, annot=True, cmap=plt.cm.gray)