Open in Colab

IMPORTING LIBRARIES

import pandas as pd
import numpy as np
import os
from natsort import natsorted
import joblib

WORKING ON COMBING MULTIPLE LEAD FILES

#creating list to store file_names

NORMAL_=[]
MI_=[]
PMI_=[]
HB_=[]

normal = '/content/drive/MyDrive/CMPE255_PROJECT/NORMAL'
abnormal = '/content/drive/MyDrive/CMPE255_PROJECT/AHB'
MI = '/content/drive/MyDrive/CMPE255_PROJECT/MI'
MI_history = '/content/drive/MyDrive/CMPE255_PROJECT/PM'

Types_ECG = {'normal':normal,'Abnormal_hear_beat':abnormal,'MI':MI,'History_MI':MI_histo

for types,folder in Types_ECG.items():

for files in os.listdir(folder):
if types=='normal':
NORMAL_.append(files)
elif types=='Abnormal_hear_beat':
HB_.append(files)
elif types=='MI':
MI_.append(files)
elif types=='History_MI':
PMI_.append(files)

NORMAL_ = natsorted(NORMAL_)
NORMAL_

['scaled_data_1D_1.csv',
'scaled_data_1D_2.csv',
'scaled_data_1D_3.csv',
'scaled_data_1D_4.csv',
'scaled_data_1D_5.csv',
'scaled_data_1D_6.csv',
'scaled_data_1D_7.csv',
'scaled_data_1D_8.csv',
'scaled_data_1D_9.csv',
'scaled_data_1D_10.csv',
'scaled_data_1D_11.csv',
'scaled_data_1D_12.csv',
'scaled_data_1D_13.csv']

D bl li k ( ) di
 Double-click (or enter) to edit

Double-click (or enter) to edit

MI_ = natsorted(MI_)
MI_

['scaled_data_1D_1.csv',
'scaled_data_1D_2.csv',
'scaled_data_1D_3.csv',
'scaled_data_1D_4.csv',
'scaled_data_1D_5.csv',
'scaled_data_1D_6.csv',
'scaled_data_1D_7.csv',
'scaled_data_1D_8.csv',
'scaled_data_1D_9.csv',
'scaled_data_1D_10.csv',
'scaled_data_1D_11.csv',
'scaled_data_1D_12.csv',
'scaled_data_1D_13.csv']

PMI_ = natsorted(PMI_)
PMI_

['scaled_data_1D_1.csv',
'scaled_data_1D_2.csv',
'scaled_data_1D_3.csv',
'scaled_data_1D_4.csv',
'scaled_data_1D_5.csv',
'scaled_data_1D_6.csv',
'scaled_data_1D_7.csv',
'scaled_data_1D_8.csv',
'scaled_data_1D_9.csv',
'scaled_data_1D_10.csv',
'scaled_data_1D_11.csv',
'scaled_data_1D_12.csv',
'scaled_data_1D_13.csv']

HB_ = natsorted(HB_)
HB_

['scaled_data_1D_1.csv',
'scaled_data_1D_2.csv',
'scaled_data_1D_3.csv',
'scaled_data_1D_4.csv',
'scaled_data_1D_5.csv',
'scaled_data_1D_6.csv',
'scaled_data_1D_7.csv',
'scaled_data_1D_8.csv',
'scaled_data_1D_9.csv',
'scaled_data_1D_10.csv',
'scaled_data_1D_11.csv',
'scaled_data_1D_12.csv',
'scaled_data_1D_13.csv']
 COMBINED CSV OF EACH LEAD(1-12) FROM ALL IMAGES

#loop over and create combined csv files for each leads.
for x in range(len(MI_)):
df1=pd.read_csv('/content/drive/MyDrive/CMPE255_PROJECT/NORMAL/{}'.format(NORMAL_[x]))
df2=pd.read_csv('/content/drive/MyDrive/CMPE255_PROJECT/AHB/{}'.format(HB_[x]))
df3=pd.read_csv('/content/drive/MyDrive/CMPE255_PROJECT/MI/{}'.format(MI_[x]))
df4=pd.read_csv('/content/drive/MyDrive/CMPE255_PROJECT/PM/{}'.format(PMI_[x]))
final_df = pd.concat([df1,df2,df3,df4],ignore_index=True)
final_df.to_csv('/content/drive/MyDrive/CMPE255_PROJECT/Combined_IDLead_{}.csv'.format

#now reading just lead1

df=pd.read_csv('/content/drive/MyDrive/CMPE255_PROJECT/Combined_IDLead_1.csv')
df['Target'].unique()

array(['No', 'HB', 'MI', 'PM'], dtype=object)

df.drop(columns=['Unnamed: 0'],inplace=True)

#convert Target column values as Numeric using ngroups

encode_target_label = df.groupby('Target').ngroup().rename("target").to_frame()
test_final = df.merge(encode_target_label, left_index=True, right_index=True)
test_final.drop(columns=['Target'],inplace=True)
test_final

0 1 2 3 4 5 6 7

0 0.728449 0.680755 0.619010 0.645367 0.681570 0.732488 0.758448 0.750660 0.7

1 0.957972 0.950695 0.941024 0.930501 0.913601 0.892244 0.868016 0.855127 0.8

2 0.611084 0.661575 0.695790 0.741113 0.716666 0.595794 0.425022 0.286457 0.4

3 0.839213 0.861690 0.866457 0.865756 0.855027 0.855606 0.845561 0.843187 0.8

4 0.917753 0.924369 0.873765 0.791381 0.699513 0.604927 0.500312 0.446012 0.5

... ... ... ... ... ... ... ... ...

923 0.874246 0.877014 0.864280 0.860505 0.871349 0.912404 0.958148 0.977826 0.9

924 0.829815 0.832084 0.852396 0.909665 0.988242 1.000000 0.923323 0.821865 0.7

925 0.469048 0.417983 0.362322 0.351995 0.391493 0.418305 0.440135 0.444598 0.4

926 0.682510 0.682286 0.641051 0.620212 0.608210 0.576331 0.603596 0.645714 0.6

927 0.792175 0.815695 0.819518 0.820559 0.847985 0.880933 0.902061 0.878266 0.8

928 rows × 256 columns

PERFORM DIMENSIONALITY REDUCTION JUST FOR CHECKING/UNDERSTANDING

#just for testing

# Now Perform Dimensionality reduction (PCA) on that Dataframe and check
from sklearn.decomposition import PCA

#do PCA and choose componeents as 100

pca = PCA(n_components=100)
x_pca = pca.fit_transform(test_final.iloc[:,0:-1])
x_pca = pd.DataFrame(x_pca)

# Calculate the variance explained by priciple components

explained_variance = pca.explained_variance_ratio_
print('Variance of each component:', pca.explained_variance_ratio_)
print('\n Total Variance Explained:', round(sum(list(pca.explained_variance_ratio_))*100

#store the new pca generated dimensions in a dataframe

pca_df = pd.DataFrame(data = x_pca)
target = pd.Series(test_final['target'], name='target')
result_df = pd.concat([pca_df, target], axis=1)
result_df
 Variance of each component: [1.76145888e-01 9.50265614e-02 6.99060614e-02 6.1
5.34876630e-02 4.23664893e-02 3.68320213e-02 3.38541791e-02
3.00884979e-02 2.90396728e-02 2.64962509e-02 2.42272738e-02
2.10221030e-02 1.99751559e-02 1.77321042e-02 1.63016802e-02
1.53898622e-02 1.48412074e-02 1.33644825e-02 1.19674074e-02
1.16813409e-02 1.05807650e-02 9.68875480e-03 9.47385060e-03
8.65347748e-03 8.47506998e-03 7.93382172e-03 7.30163338e-03
6.76380665e-03 6.36886390e-03 6.02004791e-03 5.46823032e-03
3 2299 03 4 9 82 89 03 4 4686092 03 4 4608 684 03
result_df

0 1 2 3 4 5 6 7

0 1.018578 1.148263 -0.589582 0.193617 0.047950 -0.309400 -0.161566 0.478471

1 -1.098692 0.289832 -1.766388 1.076165 -0.261201 -0.820446 -0.474188 -0.515238

2 0.275021 -0.451289 0.106750 -0.426415 0.066133 0.692474 0.634894 -0.035867

3 -1.517085 1.662693 -1.021167 0.804267 -0.281985 0.518180 0.355748 -0.344235

4 -0.152840 -1.046283 0.351278 1.100381 -1.613642 1.484188 -0.113277 -0.251152

... ... ... ... ... ... ... ... ...

923 -1.321884 2.153021 0.788596 -1.304253 0.458186 -0.859346 -0.069127 -0.392796

924 -0.867163 -0.040504 0.940680 0.302648 -0.469672 -0.368255 1.065579 0.801522

925 3.753012 0.841636 -0.317393 -0.296117 0.593769 -0.255474 -0.057091 -0.072048

926 0.603083 0.126259 0.003433 0.283612 0.169559 -0.156326 -0.068399 -0.184308

927 -1.452945 1.233599 0.439472 0.278517 0.165928 -0.171830 -0.075000 0.033859

928 rows × 101 columns

TRYING DIFFERENT ML MODELS ON A SINGLE LEAD(EX : 1) POST

DIMENSIONALITY REDUCTION

KNN

# Import the necessary modules for ML model

from sklearn.pipeline import Pipeline
from sklearn.neighbors import KNeighborsClassifier
from sklearn.model_selection import train_test_split,GridSearchCV
from sklearn.metrics import confusion_matrix, classification_report

# Setup the pipeline steps:

steps = [('knn', KNeighborsClassifier())]

# Create the pipeline: pipeline

pipeline = Pipeline(steps)

# have paased less range value of hyperparamter since i'm using free tier version of goo
k_range = list(range(1, 9))
parameters = dict(knn__n_neighbors=k_range)

#input
X = result_df.iloc[:,0:-1]

#target
y=result_df.iloc[:,-1]

# Create train and test sets

X_train, X_test, y_train, y_test = train_test_split(X,y,test_size=0.4,random_state=42)

#increasing cv score takes lot of time in gooogle colab, so kept it just 2.

cv = GridSearchCV(pipeline,parameters,cv=2)

cv.fit(X_train,y_train)

# Predict the labels of the test set: y_pred

y_pred = cv.predict(X_test)

Knn_Accuracy = cv.score(X_test, y_test)

# Compute and print metrics

print("Accuracy: {}".format(Knn_Accuracy))
print(classification_report(y_test, y_pred))
print("Tuned Model Parameters: {}".format(cv.best_params_))

Accuracy: 0.782258064516129
precision recall f1-score support

0 0.87 0.63 0.73 105

1 0.91 0.91 0.91 94
2 0.72 0.88 0.79 112
3 0.63 0.67 0.65 61

accuracy 0.78 372

macro avg 0.78 0.77 0.77 372
weighted avg 0.80 0.78 0.78 372

Tuned Model Parameters: {'knn__n_neighbors': 1}

Double-click (or enter) to edit

LOGISTIC REGRESSION

from sklearn.pipeline import Pipeline

from sklearn.linear_model import LogisticRegression
from sklearn.model_selection import train_test_split,GridSearchCV
from sklearn.metrics import confusion_matrix, classification_report

# Setup the pipeline steps:

steps = [('lr', LogisticRegression())]

# Create the pipeline: pipeline

pipeline = Pipeline(steps)

#input
X = result_df.iloc[:,0:-1]

#target
y=result_df.iloc[:,-1]

#parameters for gridsearchcv if we increase range of entries from 5 to higher value, we

c_space = np.logspace(-4, 4, 3)
parameters = {'lr__C': c_space,'lr__penalty': ['l2']}

# Create train and test sets

X_train, X_test, y_train, y_test = train_test_split(X,y,test_size=0.4,random_state=42)

#call GridSearchCV and set crossvalscore to 2

cv = GridSearchCV(pipeline,parameters,cv=2)

cv.fit(X_train,y_train)

# Predict the labels of the test set: y_pred

y_pred = cv.predict(X_test)
LR_Accuracy = cv.score(X_test, y_test)

# Compute and print metrics

print("Accuracy: {}".format(LR_Accuracy))
print(classification_report(y_test, y_pred))
print("Tuned Model Parameters: {}".format(cv.best_params_))

Accuracy: 0.543010752688172
precision recall f1-score support

0 0.36 0.33 0.35 105

1 0.73 0.91 0.81 94
2 0.56 0.58 0.57 112
3 0.38 0.26 0.31 61

accuracy 0.54 372

macro avg 0.51 0.52 0.51 372
weighted avg 0.52 0.54 0.53 372

Tuned Model Parameters: {'lr__C': 10000.0, 'lr__penalty': 'l2'}

SVM

# Import the necessary modules for ML model

from sklearn.pipeline import Pipeline
from sklearn.svm import SVC
from sklearn.model_selection import train_test_split,GridSearchCV
from sklearn.metrics import confusion_matrix, classification_report

# Setup the pipeline

steps = [('SVM', SVC())]

pipeline = Pipeline(steps)

#input
X = result_df.iloc[:,0:-1]

#target
y=result_df.iloc[:,-1]

# Specify the hyperparameter space, if we increase the penalty(c) and gamma value the ac
#since it takes lots of time in google colab provided only a single value
parameters = {'SVM__C':[10],'SVM__gamma':[1]}
# Create train and test sets
X_train, X_test, y_train, y_test = train_test_split(X,y,test_size=0.4,random_state=21)

cv = GridSearchCV(pipeline,parameters,cv=3)
cv.fit(X_train,y_train)

y_pred = cv.predict(X_test)
SVM_Accuracy = cv.score(X_test, y_test)

# Compute and print metrics

SVM_Accuracy=cv.score(X_test, y_test)

print("Accuracy: {}".format(SVM_Accuracy))
print(classification_report(y_test, y_pred))

Accuracy: 0.8225806451612904
precision recall f1-score support

0 0.58 1.00 0.74 93

1 1.00 1.00 1.00 99
2 1.00 0.61 0.76 117
3 1.00 0.68 0.81 63

accuracy 0.82 372

macro avg 0.90 0.82 0.83 372
weighted avg 0.90 0.82 0.83 372

NOW COMBINING ALL 12 LEADS INTO A SINGLE CSV FILE AND THEN PERFROM
MODEL ANALYSIS

#lets try combining all 12 leads in a single csv

location= '/content/drive/MyDrive/CMPE255_PROJECT/'
for files in natsorted(os.listdir(location)):
if files.endswith(".csv") and not files.endswith("13.csv"):
if files!='Combined_IDLead_1.csv':
df=pd.read_csv('/content/drive/MyDrive/CMPE255_PROJECT/{}'.format(files))
df.drop(columns=['Unnamed: 0'],inplace=True)
test_final=pd.concat([test_final,df],axis=1,ignore_index=True)
test_final.drop(columns=test_final.columns[-1],axis=1,inplace=True)

#drop the target column

test_final.drop(columns=[255],axis=1,inplace=True)
test_final
 0 1 2 3 4 5 6 7

0 0.728449 0.680755 0.619010 0.645367 0.681570 0.732488 0.758448 0.750660 0.7

1 0.957972 0.950695 0.941024 0.930501 0.913601 0.892244 0.868016 0.855127 0.8

2 0.611084 0.661575 0.695790 0.741113 0.716666 0.595794 0.425022 0.286457 0.4

3 0.839213 0.861690 0.866457 0.865756 0.855027 0.855606 0.845561 0.843187 0.8

4 0.917753 0.924369 0.873765 0.791381 0.699513 0.604927 0.500312 0.446012 0.5

#write the final file to csv

test_final.to_csv('final_1D.csv',header=False,index=False)

TEST DIMENSIONALITY REDUCTION EXPLAINED VARIANCE ON THE DATA

# Now Perform Dimensionality reduction (PCA) on that Dataframe and check

from sklearn.decomposition import PCA

#do PCA and choose componeents as 400

pca = PCA(n_components=400)
x_pca = pca.fit_transform(test_final)
x_pca = pd.DataFrame(x_pca)

# Calculate the variance explained by priciple components

explained_variance = pca.explained_variance_ratio_
print('Variance of each component:', pca.explained_variance_ratio_)
print('\n Total Variance Explained:', round(sum(list(pca.explained_variance_ratio_))*100

#store the new pca generated dimensions in a dataframe

#store the new pca generated dimensions in a dataframe
pca_df = pd.DataFrame(data = x_pca)
target = pd.Series(result_df.iloc[:,-1], name='target')
final_result_df = pd.concat([pca_df, target], axis=1)
final_result_df
 Variance of each component: [8.04649534e-02 4.68818003e-02 3.76212504e-02 2.9
2.57031130e-02 2.32574514e-02 2.14376788e-02 2.04315151e-02
1.94482863e-02 1.79877408e-02 1.64766264e-02 1.53241665e-02
1.50689862e-02 1.41398267e-02 1.36330466e-02 1.33375324e-02
1.26355566e-02 1.25577001e-02 1.16968257e-02 1.11671338e-02
1.07975552e-02 1.06183806e-02 1.03402122e-02 1.01248410e-02
9.73197948e-03 9.25504395e-03 9.16367637e-03 8.76267060e-03
8.54270112e-03 8.20665462e-03 8.07642149e-03 7.90742343e-03
7.54929819e-03 7.21938018e-03 7.07604659e-03 6.89135251e-03
6.80575532e-03 6.71875790e-03 6.38252148e-03 6.33951897e-03
6.10254734e-03 5.94560955e-03 5.76371295e-03 5.71788829e-03
5.55354810e-03 5.42316932e-03 5.35640711e-03 5.08429353e-03
5.03302777e-03 4.96811576e-03 4.87696491e-03 4.63686128e-03
4.55349933e-03 4.45390625e-03 4.31579996e-03 4.28316592e-03
4.17213140e-03 4.12346241e-03 4.09072049e-03 3.99349122e-03
3.92129459e-03 3.81982060e-03 3.78116652e-03 3.73307150e-03
3.68894307e-03 3.55238746e-03 3.49148625e-03 3.40490507e-03
3.33593814e-03 3.25467389e-03 3.20023474e-03 3.14871964e-03
3.09091665e-03 3.07180393e-03 3.05651457e-03 2.95447952e-03
2.90507083e-03 2.84618700e-03 2.80939396e-03 2.76324718e-03
2.71487874e-03 2.68959207e-03 2.67378836e-03 2.62085254e-03
2.55991613e-03 2.53614502e-03 2.47015404e-03 2.45768102e-03
2.41851536e-03 2.39477316e-03 2.35560704e-03 2.29236345e-03
2.26928539e-03 2.24965527e-03 2.22764534e-03 2.19258829e-03
2.14654982e-03 2.09081474e-03 2.08656961e-03 2.04315332e-03
2.01191187e-03 1.99715030e-03 1.98092986e-03 1.93183566e-03
1.90133601e-03 1.86628808e-03 1.85847904e-03 1.79040117e-03
1.77318190e-03 1.76278440e-03 1.73682193e-03 1.70177712e-03
1.69142157e-03 1.66289246e-03 1.64192361e-03 1.62455779e-03
1.59836820e-03 1.57166872e-03 1.56017874e-03 1.55193712e-03
1.52130395e-03 1.50860404e-03 1.48563216e-03 1.45667689e-03
1.44862677e-03 1.43014707e-03 1.42443426e-03 1.39341888e-03
1.38941740e-03 1.38032166e-03 1.35292505e-03 1.33403513e-03
1.33300728e-03 1.31774024e-03 1.29238722e-03 1.24574072e-03
1.23408862e-03 1.21598644e-03 1.20568485e-03 1.19391143e-03
1.18690274e-03 1.16630751e-03 1.16159095e-03 1.14539199e-03
#save to dimensionally reduced csv file
1.13634359e-03 1.11858663e-03 1.10460060e-03 1.08515359e-03
final_result_df.to_csv("pca_final.csv")
1.07679695e-03 1.06488284e-03 1.05861426e-03 1.04012565e-03
1.03222232e-03 1.02519590e-03 1.01169941e-03 9.96444257e-04
import 9.76134514e-04
joblib 9.61104386e-04 9.57134099e-04 9.48294848e-04
9.35386446e-04
#save the PCA model 9.29858628e-04 9.24107282e-04 9.20229599e-04
9.00136970e-04 8.84392791e-04 8.60041244e-04 8.58222437e-04
joblib_file='PCA_ECG.pkl'
8.39586154e-04 8.34156616e-04 8.24745137e-04 8.19630377e-04
joblib.dump(pca,joblib_file)
8.11755902e-04 8.09589697e-04 7.93351930e-04 7.83229226e-04
['PCA_ECG.pkl']
7.69323633e-04 7.62916710e-04 7.61217310e-04 7.49412461e-04
7.41978508e-04 7.32319449e-04 7.28386324e-04 7.15766463e-04
7.00416470e-04 6.92792928e-04 6.87860571e-04 6.77118996e-04
TRYING DIFFERENT ML
6.69195650e-04 MODELS ON 6.52787237e-04
6.62776506e-04 THE ALL 12 LEADS COMBINED FILE WITHOUT
6.41350808e-04
6.31671343e-04 6.25941688e-04 6.20986818e-04 6.12964320e-04
DIMENSIONALITY
6.06757241e-04REDUCTION
6.00414979e-04 5.90442751e-04 5.85447566e-04
5.82053388e-04 5.72736727e-04 5.64768427e-04 5.62060875e-04
5.53942338e-04 5.47413376e-04 5.43815848e-04 5.39018247e-04
KNN 5.31538796e-04 5.21422265e-04 5.16620308e-04 5.13730678e-04
5.08883050e-04 5.04308686e-04 4.96238364e-04 4.91958416e-04
4.80055673e-04 4.74422583e-04 4.69414331e-04 4.65649131e-04
4.62052065e-04
# Import the necessary 4.58664175e-04 4.49131977e-04 4.46512859e-04
modules for ML model
from sklearn.pipeline import Pipeline 4.30056931e-04 4.24233413e-04
4.45747677e-04 4.36928354e-04
4 21656145 04 4
from sklearn.neighbors 20467965
import 04 4 16760267 04 4 15888840 04
KNeighborsClassifier
from sklearn.model_selection import train_test_split,GridSearchCV
from sklearn.metrics import confusion_matrix, classification_report

# Setup the pipeline steps:

steps = [('knn', KNeighborsClassifier())]

# Create the pipeline: pipeline

pipeline = Pipeline(steps)

# have paased less range value of hyperparamter since i'm using free tier version of goo
k_range = list(range(1, 30))
parameters = dict(knn__n_neighbors=k_range)

#input
X = final_result_df.iloc[:,:-1]

#target
y=final_result_df.iloc[:,-1]

# Create train and test sets

X_train, X_test, y_train, y_test = train_test_split(X,y,test_size=0.4,random_state=42)

#increasing cv score takes lot of time in gooogle colab, so kept it just 2.

cv = GridSearchCV(pipeline,parameters,cv=2)

cv.fit(X_train,y_train)

# Predict the labels of the test set: y_pred

y_pred = cv.predict(X_test)

Knn_Accuracy = cv.score(X_test, y_test)

# Compute and print metrics

print("Accuracy: {}".format(Knn_Accuracy))
print(classification_report(y_test, y_pred))
print("Tuned Model Parameters: {}".format(cv.best_params_))

Accuracy: 0.793010752688172
precision recall f1-score support

0 0.92 0.65 0.76 105

1 0.95 0.91 0.93 94
2 0.70 0.86 0.77 112
3 0.65 0.74 0.69 61

accuracy 0.79 372

macro avg 0.80 0.79 0.79 372
weighted avg 0.81 0.79 0.79 372

Tuned Model Parameters: {'knn__n_neighbors': 1}

LOGISTIC REGRESSION

from sklearn.pipeline import Pipeline

from sklearn.linear_model import LogisticRegression
from sklearn.model_selection import train_test_split,GridSearchCV
from sklearn.metrics import confusion_matrix, classification_report

# Setup the pipeline steps:

steps = [('lr', LogisticRegression())]

# Create the pipeline: pipeline

pipeline = Pipeline(steps)

#input
X = final_result_df.iloc[:,:-1]

#target
y=final_result_df.iloc[:,-1]

#parameters for gridsearchcv

c_space = np.logspace(-4, 4, 10)
parameters = {'lr__C': c_space,'lr__penalty': ['l2']}

# Create train and test sets

X_train, X_test, y_train, y_test = train_test_split(X,y,test_size=0.4,random_state=42)

#call GridSearchCV and set crossvalscore to 2

cv = GridSearchCV(pipeline,parameters,cv=2)

cv.fit(X_train,y_train)

# Predict the labels of the test set: y_pred

y_pred = cv.predict(X_test)
LR_Accuracy = cv.score(X_test, y_test)

# Compute and print metrics

print("Accuracy: {}".format(LR_Accuracy))
print(classification_report(y_test, y_pred))
print("Tuned Model Parameters: {}".format(cv.best_params_))

Accuracy: 0.7768817204301075
precision recall f1-score support

0 0.83 0.57 0.68 105

1 0.83 0.91 0.87 94
2 0.82 0.86 0.84 112
3 0.59 0.77 0.67 61

accuracy 0.78 372

macro avg 0.77 0.78 0.76 372
weighted avg 0.79 0.78 0.77 372

Tuned Model Parameters: {'lr__C': 0.3593813663804626, 'lr__penalty': 'l2'}

SVM

# Import the necessary modules for ML model

from sklearn.pipeline import Pipeline
from sklearn.svm import SVC
from sklearn.model_selection import train_test_split,GridSearchCV
from sklearn.metrics import confusion_matrix, classification_report

# Setup the pipeline

steps = [('SVM', SVC())]

pipeline = Pipeline(steps)

#input
X = final_result_df.iloc[:,:-1]

#target
y=final_result_df.iloc[:,-1]

# Specify the hyperparameter space, if we increase the penalty(c) and gamma value the ac
#since it takes lots of time in google colab provided only a single value
parameters = {'SVM__C':[1, 10, 100],
'SVM__gamma':[0.1, 0.01]}

# Create train and test sets

X_train, X_test, y_train, y_test = train_test_split(X,y,test_size=0.5,random_state=21)

cv = GridSearchCV(pipeline,parameters,cv=3)
cv.fit(X_train,y_train)

y_pred = cv.predict(X_test)
SVM_Accuracy = cv.score(X_test, y_test)

# Compute and print metrics

SVM_Accuracy=cv.score(X_test, y_test)

print("Accuracy: {}".format(SVM_Accuracy))
print(classification_report(y_test, y_pred))
print("Tuned Model Parameters: {}".format(cv.best_params_))

Accuracy: 0.9051724137931034
precision recall f1-score support

0 0.81 0.92 0.86 119

1 1.00 1.00 1.00 125
2 0.91 0.89 0.90 140
3 0.93 0.78 0.84 80

accuracy 0.91 464

macro avg 0.91 0.89 0.90 464
weighted avg 0.91 0.91 0.91 464

Tuned Model Parameters: {'SVM__C': 10, 'SVM__gamma': 0.01}

XGBOOST

from xgboost import XGBClassifier

from sklearn.metrics import accuracy_score

model = XGBClassifier()
model.fit(X_train, y_train)

y_pred = model.predict(X_test)
predictions = [round(value) for value in y_pred]

# evaluate predictions
accuracy = accuracy_score(y_test, predictions)
print("Accuracy: {}".format(accuracy))
print(classification_report(y_test, y_pred))

Accuracy: 0.853448275862069
precision recall f1-score support

0 0.79 0.70 0.74 119

1 0.98 1.00 0.99 125
 2 0.82 0.87 0.84 140
3 0.80 0.82 0.81 80

accuracy 0.85 464

macro avg 0.85 0.85 0.85 464
weighted avg 0.85 0.85 0.85 464

SAVING A VERY BASIC ML MODEL AND USING IT ON REALTIME PIPELINE TO

CHECK WORKING

# Import the necessary modules for ML model

from sklearn.neighbors import KNeighborsClassifier
from sklearn.metrics import confusion_matrix, classification_report
import joblib
#input
X = final_result_df.iloc[:,:-1]
#target
y=final_result_df.iloc[:,-1]

# Create train and test sets

X_train, X_test, y_train, y_test = train_test_split(X,y,test_size=0.4,random_state=42)

knn = KNeighborsClassifier(n_neighbors=1)

knn.fit(X_train,y_train)

joblib_file='knn_model_test.pkl'
joblib.dump(knn,joblib_file)

['knn_model_test.pkl']

# Import the necessary modules for ML model

from sklearn.pipeline import Pipeline
from sklearn.svm import SVC
from sklearn.model_selection import train_test_split,GridSearchCV
from sklearn.metrics import confusion_matrix, classification_report

#input
X = pd.read_csv('final_1D.csv',header=None)

#target
y=final_result_df.iloc[:,-1]

# Create train and test sets

X_train, X_test, y_train, y_test = train_test_split(X,y,test_size=0.5,random_state=21)

svm=SVC(C=10,gamma=0.01)

svm.fit(X_train,y_train)

joblib_file='svm_model_test.pkl'
joblib.dump(svm,joblib_file)

['svm_model_test.pkl']
 ENSEMBLE

# Importing required modules

from sklearn import linear_model, tree, ensemble
from sklearn.naive_bayes import GaussianNB
import xgboost
from xgboost import XGBClassifier
from sklearn.neighbors import KNeighborsClassifier
from sklearn.ensemble import VotingClassifier
from sklearn.metrics import accuracy_score
import pickle

#input
X = final_result_df.iloc[:,0:-1]

#target
y=final_result_df.iloc[:,-1]

# Create train and test sets

X_train, X_test, y_train, y_test = train_test_split(X,y,test_size=0.3,random_state=42)

# Stacking of ML Models
eclf = VotingClassifier(estimators=[
('SVM', SVC(probability=True)),
('knn', KNeighborsClassifier()),
('rf', ensemble.RandomForestClassifier()),
('bayes',GaussianNB()),
('logistic',LogisticRegression()),
], voting='soft')

# Hyperparameter Tuning using gridSearch

params = {'SVM__C':[1, 10, 100],
'SVM__gamma':[0.1, 0.01],
'knn__n_neighbors': [1,3,5],
'rf__n_estimators':[300, 400],
}

grid = GridSearchCV(estimator=eclf, param_grid=params, cv=5)

voting_clf = grid.fit(X_train, y_train)

print(grid.best_params_)
y_pred = voting_clf.predict(X_test)

# Compute and print metrics

Voting_Accuracy=voting_clf.score(X_test, y_test)

print("Accuracy: {}".format(Voting_Accuracy))
print(classification_report(y_test, y_pred))
print(voting_clf.best_params_)

{'SVM__C': 1, 'SVM__gamma': 0.1, 'knn__n_neighbors': 1, 'rf__n_estimators': 3

Accuracy: 0.9247311827956989
precision recall f1-score support

0 0.89 0.96 0.92 80

1 1.00 1.00 1.00 72
2 0.92 0.92 0.92 79
3 0.88 0.75 0.81 48
 accuracy 0.92 279
macro avg 0.92 0.91 0.91 279
weighted avg 0.92 0.92 0.92 279

{'SVM__C': 1, 'SVM__gamma': 0.1, 'knn__n_neighbors': 1, 'rf__n_estimators': 3

# open a file, where you ant to store the data

file = open('Heart_Disease_Prediction_using_ECG.pkl', 'wb')
# dump information to that file
pickle.dump(voting_clf, file)

SAVE AND USE THE ABOVE MODEL IN THE STREAMLIT APP :

https://colab.research.google.com/drive/139YVmcUBCiP52J2sX
3QE_eiu2sukVgpn?usp=sharing