Fetal Health Analysis Dense Model: # Importing The Dependince

fetal-health-analysis-dense-model-acc-94 - Jupyter Notebook 28/02/23, 11:07 PM
Fetal health analysis dense Model

In [1]: # importing the dependince
import os
import numpy as np
# Data
import pandas as pd
# Data Visualization
import seaborn as sns
import matplotlib.pyplot as plt
# Data Preprocessing
from sklearn.preprocessing import StandardScaler
from sklearn.model_selection import StratifiedShuffleSplit, train_test_
# Model
from keras import Sequential
from keras.layers import Dense, Dropout
# Callbacks
from keras.callbacks import EarlyStopping, ModelCheckpoint
# Model Performance
from sklearn.metrics import accuracy_score, precision_score, recall_sco
Data Collection
In [3]: #loading the dataset in pandas datafram
file_path = 'fetal_health.csv'
data = pd.read_csv(file_path)
http://localhost:8888/notebooks/Downloads/fetal-health-analysis-dense-model-acc-94.ipynb#Data Page 1 of 35
In [4]: #check first five rows of the dataset

data.head()
Out[4]:
baseline
accelerations fetal_movement uterine_contractions light_decelerations severe_dece
value
0 120.0 0.000 0.0 0.000 0.000
1 132.0 0.006 0.0 0.006 0.003
2 133.0 0.003 0.0 0.008 0.003
3 134.0 0.003 0.0 0.008 0.003
4 132.0 0.007 0.0 0.008 0.000
5 rows × 22 columns
In [5]: #check last five rows of the dataset

data.tail()
Out[5]:
baseline
accelerations fetal_movement uterine_contractions light_decelerations severe_d
value
2121 140.0 0.000 0.000 0.007 0.0
2122 140.0 0.001 0.000 0.007 0.0
2123 140.0 0.001 0.000 0.007 0.0
2124 140.0 0.001 0.000 0.006 0.0
2125 142.0 0.002 0.002 0.008 0.0
In [6]: #check shape of the dataset

data.shape
Out[6]: (2126, 22)
In [7]: #check mathamatci realtionship of the dataset

data.describe()
Out[7]:
baseline
accelerations fetal_movement uterine_contractions light_decelerations
value
count 2126.000000 2126.000000 2126.000000 2126.000000 2126.000000
mean 133.303857 0.003178 0.009481 0.004366 0.001889
std 9.840844 0.003866 0.046666 0.002946 0.002960
min 106.000000 0.000000 0.000000 0.000000 0.000000
25% 126.000000 0.000000 0.000000 0.002000 0.000000
50% 133.000000 0.002000 0.000000 0.004000 0.000000
75% 140.000000 0.006000 0.003000 0.007000 0.003000
max 160.000000 0.019000 0.481000 0.015000 0.015000
In [8]: #check more infomation of the dataset

data.info()
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 2126 entries, 0 to 2125
Data columns (total 22 columns):
# Column Non-N
ull Count Dtype
--- ------ -----
--------- -----
0 baseline value 2126
non-null float64
1 accelerations 2126
non-null float64
2 fetal_movement 2126
non-null float64
3 uterine_contractions 2126
non-null float64
4 light_decelerations 2126
non-null float64
5 severe_decelerations 2126
non-null float64
6 prolongued_decelerations 2126
non-null float64
7 abnormal_short_term_variability 2126
non-null float64
8 mean_value_of_short_term_variability 2126
non-null float64
9 percentage_of_time_with_abnormal_long_term_variability 2126
non-null float64
10 mean_value_of_long_term_variability 2126
non-null float64
non-null float64
11 histogram_width 2126
non-null float64
12 histogram_min 2126
non-null float64
13 histogram_max 2126
non-null float64
14 histogram_number_of_peaks 2126
non-null float64
15 histogram_number_of_zeroes 2126
non-null float64
16 histogram_mode 2126
non-null float64
17 histogram_mean 2126
non-null float64
18 histogram_median 2126
non-null float64
19 histogram_variance 2126
non-null float64
20 histogram_tendency 2126
non-null float64
21 fetal_health 2126
non-null float64
dtypes: float64(22)
memory usage: 365.5 KB
In [10]: #check missing values of the dataset

data.isnull().sum()
Out[10]: baseline value 0

accelerations 0
fetal_movement 0
uterine_contractions 0
light_decelerations 0
severe_decelerations 0
prolongued_decelerations 0
abnormal_short_term_variability 0
mean_value_of_short_term_variability 0
percentage_of_time_with_abnormal_long_term_variability 0
mean_value_of_long_term_variability 0
histogram_width 0
histogram_min 0
histogram_max 0
histogram_number_of_peaks 0
histogram_number_of_zeroes 0
histogram_mode 0
histogram_mean 0
histogram_median 0
histogram_variance 0
histogram_tendency 0
fetal_health 0
dtype: int64
Data Visualization
In [13]: for i in data.columns:
plt.figure()
plt.hist(data[i])
In [17]: for i in data.columns:

sns.distplot(data[i])
tributions.py:2619: FutureWarning: `distplot` is a deprecated func
tion and will be removed in a future version. Please adapt your co
de to use either `displot` (a figure-level function with similar f
lexibility) or `histplot` (an axes-level function for histograms).
warnings.warn(msg, FutureWarning)
/Users/vikky/opt/anaconda3/lib/python3.9/site-packages/seaborn/dis
Let's have a look a the Correlation among Features.
In [11]: corr = data.corr()
plt.figure(figsize=(15,10))
sns.heatmap(corr, annot=True)
plt.show()
Their are many relations let's have a look at the most strongest ones.
In [214]: def search_strong_relation(value):

if value >= 0.7:
return value
elif value<=-0.7:
return value
else:
return 0

corr = (corr>=0.7).astype('int')
plt.title("Strong Positive Relation")
plt.show()

corr = (corr<-0.7).astype('int')
plt.title("Strong Negative Relation")
plt.show()
You can change the Threshold Values if you want to, but I consider that the relation
should have a score >0.7(70%). Let's plot these relations. Before plotting the relations
we will also plot the individual column.
In [217]: plt.figure(figsize=(10,8))
sns.histplot(data['histogram_mode'], kde=True)
plt.show()
In [218]: sns.catplot(
data=data,
y='histogram_mode',
x='fetal_health',
kind='box',
aspect=1.5
)
plt.show()
The values are concentrated between 130 - 150.
sns.histplot(data['baseline value'], kde=True)
plt.show()
data=data,
y='baseline value',
x='fetal_health',
kind='box',
aspect=1.5
)
plt.show()
Here, the values are concentrated between 120 - 145.
Theya had a positive relation, let's see
sns.scatterplot(
data=data,
x='histogram_mode',
y='baseline value',
hue='fetal_health'
)
plt.show()
Clearly they have a Positive Linear Relationship, the interesting this is the *class
distribution :
Class 1 and 3 are easily seperable.

But class 2 and 3 are mixed together.
sns.histplot(data['histogram_mean'], kde=True)
plt.show()
data=data,
y='histogram_mean',
x='fetal_health',
kind='box',
aspect=1.5
)
plt.show()
sns.histplot(data['histogram_median'], kde=True)
plt.show()
data=data,
y='histogram_median',
x='fetal_health',
kind='box',
aspect=1.5
)
plt.show()
Both belong to the same distribution that's why the data distribution range is roughly
the same.
sns.scatterplot(
data=data,
x='histogram_mean',
y='baseline value',
hue='fetal_health'
)
plt.show()
sns.scatterplot(
data=data,
x='histogram_median',
y='baseline value',
hue='fetal_health'
)
plt.show()
This distribution is much more the same.
sns.histplot(data['histogram_width'], kde=True)
plt.show()
sns.histplot(data['histogram_number_of_peaks'], kde=True)
plt.show()
Number of Peaks continuosly decreases.
sns.scatterplot(
data=data,
x='histogram_number_of_peaks',
y='histogram_width',
hue='fetal_health'
)
plt.show()
The Linear Relation is clear but the classes are not clearly seperable.
data=data,
x='fetal_health',
kind='box',
aspect=1.5
)
plt.show()
This confirms that the classes are overlapping each other. The overlapping is high, this
variable alone cannot decide the class.
sns.scatterplot(
data=data,
x='histogram_mode',
y='histogram_mean',
hue='fetal_health'
)
plt.show()
sns.scatterplot(
data=data,
x='histogram_mode',
y='histogram_median',
hue='fetal_health'
)
plt.show()
sns.scatterplot(
data=data,
x='histogram_median',
y='histogram_mean',
hue='fetal_health'
)
plt.show()
This had to be the case after-all they are statically related. The classes are seperable,
thus it can be easy for the model to detect the hidden pattern.
It's time for Negative Relations.
sns.histplot(data['histogram_min'], kde=True)
plt.show()
data=data,
y='histogram_min',
x='fetal_health',
kind='box',
aspect=1.5
)
plt.show()
sns.scatterplot(
data=data,
x='histogram_min',
hue='fetal_health'
)
plt.show()
The Negative Relationship can easily be noted, however the classes are not easily
seperable. We are done with data understanding it's time for the model to understand
the data. But before doing that, we need to prepared the data.
Data Splitting
The Features are not on the same scale, let's bring them all to one scale.
In [238]: y = data.pop('fetal_health') - 1 # To bring values from 1-3 to 0-2

X = data
In [239]: scaler = StandardScaler()

X_scaled = scaler.fit_transform(X) # Return A Numpy Array
Let's split the data into training, testing and validation data. To maintain the Class
Distribution, we will using Stratified Splitting.
In [240]: splitter = StratifiedShuffleSplit(n_splits=3)

for train_ids, test_ids in splitter.split(X_scaled, y):
X_train_full, y_train_full = X_scaled[train_ids], y[train_ids]
X_test, y_test = X_scaled[test_ids], y[test_ids]
In [241]: X_train, X_valid, y_train, y_valid = train_test_split(X_train_full,
Feature Selection
This is an optional step, as we will be using a Dense Model it will automatically perform
Feature Selection, but if you are using a Machine Learning Model you can use it.
In [242]: from sklearn.ensemble import RandomForestClassifier
In [243]: rfc = RandomForestClassifier()

rfc.fit(X_train, y_train)
feature_imps = rfc.feature_importances_
feature_names = data.columns
sns.barplot(
x=feature_names,
y=feature_imps
)
plt.axhline(np.mean(feature_imps), color='k', linestyle="--", alpha
plt.xticks(rotation=90)
plt.legend()
plt.show()
Many features are below the Mean Importance, you can drop them it will no aﬀect the
model much. Dense Models can do this on their own, so we don't have to worry about
it. You can do multiple runs and get the mean, that will be a Robust Prediction but it
will generally be the same.
Dense Model
In [325]: name = "dense"
# Model Architecture
model = Sequential([
Dense(32, activation='relu', kernel_initializer='he_normal'),
Dropout(0.2),
Dense(3, activation='softmax'),
], name=name)
# Compile
model.compile(
loss='sparse_categorical_crossentropy',
optimizer='adam',
metrics=['accuracy']
)
# Callbacks
cbs = [EarlyStopping(patience=3, restore_best_weights=True), ModelCheck
In [326]: model.fit(X_train, y_train, validation_data=(X_valid, y_valid), epochs
Epoch 1/15
48/48 [==============================] - 1s 7ms/step - loss: 0.627
3 - accuracy: 0.7876 - val_loss: 0.3977 - val_accuracy: 0.8590
Epoch 2/15
48/48 [==============================] - 0s 3ms/step - loss: 0.380
Epoch 3/15
48/48 [==============================] - 0s 3ms/step - loss: 0.313
Epoch 4/15
48/48 [==============================] - 0s 3ms/step - loss: 0.283
Epoch 5/15
48/48 [==============================] - 0s 3ms/step - loss: 0.245
Epoch 6/15
48/48 [==============================] - 0s 3ms/step - loss: 0.224
Epoch 7/15
48/48 [==============================] - 0s 5ms/step - loss: 0.219
Epoch 8/15
48/48 [==============================] - 0s 3ms/step - loss: 0.203
Epoch 9/15
48/48 [==============================] - 0s 3ms/step - loss: 0.194
Epoch 10/15
48/48 [==============================] - 0s 3ms/step - loss: 0.182
Epoch 11/15
48/48 [==============================] - 0s 3ms/step - loss: 0.171
Epoch 12/15
48/48 [==============================] - 0s 3ms/step - loss: 0.158
Epoch 13/15
48/48 [==============================] - 0s 3ms/step - loss: 0.145
Epoch 14/15
48/48 [==============================] - 0s 3ms/step - loss: 0.154
Epoch 15/15
48/48 [==============================] - 0s 3ms/step - loss: 0.144
Out[326]: <keras.callbacks.History at 0x7fa7cc6ed190>
Evaluation
The training is promising, let's check model's performance on the data ir has never seen
before.
In [327]: model.evaluate(X_test, y_test)
7/7 [==============================] - 0s 2ms/step - loss: 0.2408

- accuracy: 0.9202
Out[327]: [0.24079568684101105, 0.920187771320343]
Performance Analysis
Let's see model's performance on various metrics.
In [328]: # Make Prediction

pred = np.argmax(model.predict(X_test), axis=1)
# Calculate Score
acc_score = accuracy_score(y_test, pred)
pre_score = precision_score(y_test, pred, average='macro')
re_score = recall_score(y_test, pred, average='macro')
f_score = f1_score(y_test, pred, average='macro')
# Show Prediction
print("Accuracy Score : {:.6}".format(acc_score))
print("Precision Score : {:.6}".format(pre_score))
print("Recall Score : {:.6}".format(re_score))
print("f1 Score : {:.6}".format(f_score))
Accuracy Score : 0.920188

Precision Score : 0.857048
Recall Score : 0.849928
f1 Score : 0.853406
In [331]: print(classification_report(y_test, pred))
precision recall f1-score support
0.0 0.95 0.96 0.95 166

1.0 0.79 0.76 0.77 29
2.0 0.83 0.83 0.83 18
accuracy 0.92 213

macro avg 0.86 0.85 0.85 213
weighted avg 0.92 0.92 0.92 213

Fetal Health Analysis Dense Model: # Importing The Dependince

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Fetal Health Analysis Dense Model: # Importing The Dependince

Uploaded by

Copyright:

Available Formats

fetal-health-analysis-dense-model-acc-94 - Jupyter Notebook 28/02/23, 11:07 PM

Fetal health analysis dense Model

In [4]: #check first five rows of the dataset

0 120.0 0.000 0.0 0.000 0.000

1 132.0 0.006 0.0 0.006 0.003

2 133.0 0.003 0.0 0.008 0.003

3 134.0 0.003 0.0 0.008 0.003

4 132.0 0.007 0.0 0.008 0.000

In [5]: #check last five rows of the dataset

2121 140.0 0.000 0.000 0.007 0.0

2122 140.0 0.001 0.000 0.007 0.0

2123 140.0 0.001 0.000 0.007 0.0

2124 140.0 0.001 0.000 0.006 0.0

2125 142.0 0.002 0.002 0.008 0.0

In [6]: #check shape of the dataset

Out[6]: (2126, 22)

In [7]: #check mathamatci realtionship of the dataset

count 2126.000000 2126.000000 2126.000000 2126.000000 2126.000000

mean 133.303857 0.003178 0.009481 0.004366 0.001889

std 9.840844 0.003866 0.046666 0.002946 0.002960

min 106.000000 0.000000 0.000000 0.000000 0.000000

25% 126.000000 0.000000 0.000000 0.002000 0.000000

50% 133.000000 0.002000 0.000000 0.004000 0.000000

75% 140.000000 0.006000 0.003000 0.007000 0.003000

max 160.000000 0.019000 0.481000 0.015000 0.015000

In [8]: #check more infomation of the dataset

In [10]: #check missing values of the dataset

Out[10]: baseline value 0

In [17]: for i in data.columns:

Let's have a look a the Correlation among Features.

In [11]: corr = data.corr()

In [214]: def search_strong_relation(value):

In [215]: corr = data.corr()

In [216]: corr = data.corr()

The values are concentrated between 130 - 150.

Here, the values are concentrated between 120 - 145.

Theya had a positive relation, let's see

Class 1 and 3 are easily seperable.

This distribution is much more the same.

Number of Peaks continuosly decreases.

It's time for Negative Relations.

In [238]: y = data.pop('fetal_health') - 1 # To bring values from 1-3 to 0-2

In [239]: scaler = StandardScaler()

In [240]: splitter = StratifiedShuffleSplit(n_splits=3)

In [241]: X_train, X_valid, y_train, y_valid = train_test_split(X_train_full,

In [242]: from sklearn.ensemble import RandomForestClassifier

In [243]: rfc = RandomForestClassifier()

In [326]: model.fit(X_train, y_train, validation_data=(X_valid, y_valid), epochs

Out[326]: <keras.callbacks.History at 0x7fa7cc6ed190>

In [327]: model.evaluate(X_test, y_test)

7/7 [==============================] - 0s 2ms/step - loss: 0.2408

Out[327]: [0.24079568684101105, 0.920187771320343]

In [328]: # Make Prediction

Accuracy Score : 0.920188

In [331]: print(classification_report(y_test, pred))

precision recall f1-score support

0.0 0.95 0.96 0.95 166

accuracy 0.92 213

You might also like