Built this data-driven lung cancer detection system by employing multiple Machine Learning algorithms such as Logistic Regression, Decision Tree, SVM, KNN, Random Forest, and Naive Bayes on a dataset from Kaggle, achieving a combined accuracy of 94.6183%, using python modules such as numpy, pandas, matplotlib,scikit learn, streamlit, tensorflow.
Use the package manager pip to install numpy, pandas, sklearn, streamlit, imblearn and tensorflow.
pip install numpy
pip install pandas
pip install matplotlib
pip install sklearn
pip install streamlit
pip install imblearn
pip install tensorflow
pip install streamlit-marquee 1. Download the source code, give it a name
2. Open any terminal and locate the file
3. Use the command "python -m streamlit run <file_name.py>"
import numpy as np
import pandas as pd
from sklearn.model_selection import train_test_split
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import accuracy_score
from sklearn.tree import DecisionTreeClassifier
import matplotlib.pyplot as plt
from sklearn.tree import plot_tree
from sklearn.svm import SVC
from sklearn.neighbors import KNeighborsClassifier
from sklearn.ensemble import RandomForestClassifier
from sklearn.naive_bayes import GaussianNB
import streamlit
import tensorflow as tf
from tensorflow import keras
from imblearn.combine import SMOTETomek
from streamlit_marquee import streamlit_marquee* Logistic Regression
* Decision Tree
* Support Vector Machine
* K - Nearest Neighbours
* Random Forest Classifier
* Naive Bayes
* Artificial Neural Networks
Specific features in each model are described in the implementation.
I made use of a function called accuracy_score, which is available in sklearn.metrics.
At first, I predicted data and stored it in variable ptrain, then compared it with original data (Y_train).
ptrain = Model_Name.predict(X_train.values)
accuracy = accuracy_score(Y_train, ptrain)
print(accuracy)The above task is performed again on test data, to check for overfitting.
ptest = Model_Name.predict(X_test.values)
accuracy = accuracy_score(Y_test, ptest)
print(accuracy)This dictionary is used to map strings to integers and vice versa. Pleas have a look at the main function to better understand convert dictionary.
convert = {
'Yes': 2,
'No': 1,
'M': 1,
'F': 0,
1: 'High chances of Lung Cancer',
0: 'Low chances of Lung Cancer'
}def main():
#data retrieval
data = pd.read_csv('D:/Programming/Streamlit/Lung_Cancer/cancer_data.csv')
#plotting labels to check data
# label_counts = data['LUNG_CANCER'].value_counts()
# fig, ax = plt.subplots()
# colors = ['Red', 'Green']
# ax.pie(label_counts, labels=label_counts.index, autopct='%1.1f%%', colors=colors)
# st.pyplot(fig)
data['LUNG_CANCER'] = data['LUNG_CANCER'].replace({'YES': 1, 'NO': 0})
data['GENDER'] = data['GENDER'].replace({'M': 1, 'F': 0})
X = data.drop(['LUNG_CANCER'], axis = 1)
Y = data['LUNG_CANCER'].astype(int)
#Handling imbalance
smk = SMOTETomek(random_state = 42)
X, Y = smk.fit_resample(X, Y)
data = X
data['LUNG_CANCER'] = Y
X = data.drop(['LUNG_CANCER'], axis = 1).astype(int)
Y = data['LUNG_CANCER'].astype(int)
view_dataset = st.button("View dataset")
if view_dataset:
st.write(data)The below part is used to handle input from user.
st.selectbox is a dropdown.
st.write, puts content on the screen.
user_inputs = []
gender = st.selectbox("Please select your gender : ", ("Select", "M", "F"))
user_inputs.append(convert.get(gender))
age = st.number_input("Please enter your age : ", value=0, step=1)
user_inputs.append(age)
smoke = st.selectbox("Do you smoke?", ("Select", "Yes", "No"))
user_inputs.append(convert.get(smoke))
yellow_fingers = st.selectbox(
"Do you have yellow fingers?", ("Select", "Yes", "No"))
user_inputs.append(convert.get(yellow_fingers))
anxiety = st.selectbox("Do you have anxiety issues?",
("Select", "Yes", "No"))
user_inputs.append(convert.get(anxiety))
chronic = st.selectbox(
"Do you have any chronic diseases?", ("Select", "Yes", "No"))
user_inputs.append(convert.get(chronic))
fatigue = st.selectbox("Do you experience fatigue?",
("Select", "Yes", "No"))
user_inputs.append(convert.get(fatigue))
allergy = st.selectbox("Do you have any allergies?",
("Select", "Yes", "No"))
user_inputs.append(convert.get(allergy))
wheezing = st.selectbox("Do you wheeze?", ("Select", "Yes", "No"))
user_inputs.append(convert.get(wheezing))
alcohol = st.selectbox("Do you consume alcohol?", ("Select", "Yes", "No"))
user_inputs.append(convert.get(alcohol))
cough = st.selectbox("Do you cough on a regular basis?",
("Select", "Yes", "No"))
user_inputs.append(convert.get(cough))
breathing = st.selectbox(
"Do you have any kind of difficulity in breathing?", ("Select", "Yes", "No"))
user_inputs.append(convert.get(breathing))
swallowing = st.selectbox(
"Do you have any kind of difficulity in swallowing?", ("Select", "Yes", "No"))
user_inputs.append(convert.get(swallowing))
chest_pain = st.selectbox(
"Are you suffering from chest pain?", ("Select", "Yes", "No"))
user_inputs.append(convert.get(chest_pain)) selected_model = st.selectbox("select your model : ", ("None", "Logistic Regression", "Decision Tree", "Support Vector Machine", "K Nearest Neighbour", "Random Forest", "Naive Bayes", "Artificial Neural Networks"))
# handle errors in input
if not (gender and age and smoke and yellow_fingers and anxiety and chronic and fatigue and allergy and wheezing and alcohol and cough and breathing and swallowing and chest_pain) or None in user_inputs:
pass
elif not selected_model or selected_model == "None":
st.write("Please select one of the given models")
else:
# st.write('Model : ', selected_model)
X_train, X_test, Y_train, Y_test = train_test_split(X, Y, test_size=0.3, random_state=42)
# conversion of array
narry = np.asarray(user_inputs, dtype=int)
reshaped_array = narry.reshape(1, -1) if selected_model == "Logistic Regression":
solver = st.selectbox("select solver", ("None", "liblinear", "lbfgs", "sag", "saga"))
if solver and solver != "None":
if solver == "liblinear":
penalty = st.selectbox("Please select penalty", ('l1', 'l2'))
elif solver == "lbfgs" or solver == "sag":
penalty = st.selectbox("Please select penalty", ('l2', 'none'))
else:
penalty = st.selectbox("Please select penalty", ('l1', 'l2', 'none'))
Logistic_Regression_model = LogisticRegression(max_iter=1000, solver=solver, penalty=penalty)
Logistic_Regression_model.fit(X_train.values, Y_train.values)
st.write("Prediction using Logistic regression : ", convert.get(
Logistic_Regression_model.predict(reshaped_array)[0]))
view_train_accuracy = st.button("View train accuracy")
if view_train_accuracy:
ptrain = Logistic_Regression_model.predict(X_train.values)
accuracy = accuracy_score(Y_train, ptrain)
st.write(f"Train accuracy for {selected_model} is {round(accuracy*100, 5)}%")
view_test_accuracy = st.button("View test accuracy")
if view_test_accuracy:
ptest = Logistic_Regression_model.predict(X_test.values)
accuracy = accuracy_score(Y_test, ptest)
st.write(f"Test accuracy for {selected_model} {round(accuracy*100, 5)}%") elif selected_model == "Decision Tree":
criterion = st.selectbox("Criterion", ('Select', 'gini', 'entropy'))
if criterion != 'Select':
Decision_Tree_model = DecisionTreeClassifier(criterion=criterion)
Decision_Tree_model.fit(X_train.values, Y_train.values)
st.write("Prediction using Decision Tree : ", convert.get(
Decision_Tree_model.predict(reshaped_array)[0]))
view_tree = st.button("View Tree")
if view_tree:
# plt.figure(figsize=(width, height))
with st.spinner('Please wait while the decision tree is being loaded...'):
plt.figure(figsize=(50, 10))
plot_tree(decision_tree=Decision_Tree_model, filled=True,
feature_names=X.columns, rounded=True)
st.pyplot(plt)
view_train_accuracy = st.button("View train accuracy")
if view_train_accuracy:
ptrain = Decision_Tree_model.predict(X_train.values)
accuracy = accuracy_score(Y_train, ptrain)
st.write(f"Train accuracy for {selected_model} is {round(accuracy*100, 5)}%")
view_test_accuracy = st.button("View test accuracy")
if view_test_accuracy:
ptest = Decision_Tree_model.predict(X_test.values)
accuracy = accuracy_score(Y_test, ptest)
st.write(f"Test accuracy for {selected_model} is {round(accuracy*100, 5)}%") elif selected_model == "Support Vector Machine":
# svm_model = SVC() #By default Radial - Basis function kernel
selected_kernel = st.selectbox(
"Please select a kernel", ("None", "linear", "rbf", "poly"))
if selected_kernel and selected_kernel != "None":
svm_model = SVC(kernel=selected_kernel)
svm_model.fit(X_train.values, Y_train.values)
st.write("Prediction using SVM Model : ", convert.get(
svm_model.predict(reshaped_array)[0]))
view_train_accuracy = st.button("View train accuracy")
if view_train_accuracy:
ptrain = svm_model.predict(X_train.values)
accuracy = accuracy_score(Y_train, ptrain)
st.write(f"Train accuracy for {selected_model} is {round(accuracy*100, 5)}%")
view_test_accuracy = st.button("View test accuracy")
if view_test_accuracy:
ptest = svm_model.predict(X_test.values)
accuracy = accuracy_score(Y_test, ptest)
st.write(f"Test accuracy for {selected_model} is {round(accuracy*100, 5)}%") elif selected_model == "K Nearest Neighbour":
# knn_model = KNeighborsClassifier(value of k) #By default, value of k is 5
weights = st.selectbox("select weights", ("Select", "uniform", "distance"))
if weights != "Select":
kval = st.slider("Please select the value of k", 0, len(X_train.values))
if kval:
knn_model = KNeighborsClassifier(n_neighbors = kval, weights=weights)
knn_model.fit(X_train.values, Y_train.values)
st.write("Prediction using KNN Model : ", convert.get(
knn_model.predict(reshaped_array)[0]), " with k as ", kval)
view_train_accuracy = st.button("View train accuracy")
if view_train_accuracy:
ptrain = knn_model.predict(X_train.values)
accuracy = accuracy_score(Y_train, ptrain)
st.write(f"Train accuracy for {selected_model} is {round(accuracy*100, 5)}%")
view_test_accuracy = st.button("View test accuracy")
if view_test_accuracy:
ptest = knn_model.predict(X_test.values)
accuracy = accuracy_score(Y_test, ptest)
st.write(f"Test accuracy for {selected_model} is {round(accuracy*100, 5)}%") elif selected_model == "Random Forest":
n = st.slider("select number of estimators", 1, 5000)
criterion = st.selectbox("Criterion", ('gini', 'entropy'))
Random_Forest_Classifier = RandomForestClassifier(n_estimators = n, criterion=criterion)
Random_Forest_Classifier.fit(X_train.values, Y_train.values)
st.write("Prediction using Random Forest Classifier : ", convert.get(
Random_Forest_Classifier.predict(reshaped_array)[0]))
view_train_accuracy = st.button("View train accuracy")
if view_train_accuracy:
ptrain = Random_Forest_Classifier.predict(X_train.values)
accuracy = accuracy_score(Y_train, ptrain)
st.write(f"Train accuracy for {selected_model} is {round(accuracy*100, 5)}%")
view_test_accuracy = st.button("View test accuracy")
if view_test_accuracy:
ptest = Random_Forest_Classifier.predict(X_test.values)
accuracy = accuracy_score(Y_test, ptest)
st.write(f"Test accuracy for {selected_model} is {round(accuracy*100, 5)}%") elif selected_model == "Naive Bayes":
Naive_Bayes_model = GaussianNB()
Naive_Bayes_model.fit(X_train.values, Y_train.values)
st.write("Prediction using Naive Bayes Model : ", convert.get(
Naive_Bayes_model.predict(reshaped_array)[0]))
view_train_accuracy = st.button("View train accuracy")
if view_train_accuracy:
ptrain = Naive_Bayes_model.predict(X_train.values)
accuracy = accuracy_score(Y_train, ptrain)
st.write(f"Train accuracy for {selected_model} is {round(accuracy*100, 5)}%")
view_test_accuracy = st.button("View test accuracy")
if view_test_accuracy:
ptest = Naive_Bayes_model.predict(X_test.values)
accuracy = accuracy_score(Y_test, ptest)
st.write(f"Test accuracy for {selected_model} is {round(accuracy*100, 5)}%") elif selected_model == "Artificial Neural Networks":
model = keras.Sequential([
keras.layers.Dense(100, input_shape=(14,), activation = 'sigmoid'),
keras.layers.Dense(100, activation = 'sigmoid'),
keras.layers.Dense(100, activation = 'sigmoid'),
keras.layers.Dense(2, activation = 'sigmoid'),
])
model.compile(optimizer = 'adam', loss = 'sparse_categorical_crossentropy', metrics=['accuracy'])
with st.spinner('Please wait while model is running'):
history = model.fit(X_train, Y_train, epochs = 120)
predictArray = [reshaped_array]
predictAns = model.predict(predictArray)
st.write("Prediction using Artificial Neural networks : ", convert.get(np.argmax(predictAns)))
view_train_accuracy = st.button("View train accuracy")
if view_train_accuracy:
accuracy = history.history['accuracy'][-1]
st.write(f"Train accuracy for {selected_model} is {round(accuracy*100, 5)}%")
view_test_accuracy = st.button("View test accuracy")
if view_test_accuracy:
loss, accuracy = model.evaluate(X_test, Y_test)
st.write(f"Test accuracy for {selected_model} is {round(accuracy*100, 5)}%")if __name__ == "__main__":
main()footer = """
<style>
a:hover
{
opacity:50%;
}
.footer
{
position: fixed;
left: 0;
bottom: 0;
width: 100%;
background-color: white;
text-align: center;
}
</style>
<div class="footer">
<p> Developed with ❤ by <a style='display: block; text-align: center;' href="https://iamssuraj.netlify.app/" target="_blank"> Suraj Sanganbhatla </a> </p>
</div>
"""
st.markdown(footer, unsafe_allow_html=True)Output Samples:
