app.py

import streamlit as st
import pandas as pd
from joblib import load
import numpy as np
import matplotlib.pyplot as plt

@st.cache_resource
def load_data() -> pd.DataFrame:
    """
    Loads the `.csv` data using pandas
    """
    df = pd.read_csv('./lung_disease_data.csv')
    
    numerical_columns = ['Age', 'Lung Capacity', 'Hospital Visits']
    df[numerical_columns] = df[numerical_columns].fillna(df[numerical_columns].mean())

    # Impute categorical features with their mode (most frequent value)
    categorical_columns = ['Gender', 'Smoking Status', 'Disease Type', 'Treatment Type', 'Recovered']
    df[categorical_columns] = df[categorical_columns].fillna(df[categorical_columns].mode().iloc[0])
    
    return df

@st.cache_resource
def load_models() -> dict:
    """
    Loads the trained models for prediction.
    """
    nb = load('./models/GaussianNB.pkl')
    lg = load('./models/LogisticRegression.pkl')
    rf = load('./models/RandomForests.pkl')
    svm = load('./models/SVM.pkl')
    xgb = load('./models/XGBoost.pkl')
    
    models = dict({
        'Gaussian Naive Bayes': nb,
        'Logistic Regression': lg,
        'Random Forest': rf,
        'Support Vector Machines': svm,
        'XG Boost': xgb
    })
    
    return models

def prediction(model, age: int, gender: str,
               smoke_status: str, lung_capacity: float,
               disease_type: str, treatment_type: str,
               hospital_visits: int
               ) -> int:
    
    df_input = pd.DataFrame(
        {'Age': [age],
         'Hospital Visits': [hospital_visits],
         'Lung Capacity': [lung_capacity],
         'Gender': [1 if gender == "Male" else 0],
         'Smoking Status': [1 if smoke_status == "Yes" else 0],
         'Disease Type_Asthma': [1 if disease_type in 'Disease Type_Asthma' else 0],
         'Disease Type_Bronchitis': [1 if disease_type in 'Disease Type_Bronchitis' else 0],
        'Disease Type_COPD': [1 if disease_type in 'Disease Type_COPD' else 0],
        'Disease Type_Lung Cancer': [1 if disease_type in 'Disease Type_Lung Cancer' else 0],
        'Disease Type_Pneumonia': [1 if disease_type in 'Disease Type_Pneumonia' else 0],
        
        'Treatment Type_Medication': [1 if treatment_type in 'Treatment Type_Medication' else 0],
        'Treatment Type_Surgery': [1 if treatment_type in 'Treatment Type_Surgery' else 0],
        'Treatment Type_Therapy': [1 if treatment_type in 'Treatment Type_Therapy' else 0]
         }
    )
    
    input_arr = np.array(df_input)
    
    prediction = model.predict(input_arr)[0]
    
    return prediction.item()

def main():
    st.header("Lung Disease Recovery Predictor")
    st.caption('Prepared by `hydraadra112` | John Manuel Carado')
    
    data_tab, pred_tab, data_viz = st.tabs(['About Data', 'Prediction', 'Data Viz'])
    df = load_data()
    
    with data_tab:
        st.header('About the Data')
        st.caption('In this tab, we will explore the particular details about our data.')
        
        st.caption('Take a look at the data table.')
        st.dataframe(df)
        
        col1, col2 = st.columns(2)
        
        with col1:
            st.caption('This dataset captures detailed information about patients suffering from various lung conditions. It includes:')
            st.caption('**Age & Gender**: Patient demographics to understand the spread across age groups and gender.')
            st.caption('**Smoking Status**: Whether the patient is a smoker or non-smoker.')
            st.caption('**Lung Capacity**: Measured lung function to assess disease severity.')
        st.caption('**Disease Type**: The specific lung condition, like COPD or Bronchitis.')
        
        with col2:
            st.caption('**Treatment Type**: Different treatments patients received, including therapy, medication, or surgery.')
            st.caption('**Hospital Visits**: Number of visits to the hospital for managing the condition.')
            st.caption('**Recovery Status**: Indicates whether the patient recovered after treatment.')
        
        url = 'https://www.kaggle.com/datasets/samikshadalvi/lungs-diseases-dataset'
        st.caption('For more details, check out the the original [source](%s) of the dataset.' % url)
    
    with pred_tab:
        st.header('Prediction Tab')
        st.caption('In this tab, our ML models will predict if you will recover based on your data.')
        
        models = load_models()
        
        model = st.selectbox('Select preferred model for prediction', models.keys())
        model_predictor = models[model]
        
        col1, col2 = st.columns(2)        
        
        
        with col1:
            # age
            age = st.number_input('What is your age?', min_value=0, max_value=100)
            gender = st.radio('What is your gender?', df['Gender'].unique())
            disease = st.selectbox('What is your lung condition?', df['Disease Type'].unique())
            treatment = st.selectbox('Which treatment did you receive?', df['Treatment Type'].unique())
        
        with col2:
            visits = st.number_input('How many times do you visit the hospital? (Annually)', min_value=0, max_value=365)
            capacity = st.slider('What is your lung capacity?', min_value=1.00, max_value=df['Lung Capacity'].max()+5)
            smoke = st.radio('Do you smoke?', ['Yes', 'No'])

        if st.button('Predict!'):
            pred = prediction(model_predictor, age, gender, smoke, capacity, disease, treatment, visits)
            rec = 'Recovered!' if pred == 1 else 'I am sorry.'
            st.header(rec)
            
    with data_viz:
        st.title('Data Viz Tab')
        st.caption('In this tab, we can visualize the relationships among our data.')
        st.caption('See our pre-existing plots and you can also plot your own!')
        
        dviz_tab1, dviz_tab2 = st.tabs(['Plots', 'Custom Plot'])
        
        with dviz_tab1:
            st.title('Feature Distribution and Relationships')
            st.caption('In this tab we will see the feature distributions of the dataset.')
            st.caption('We can see the relationships of the features among each other.')

            # Create subplots
            fig, axes = plt.subplots(nrows=4, ncols=2, figsize=(15, 25))

            # Age distribution
            axes[0, 0].hist(df['Age'])
            axes[0, 0].set_xlabel('Age')
            axes[0, 0].set_ylabel('Frequency')
            axes[0, 0].set_title('Age Distribution')

            # Lung Capacity distribution
            axes[0, 1].hist(df['Lung Capacity'])
            axes[0, 1].set_xlabel('Lung Capacity')
            axes[0, 1].set_ylabel('Frequency')
            axes[0, 1].set_title('Lung Capacity Distribution')

            # Hospital Visits distribution
            axes[1, 0].hist(df['Hospital Visits'])
            axes[1, 0].set_xlabel('Hospital Visits')
            axes[1, 0].set_ylabel('Frequency')
            axes[1, 0].set_title('Hospital Visits Distribution')

            # Gender vs Recovered
            count_data = df.groupby(['Gender', 'Recovered']).size().unstack(fill_value=0)
            count_data.plot(kind='bar', stacked=False, ax=axes[1, 1])
            axes[1, 1].set_xlabel('Gender')
            axes[1, 1].set_ylabel('Count')
            axes[1, 1].set_title('Gender Count by Recovery')
            axes[1, 1].legend(title='Recovered')

            # Smoking vs Recovered
            count_data = df.groupby(['Smoking Status', 'Recovered']).size().unstack(fill_value=0)
            count_data.plot(kind='bar', stacked=False, ax=axes[2, 0])
            axes[2, 0].set_xlabel('Smoking Status')
            axes[2, 0].set_ylabel('Count')
            axes[2, 0].set_title('Smoking Status by Recovery')
            axes[2, 0].legend(title='Recovered')

            # Disease Type vs Recovered
            count_data = df.groupby(['Disease Type', 'Recovered']).size().unstack(fill_value=0)
            count_data.plot(kind='bar', stacked=False, ax=axes[2, 1])
            axes[2, 1].set_xlabel('Disease Type')
            axes[2, 1].set_ylabel('Count')
            axes[2, 1].set_title('Disease Type by Recovery')
            axes[2, 1].legend(title='Recovered')

            # Treatment Type vs Recovered
            count_data = df.groupby(['Treatment Type', 'Recovered']).size().unstack(fill_value=0)
            count_data.plot(kind='bar', stacked=False, ax=axes[3, 0])
            axes[3, 0].set_xlabel('Treatment Type')
            axes[3, 0].set_ylabel('Count')
            axes[3, 0].set_title('Treatment Type by Recovery')
            axes[3, 0].legend(title='Recovered')

            # Disease Type vs Treatment Type
            count_data = df.groupby(['Disease Type', 'Treatment Type']).size().unstack(fill_value=0)
            count_data.plot(kind='bar', stacked=False, ax=axes[3, 1])
            axes[3, 1].set_xlabel('Disease Type')
            axes[3, 1].set_ylabel('Count')
            axes[3, 1].set_title('Disease Type by Treatment Type')
            axes[3, 1].legend(title='Treatment')
            
            st.pyplot(fig)
            plt.tight_layout()

        with dviz_tab2:
            x = st.selectbox("Choose X for plotting.", tuple(df.columns))
            y = st.selectbox("Choose Y for plotting.", tuple(df.drop(x, axis=1).columns))
            
            plot = st.selectbox("Select type of plot.", ("Scatter", "Bar", "Line"))
                
            if st.button("Plot X and Y!"):
                if plot == "Scatter":
                    st.scatter_chart(
                        data=df,
                        x=x,
                        y=y,
                        size='Recovered'
                    )
                elif plot == "Bar":
                    st.bar_chart(
                        data=df,
                        x=x,
                        y=y
                    )
                elif plot == "Line":
                    st.line_chart(
                        data=df,
                        x=x,
                        y=y
                        )
    
if __name__ == "__main__":
    main()