Breast Cancer Prediction

• Dataset.
• The dataset can be found in folder data

Model Building

Refer to the file model/main.py

It imports the necessary dependencies, including pandas for data manipulation, scikit-learn for data preprocessing and modeling, and pickle for serializing the trained model and scaler.

1. The create_model function is defined, which takes the input data and performs the following steps:

• Separates the independent variables (features) and the dependent variable (diagnosis).
• Applies feature scaling using StandardScaler to standardize the feature values.
• Splits the data into training and testing sets using the train_test_split function.
• Creates a logistic regression model using LogisticRegression from scikit-learn.
• Trains the model using the training data.
• Makes predictions on the test data and evaluates the model's performance by calculating the accuracy and generating a classification report.

2. The clean_data function is defined, which performs some basic data cleaning steps:

• Reads the breast cancer dataset from a CSV file.
• Removes unnecessary columns (Unnamed: 32 and id) from the dataset.
• Maps the diagnosis column values (M for malignant, B for benign) to binary labels (1 for malignant, 0 for benign).

3. The main function is defined, which serves as the entry point of the script:

• Calls the clean_data function to obtain the cleaned dataset.
• Calls the create_model function, passing the cleaned dataset.
• Serializes the trained model and scaler objects using pickle and saves them as files ('model.pkl' and 'scaler.pkl').
• Finally, the main function is called when the script is executed.

Streamlit App

Refer to the file app/main.py

It imports the necessary dependencies, including Streamlit for creating the web application, pickle for loading the trained model and scaler, pandas for data manipulation, NumPy for numerical operations, and Plotly for creating radar charts.

• The get_clean_data function reads the breast cancer dataset from a CSV file, performs data cleaning by dropping unnecessary columns, and maps the diagnosis column values to binary labels (the same function in the modelling step).
• The add_sidebar function creates a sidebar in the Streamlit application, which displays sliders for user input of cell nuclei measurements. The function retrieves the cleaned data and defines a list of slider labels and keys for each measurement.
• The get_scaled_values function scales the user input values based on the range of the corresponding measurement in the original dataset. It uses the cleaned dataset and the input dictionary to calculate scaled values for each measurement.
• The get_radar_chart function creates a radar chart using Plotly. It takes the scaled input data, defines categories for the chart, and plots three radar traces representing different types of measurements: mean, standard error, and worst.
• The add_prediction function loads the trained model and scaler using pickle. It transforms the scaled user input into the appropriate format and makes a prediction using the logistic regression model. The function displays the predicted diagnosis (benign or malignant) and the probabilities associated with each class.
• The main function sets the configuration of the Streamlit application, including the page title, icon, layout, and initial sidebar state. It also includes custom CSS styling for the application. Inside the main function, the user input data is collected using the add_sidebar function, and a container is created to display the application title and description.
• The application layout is divided into two columns. In the first column, the radar chart is displayed based on the user input data, using the get_radar_chart function. In the second column, the diagnosis prediction and associated probabilities are shown using the add_prediction function.
• Finally, the main function is called to run the Streamlit application.

Note: This block of CSS code in assets/style.css defines some custom styles that are used in the Streamlit application for breast cancer prediction.

App Link