Breast Cancer Prediction Using ANN

Breast Cancer Prediction using Artificial Neural Network (ANN)

Breast Cancer Prediction : Download Dataset 

Problem Statement

Breast cancer is one of the most common cancers among women worldwide. Early and accurate diagnosis is essential for timely treatment and improved survival rates. The Breast Cancer Wisconsin Dataset contains various medical features extracted from cell nuclei of breast mass images. The objective of this practical is to develop an Artificial Neural Network (ANN) model that can classify a tumor as Malignant (M) or Benign (B) based on the given diagnostic features.

Dataset Description

Attribute	Description
Total Records	569
Input Features	30 Numerical Features
Target Variable	Diagnosis (M = Malignant, B = Benign)
Objective	Binary Classification

Research Questions

RQ1: Dataset Understanding

How can the Breast Cancer Wisconsin dataset be loaded and understood for breast cancer prediction?

---

RQ2: Data Preprocessing

How can the dataset be preprocessed by handling missing values, duplicate records, and unnecessary attributes to prepare it for Deep Learning?

---

RQ3: Data Transformation

How can categorical variables be converted into numerical form using Label Encoding, and why is it necessary for ANN models?

---

RQ4: Feature Scaling

How does Normalization (or Standardization) improve the performance and convergence of an Artificial Neural Network?

---

RQ5: Feature Selection

Which input features are most relevant for predicting whether a tumor is Malignant or Benign?

---

RQ6: Data Preparation

How can the dataset be divided into input features (X) and the target variable (Y) before training the model?

---

RQ7: Data Splitting

How should the dataset be divided into Training and Testing sets to ensure reliable model evaluation?

---

RQ8: ANN Model Design

How can an Artificial Neural Network be designed by selecting appropriate input, hidden, and output layers?

---

RQ9: ANN Initialization

Which Deep Learning libraries and framework are required to initialize and implement an Artificial Neural Network model?

---

RQ10: Model Compilation

How can the ANN model be compiled by selecting an appropriate Optimizer, Loss Function, and Evaluation Metrics?

---

RQ11: Model Training

How can the Artificial Neural Network be trained using the training dataset by selecting suitable Epochs and Batch Size?

---

RQ12: Model Evaluation

How can the performance of the trained ANN model be evaluated using Accuracy, Precision, Recall, F1-Score, Confusion Matrix, and ROC-AUC Curve?

---

RQ13: Model Selection

How can the trained ANN model be validated and selected as the best model for breast cancer prediction?

---

RQ14: Prediction on New Data

How can the trained Artificial Neural Network be used to predict whether a new patient's tumor is Malignant or Benign?

Objectives

1. To preprocess the Breast Cancer Wisconsin Dataset.
2. To build an Artificial Neural Network for breast cancer classification.
3. To train the model using the extracted diagnostic features.
4. To classify tumors into Malignant and Benign categories.
5. To evaluate the model using performance metrics such as Accuracy, Precision, Recall, F1-score, and Confusion Matrix.

nput Features (All 30 Features)

Sr. No.	Feature Name	Meaning
1	radius_mean	Mean distance from the center to the boundary of the tumor (average radius).
2	texture_mean	Variation in the gray-scale intensity (surface texture).
3	perimeter_mean	Average perimeter (boundary length) of the tumor.
4	area_mean	Average area occupied by the tumor.
5	smoothness_mean	Smoothness of the tumor surface (local variation in radius lengths).
6	compactness_mean	Measure of how compact or dense the tumor shape is.
7	concavity_mean	Degree of concave (inward-curved) regions in the tumor boundary.
8	concave points_mean	Number or proportion of concave portions on the tumor boundary.
9	symmetry_mean	Degree of symmetry of the tumor shape.
10	fractal_dimension_mean	Complexity or irregularity of the tumor boundary.
11	radius_se	Standard error of the radius measurement.
12	texture_se	Standard error of the texture measurement.
13	perimeter_se	Standard error of the perimeter measurement.
14	area_se	Standard error of the area measurement.
15	smoothness_se	Standard error of the smoothness measurement.
16	compactness_se	Standard error of the compactness measurement.
17	concavity_se	Standard error of the concavity measurement.
18	concave points_se	Standard error of the concave points measurement.
19	symmetry_se	Standard error of the symmetry measurement.
20	fractal_dimension_se	Standard error of the fractal dimension measurement.
21	radius_worst	Largest (worst-case) radius observed for the tumor.
22	texture_worst	Largest (worst-case) texture value observed.
23	perimeter_worst	Largest (worst-case) perimeter observed.
24	area_worst	Largest (worst-case) area observed.
25	smoothness_worst	Highest (worst-case) smoothness value observed.
26	compactness_worst	Highest (worst-case) compactness value observed.
27	concavity_worst	Highest (worst-case) concavity value observed.
28	concave points_worst	Highest (worst-case) number of concave points observed.
29	symmetry_worst	Highest (worst-case) symmetry value observed.
30	fractal_dimension_worst	Highest (worst-case) fractal dimension value observed.

Expected Output

Diagnosis	Meaning
M	Malignant (Cancerous Tumor)
B	Benign (Non-Cancerous Tumor)