Proposal Presentation

Heart Disease Prediction

Predicting Heart Disease Risk Using Health & Lifestyle Indicators

Group ID: 2026-Y2-S1-MLB-B7G2-10

IT25101623 Hathurusinghe H.D.V.

IT25101379 Perea S.K.A.N.I

IT25102452 Herath H.M.C.K

IT25103288 Namarathna N.B.A.D.M.A

IT25103491 Sathursan S.

IT25100474 Gunawardhana K.C

01

Assigned Dataset

Heart Disease

by Oktay Rdeki

Platform

Kaggle Datasets

Direct link

kaggle.com/datasets/oktayrdeki/heart-disease

Format

Single CSV file, tabular structured data

Quick Facts

~10,000

patient records (rows)

21

columns (20 features + 1 target)

1

binary target: Heart Disease Status

02

Problem Domain

Domain: Healthcare – Preventive Cardiology & Clinical Risk Screening

Primary Use Case

Predicting whether an individual is at risk of heart disease to support earlier screening and preventive intervention, before symptoms become severe.

Secondary Use Case

Identifying which clinical and lifestyle factors e.g. cholesterol, BMI, smoking, stress most strongly influence heart disease risk.

Why It Matters

Cardiovascular disease is a leading cause of death worldwide. Clinicians and health programs need quick, consistent, data-driven ways to flag at-risk individuals using measurable health and lifestyle indicators.

03

Dataset Structure

Features, target variable, and grouping

Demographic

Age
Gender

Clinical Measurements

Blood Pressure
Cholesterol Level
BMI
Triglyceride Level
Fasting Blood Sugar
CRP Level
Homocysteine Level
High Blood Pressure (Y/N)
Low HDL Cholesterol (Y/N)
High LDL Cholesterol (Y/N)

Lifestyle Factors

Exercise Habits (Low/Med/High)
Smoking (Y/N)
Alcohol Consumption
Stress Level
Sleep Hours

Medical History

Family Heart Disease (Y/N)
Diabetes (Y/N)

04

Dataset Justification

Features & Target

Key Features

age, gender, blood pressure, cholesterol level, BMI, triglyceride level, fasting blood sugar, CRP level, homocysteine level, exercise habits, smoking, alcohol consumption, stress level, sleep hours, family heart disease, diabetes

Target Variable

heart disease status (binary classification)
directly answers the prediction problem

Dataset Size

≈ 10,000 rows sufficient for robust training and validation splits

Quality Considerations

Missing Values

To be confirmed during EDA; the dataset author reports clean, consistently-coded data

Class Balance

Yes/No split in Heart Disease Status to be verified from the CSV before modelling

Derived Flags

High/Low HDL & LDL and High Blood Pressure are pre-derived Yes/No flags check for redundancy with raw values

Dataset Limitations

Missing Data

Will inspect the CSV and impute or drop, not ignore

Imbalance Target

Will use F1 / precision / recall alongside accuracy, not accuracy alone

Sensitive Attributes (age, gender)

Fairness will be assessed, not just accuracy

Data Provenance

Not a verified clinical record system findings will be treated as exploratory, not diagnostic

06

Planned Machine Learning Workflow

▲ Highlighted stages represent the core focus of this project proposal.

01

Problem Definition

Define the heart disease prediction problem
Identify healthcare risk screening as the objective

02

Data Collection

Use the assigned heart disease dataset
Review dataset source, features, and target variable

03

Data Preprocessing

Handle missing values
Detect outliers
Encode categorical variables
Scale numerical features

04

Feature Engineering

Select relevant features
Analyze feature relationships
Prepare features for modelling

05

Model Selection

Select suitable machine learning algorithms based on the problem type
Compare multiple candidate models
Select the best-performing model using evaluation metrics

06

Model Training

Train models using prepared data
Apply hyperparameter tuning

07

Model Evaluation

Use accuracy, precision, recall, F1-score, ROC-AUC
Perform k-fold cross-validation
Compare models and select the best performer

08

Deployment (Future Step)

Deploy the selected model for real-world prediction support

05

Planned Use of AI Tools

AI Tool	Planned Use	Usage Level
ChatGPT / Copilot	Coding assistance Debugging Explaining ML concepts	Moderate
Claude	Report drafting Editing technical explanations	Moderate
Gemini	Literature search Summarising references	Limited

All AI-assisted outputs will be reviewed and verified by the group before inclusion in the final project. The final AI Tool Usage Declaration will accurately reflect the tools used.

Thank You

Questions & Discussion

Dataset available at:

kaggle.com/datasets/oktayrdeki/heart-disease