UNIVERSITY INSTITUTE OF COMPUTING
An Overview
Bachelors of Computer Applications
Major Project
of Computing
22CAR-353
Smart Carezone for Disease Prediction
& Rubrics 3
Career
Student Name: Ayush Kumar Chaudhary, Akarsh Upadhyay, Jishanu Panday Supervisor Name: Ms. Shivani Chadha
UID : 22BCA10216, 22BCA10033, 22BCA10020 Designation: Asst prof
Section/Group: PH22BCA 10(A)
Planning DISCOVER . LEARN . EMPOWER
1
Presentation Outline
• Title & Introduction
• Progress vs. Initial Plan
• Preliminary Results or Prototype
• Methodology Refinement
• Challenges & Solutions
2
Title & Introduction
We are creating a smart system that uses computer data to predict if someone might
have a serious disease before it worsens. By focusing on four common diseases—
Diabetes, Heart Disease, Parkinson’s, and Breast Cancer—we ensure our
solution tackles health issues that affect many people. We begin by collecting
trustworthy health data, then build computer models that learn from this data to
make accurate predictions, and finally design an easy-to-use website. This
approach helps catch diseases early, enabling faster and better healthcare decisions.
Progress vs. Initial Plan
What We Planned:
✅ Collect medical data from reliable sources.
✅ Train smart computer models to predict four diseases.
✅ Create a simple website where users can enter their health data and get predictions.
What We Have Achieved:
✔️Collected high-quality medical datasets from trusted sources like Kaggle and UCI.
✔️Trained our first set of machine learning models to predict diseases.
✔️Built a basic web application (using Streamlit) where users can test the system.
What’s Left to Do:
🔹 Improve our models to make predictions even more accurate.
🔹 Enhance the website to make it easier and more user-friendly.
Preliminary Results or Prototype
Prototype Demo:
We have built a simple web interface using Streamlit where users can input their
medical details.
📸 We will show screenshots of the web app to demonstrate how it works.
Model Performance:
📊 Our trained machine learning models have started making predictions.
📈 Early accuracy results and confusion matrices give an idea of how well the models
are working.
Key Insights:
🔹 The models are already showing promising results, but some need improvement.
🔹 Certain diseases are easier to predict accurately than others.
🔹 More data cleaning and fine-tuning can help improve accuracy.
Preliminary Results or Prototype
Methodology Refinement
Refined Data Preprocessing:
✅ Better Handling of Missing Data: We now use smarter techniques like mean/median
imputation.
✅ Improved Normalization: We ensure all features are on a similar scale for better model
performance.
✅ Feature Selection: We identified the most important medical parameters to focus on.
Model Updates:
🔄 Algorithm Adjustments: We tested new models and improved existing ones.
⚙️Hyperparameter Tuning: We fine-tuned settings to improve accuracy.
🤖 Ensemble Methods: We combined multiple models for better predictions.
Why We Made These Changes:
📊 Better Performance: The models now give more accurate predictions.
📉 Data Limitations: We adapted our methods based on the quality and size of our dataset.
⚡ Faster Processing: Optimized methods allow quicker results without losing accuracy.
Challenges & Solutions
🚧 Challenges:
Poor Data Quality: Missing values and inconsistent data affected accuracy.
Model Accuracy Issues: Some models didn’t perform well in predictions.
Slow Processing: Large datasets caused delays in training and predictions.
💡 Solutions:
Better Data Handling: Used imputation techniques to fill missing values.
Hyperparameter Tuning: Adjusted model settings to improve accuracy.
Optimization: Reduced data size and improved code efficiency for faster results.
📖 Lessons Learned:
High-quality data is key to better predictions.
Fine-tuning models makes a big difference.
Efficient coding helps speed up processing. 🚀
THANK YOU
