Feature based heart disease prediction approach
Download as PPTX, PDF0 likes9 views
Feature based heart prediction
Feature based heart disease prediction approach
- 1. FEATURE BASE HEART DISEASE PREDICTION USING RFCNN ADVANCE MACHINE LEARNING ALGORITHMS STUDENT DETAILS: QAMAR BEGUM ROLL NO. 161022742009 M. E, COMPUTER SCIENCE PROJECT GUIDE: Dr.MOHAMMED SANAULLAH QASEEM PROFESSOR & HOD CSE DEPARTMENT
- 2. USING RFCNN ADVANCE MACHINE LEARNING ALGORITHM FEATURE BASE HEART DISEASE PREDICTION
- 3. TABLE OF CONTENT • Abstract • Existing system • Disadvantages • Proposed System • Libraries • Test Accuracy • Results • Application • Conclusion • Future Scope • Literature survey • System requirement • Dendrogram • Use case Diagram • System Implementation
- 4. ABSTRACT • Heart disease is one of the leading causes of death worldwide, • Predictive healthcare solutions frequently use machine learning (ML) techniques. But the accuracy of the majority of conventional ML models is less than 80%. • This procedure serves as a feature selector, identifying the most important features from the dataset and eliminating irrelevant or noisy ones. • Results will be useful to recognize patterns, correlations, and interactions among the selected features.
- 5. EXISTING SYSTEM • Lack of feature selection and data representation limits their effectiveness. • Most models are sensitive to noise and overfitting. • Accuracy is often below 80%, which is insufficient for critical medical predictions. • Less work on analysis identifying the most important features from the dataset
- 6. DISADVANTAGES • Inadequate feature selection results in poor data representation. • Limited ability to capture non-linear patterns in data. • High sensitivity to noise and outliers in the dataset. • Overfitting to training data, leading to poor generalization.
- 7. PROPOSED SYSTEM • We will be understanding advanced techniques to address the limitations of existing methods • We will be able to understand predictions from multiple models to improve robustness. • Algorithm designed for high accuracy and scalability. • Ensures diversity in predictions by splitting data more randomly. • This ensemble approach leverages the strengths of each algorithm to deliver higher accuracy, precision, and recall for heart disease prediction
- 8. Training Data Instance Training each decision tree on a random subset Bagging (majority) Prediction Output ............... Class A Class A class B pOutput RF ALGORITHMS
- 9. Step 1: Select random K data points from the training set. Step 2:Build the decision trees associated with the selected data points(Subsets). Step 3:Choose the number N for decision trees that you want to build. Step 4:Repeat Step 1 and 2. Step 5: For new data points, find the predictions of each decision tree, and assign the new data points to the category STEPS
- 10. DECISION TREE
- 11. input layer pooling Layer convolution layers flatten layer output layer CNN LAYERS
- 12. Feature Selection Data Preprocessing Training set Testing set Heart disease data set Data Splitting Model Training RF CNN RFCNN Trained Model Model Evaluation Heart Disease severity Prediction
- 13. LIBRARIES • Tensorflow-Building & training neural network • Numpy -Numerical computations • Flask- lightweight Python web framework • Keras- implementation of neural networks easy • sklearn-Data preprocessing, evaluation • matplotlib-Data visualization
- 14. LITERATURE SURVEY PAPER TITLE AND YEAR ALGORITHMS ADVANTAGES LIMITATIONS Deep Learning Neural Networks for Predictive Healthcare 2020 Deep Learning Neural Networks (DLNN) • High accuracy due to automated feature extraction. • Computationally expensive. • Requires extensive labeled data for training. Genetic Algorithm and TRFNN for Optimized Healthcare Predictions 2021 Genetic Algorithms (GA) with Transfer Function Neural Networks (TRFNN) • Improved feature selection enhances model accuracy. • Adaptive learning process. • Computational overhead due to GA. • Risk of local optima in feature selection.
- 15. LITERATURE SURVEY PAPER TITLE AND YEAR ALGORITHMS ADVANTAGES LIMITATIONS Hybrid SCNN-SVM for Image- Based Disease Detection 2022 Sparse Convolutional Neural Network (SCNN) with Support Vector Machines (SVM) • Reduces computational cost with sparse representations . • Limited scalability for very large datasets. • Complexity in model integration. PFFBPNN for Predictive Analytics in Healthcare 2020 Partial Feedback Feedforward Backpropagation Neural Networks (PFFBPNN) • Reduces training time compared to standard backpropagatio n. • .Sensitive to hyperparameter tuning. • Requires large memory resources for computation.
- 16. LITERATURE SURVEY PAPER TITLE AND YEAR ALGORITHMS ADVANTAGES LIMITATIONS IoT-Enabled Random Forest Model for Remote Health Monitoring 2021 IoT and Random Forest for real- time monitoring • Handles high- dimensional data effectively. • Limited by IoT device bandwidth. • Dependent on stable internet connectivity
- 17. SOFTWARE REQUIREMENTS HARDWARE REQUIREMENTS • OS: WINDOWS • PYTHON : PYTHON 3.X AND ABOVE • SETUP TOOLS AND PIP TO BE INSTALLED FOR 3.6.X AND ABOVE •RAM : 1GB AND HIGHER •Processor : Intel i3 and above •Hard Disk : 5GB minimum is required
- 18. DENDROGRAM Heart Disease Clinical Data Data Processing Feature selection Feature Normalization Classification Heart Disease Absent Heart disease present
- 19. SYSTEM IMPLEMENTATION RFCNN Algorithm Steps: • Hybrid Approach • Feature Reduction • Combination of Strengths • Feature Selection process
- 20. HYBRID APPROACH Random Forest serves as a feature selector. • Identifying the most important features. • Eliminating irrelevant or noisy ones. . CNN is used as the classification engine. • Recognize patterns • Recognize correlations • Interactions among the selected features
- 21. p(2 ) p(1 ) p(3 ) p(4 ) p(2 ) p(1 ) p(3 ) p(4 ) P(1)= P(2)= p(3)= P(4)= Final Prediction {p(1),P(2),p(3),p(4)} Instance Machine Learning Deep Learning RF CN N RF + PCNN) P ( RF + PCNN) P ( RF + PCNN) P ( RF + PCNN) P ( /2 /2 /2 /2
- 22. FEATURE REDUCTION • Random Forest calculates feature importance scores and selects the top-N features • This reduces dimensionality and improves Convolution Neural Network efficiency by focusing only on meaningful data.
- 23. COMBINATION OF STRENGTHS • Random Forest is robust to overfitting due to its ensemble structure, making it effective for feature selection. • The hybrid model can manage complex relationships in the data since CNN is excellent at learning complex, hierarchical patterns.
- 24. FEATURE SELECTION Step:1 Train the Random Forest Model step:2 Compute Feature Importance Step:3 Rank Features by Importance Step:4 Select Top Features Step:5 Feature Subset Input: The complete dataset with all features and the target variable.
- 25. RESULTS
- 26. TEST ACCURACY RF CNN Accuracy Graph 83 percent accuracy RFCNN compared with CNN
- 27. SPLIT-80:20 80 20 15 90 confusion matrix for RCNN Predicted Label Expected label 100 80 60 40 20 0 No Heart Disease Heart Disease No Heart Disease Heart Disease
- 28. APPLICATIONS Accurately detect abnormalities Analyze vast amounts of patient data Predict the likelihood of developing certain cardiovascular diseases
- 29. CONCLUSION • Prediction of heart diseases is done successfully in achieving superior performance compared to other machine learning models • 80/20 split dataset gives highest accuracy of impressive 83% underscoring its effectiveness in heart disease prediction • RFCNN's capability to handle complex medical datasets and provide reliable predictions • The Flask-based implementation ensures a user-friendly interface for medical professionals and patients, with seamless integration into healthcare systems.
- 30. FUTURE SCOPE Enhance the model's performance by incorporating additional features like genetic data or lifestyle factors. Expand the system to predict other cardiovascular conditions for broader applicability. Explore the integration of blockchain for secure and decentralized patient data management. Collaborate with healthcare providers to validate the system in real-world clinical environments
- 31. THANK YOU Any Questions ?