Hybrid Machine Learning Model For Multi-diseases Diagnosis

• Author(s):

  Tharun C | Dr.K.Annalakshmi

• Keywords:

  Multi-Disease Diagnosis, Hybrid Machine Learning, Ensemble Learning, Data Fusion, Health Informatics, Random Forest, Support Vector Machine, XGBoost, Streamlit Dashboard, Smart Health.

• Abstract:

  Co-occurring Chronic Conditions And Complex Multi-disease Pathologies Represent An Escalating Global Public Health And Socioeconomic Crisis. In Rapidly Growing Urban Patient Populations, Overlaps Between Metabolic, Cardiovascular, And Neurological Syndromes Contribute Heavily To Prolonged Diagnostic Latencies, High Multi-clinic Tracking Overheads, Secondary Clinical Omissions, And Therapeutic Cross-remediations. Accurate Tracking, Early Risk Stratification, And Concurrent Non-invasive Classification Of Heterogeneous Patient Clinical Profiles Are Therefore Essential For Modern Preventive Medicine, Automated Out-patient Triage, And Smart City Health System Asset Management. Conventional Clinical Diagnostic Strategies, Including Isolated Domain-specific Laboratory Assessments, Paper-based Single-disease Tracking Indices, And Disjointed Diagnostic Pipelines, Are Often Time-consuming To Execute, Heavily Subjective To Observer Visual Fatigue Or Missing Metadata Segments, And Difficult To Apply Efficiently Across High-throughput Health Networks. To Overcome These Operational And Computational Limitations, This Study Proposes An Automated, Computer-aided Multi-disease Screening Framework Using Multi-parametric Clinical Data Fusion And A Hybrid Machine Learning Ensemble Architecture. Heterogeneous Clinical Records Capturing Signs Of Cardiovascular, Metabolic, And Neurological Anomalies Are Aggregated Into A Single Integrated Schema And Preprocessed Via Automated Median Data Imputation, Structural Outlier Filtering, And Min-max Feature Normalization To Resolve Calibration Variances Across Different Clinical Measurement Devices. Discriminative Physiological Indicators —including Continuous-time Blood Parameters, Resting Electrocardiographic Metrics, Serum Insulin Distributions, Gray-level Texturing Variables, And Patient Demographic Indicators—are Mapped To Establish A Unified High-dimensional Physiological Feature Matrix. Predictive Modeling Is Executed Through A Optimized Hybrid Ensemble Model Combining Three Distinct Baseline Algorithms: Random Forest, Support Vector Machine (SVM), And Extreme Gradient Boosting (XGBoost). The Individual Probabilistic Predictions Are Consolidated Via A Weighted Soft-voting Ensemble Consensus Layer Optimized To Recognize Multi-label Classification Objectives Simultaneously. Model Execution Is Measured Comprehensively Using Standard Validation Metrics, Evaluating Multi-class Classification Accuracy, Precision, Recall, F1-score, And Structural Confusion Matrix Tracking To Verify Diagnostic Consistency. Experimental Results Suggest That The Proposed Hybrid Framework Achieves A Total Predictive Accuracy Of 96.44%, Completely Flattening Cross-domain Validation Errors While Maintaining Execution Runtimes Suitable For Modern Edge-device Implementation. The Complete Framework Is Deployed As An Interactive Software Application Via A Streamlit Web Interface, Integrating Responsive Data Submission Grids, Interactive Risk-distribution Plots, Multi-disease Prediction Charts, And Real-time Critical Health Warnings For Municipal Healthcare Ecosystems.

---

Other Details

• Paper id:

  IJSARTV12I6105698

• Published in:

  Volume: 12 Issue: 6 June 2026

• Publication Date:

  2026-06-18

Download Article