Exploratory Data Analysis & Model Evaluation
Deep dive into the OASIS-1 Dataset, feature relationships, and Phase 1 vs Phase 2 modeling results.
Class Distribution
Shows the dataset balance between Demented and Non-Demented patient scans.
Missing Values by Feature
Highlights variables requiring imputation, primarily Socioeconomic Status (SES).
Feature Distributions
Visualizes clinical feature overlaps and separations between patient groups.
Feature Correlation Matrix
Maps statistical relationships, revealing strong ties between MMSE and CDR.
Key Indicators by CDR Group
Demonstrates how MMSE drops and brain volume shrinks as Dementia progresses.
Age Demographics
Shows age distributions, emphasizing Alzheimer's prevalence in older patients.
Phase 1 Baseline Accuracy
Compares clinical-only models, showing XGBoost and Random Forest superiority.
Feature Ablation Study
Proves that adding tissue and regional MRI features massively improves accuracy.
Feature Importance Across Models
Detailed breakdown of exactly which clinical variables the trees rely on most.
