Current Research
Explainable AI for Early Detection of Diabetic Retinopathy
Research Area
Research Problem
Diabetic retinopathy is a leading cause of preventable blindness worldwide, and early detection is critical to prevent vision loss. Many deep learning diagnostic models function as black boxes, which makes their predictions difficult for clinicians to trust, verify or act on.
Objectives
Build a diabetic retinopathy detection model that classifies retinal images accurately while producing explanations a clinician can actually interpret and use in a diagnostic decision.
Previous Research
Predicting Urban Air Quality Using Machine Learning Models: India, 2015–2020
Dataset
"Air Quality Data in India (2015–2020)" (Kaggle): PM2.5, PM10, NO2, CO and O3 concentrations with timestamps, across several Indian cities.
Methodology
Missing values replaced by the mean, outliers removed via the interquartile range method, then MinMax normalization, before training and comparing a Random Forest regressor against an LSTM network to predict PM2.5 concentration.
Models
Results
LSTM outperformed Random Forest on both error metrics (RMSE 14.50 vs 15.23, MAE 9.85 vs 10.45), showing a better grasp of long-term pollution trends, while Random Forest stayed competitive for fast, interpretable predictions.