Early Prediction of ADHD Symptoms in Children Using Explainable Artificial Intelligence, Machine Learning and Statistical Risk Modelling.

Primary supervisor

Jesmin Nahar

This project uses explainable machine learning, predictive analytics, and statistical risk modelling to predict ADHD symptoms in children using publicly available or clinical healthcare datasets. Students will clean and analyse pediatric health and behavioral data, apply classification and risk stratification algorithms such as Logistic Regression, Random Forest, XGBoost, or multimodal deep learning, and identify key diagnostic risk factors using explainable AI methods. The project aims to show how data-driven methods can help healthcare practitioners better understand symptom trajectories, support early diagnosis, and enable timely clinical interventions.

Title: Early Prediction of ADHD Symptoms in Children Using Explainable Artificial Intelligence, Machine Learning and Statistical Risk Modelling.

Description: This project focuses on analysing pediatric clinical, behavioral, and neuroimaging data to identify distinct risk profiles and early symptom patterns associated with Childhood ADHD. Students will work with public or clinical health datasets (such as behavioral surveys, electronic health records, or neuroimaging) to clean and prepare the data, then apply machine learning models to discover critical risk indicators and predict symptom onset. Depending on the project scope, Honours students will focus on binary classification and post-hoc model explainability (SHAP/LIME) using structured tabular data, while Master's students will implement multimodal dynamic prediction models and time-to-event survival analysis on longitudinal cohorts. The results will help clinicians and researchers develop data-driven screening tools and personalized intervention strategies based on objective patient risk factors.

Aim/outline

Expected Outcomes:

  • Well-defined ADHD risk profiles with transparent clinical and behavioral indicators.

  • An explainable machine learning pipeline and evaluation report offering data-driven insights for early clinical decision support and personalized intervention.

Skills Required: Proficiency in Python or R; basic knowledge of machine learning and statistical modeling algorithms (e.g., Logistic Regression, Random Forest, XGBoost); familiarity with data preprocessing and explainable AI techniques (e.g., SHAP, LIME).

Benefits: This project helps healthcare practitioners improve early screening accuracy and timeliness by identifying key risk factors in children, leading to more objective diagnoses and personalized treatment strategies.

Required knowledge

General Project Pre-requisite Skills and/or Knowledge:

  • Basic understanding of machine learning concepts, particularly classification algorithms (e.g., Logistic Regression, Random Forest, XGBoost) and predictive modeling.

  • Familiarity with Python or R for data handling, statistical analysis, and model development.

  • Basic knowledge of data preprocessing (cleaning, missing value imputation, and dataset preparation).

  • Interest in explainable AI (XAI) and data visualisation (e.g., SHAP, LIME) to interpret clinical risk factors and diagnostic predictions.