Abstract

During the proposal phase of my Honours Project, I was advised to reconsider my idea. Integrating genetic susceptibility, environmental exposures such as air pollution, and lifestyle factors like smoking to predict lung cancer risk using machine learning was considered too complex for an undergraduate student. A simpler application-based project would have been a safer choice and potentially less risky for my final grade.

Despite these concerns, I chose to follow my intuition and passion for data science and machine learning. I knew I would be investing two years into this project, and I wanted it to address a real-world health problem with meaningful social impact. Lung cancer remains one of the leading causes of cancer-related deaths worldwide, and existing risk models often assess genetic, environmental, and behavioural factors in isolation—limiting their ability to support early and equitable risk assessment.

In this talk, I will share my journey of designing and building a predictive machine learning framework that combines these diverse risk dimensions. I will discuss key technical challenges such as multi-source data integration, feature engineering, model selection, evaluation, and the use of explainable AI tools like SHAP and LIME to ensure transparency and clinical relevance. I will also reflect on the ethical considerations of working with sensitive health and genetic data, including fairness, bias mitigation, and data privacy.

Alongside the technical aspects, I will reflect on the personal challenges of self-doubt, imposter syndrome, and navigating uncertainty as a woman in tech—particularly when choosing a complex and unconventional academic path. This session aims to inspire students and early-career professionals to trust their instincts, embrace challenging projects, and pursue work that aligns with both their technical interests and personal values.

This talk combines technical insight with personal experience to show how perseverance, curiosity, and purpose can turn doubt into meaningful impact.