مهدي علي محمدي

Cancer remains one of the leading causes of global mortality, significantly impacting public health and underscoring the need for effective predictive models. Despite ongoing advancements in identifying early signs and developing treatments, high mortality rates persist. This study investigates the effectiveness of various machine learning algorithms in predicting cancer mortality and incidence, addressing the question: Can a stacking model improve predictive accuracy over traditional models? The research problem focuses on optimizing model performance while considering execution time and complexity. A comparative analysis was conducted using a range of algorithms, including decision trees, k-nearest neighbors, logistic regression, support vector machines, naive Bayes, neural networks, random forests, gradient boosting, and gradient-boosting trees. The study eva‎luated these models as base and meta-learners to assess performance metrics, including accuracy and execution time. The newly developed stacking model achieved an accuracy of 87.45% in predicting mortality rates, slightly surpassing the 87.43% of gradient-boosting trees. However, the increase in accuracy was minimal relative to the added complexity and longer execution time of the stacking model, suggesting that simpler models may yield more reliable results efficiently. In predicting cancer incidence, the stacking model demonstrated poorer performance compared to conventional base models and exhibited signs of overfitting, likely attributed to imbalanced data. The findings indicate that while advanced models like the stacking model can offer slight improvements in certain contexts, they may not provide sufficient justification for their complexity. This research advocates for the use of simpler models to achieve reliable outcomes in a more time-efficient manner and highlights the need for further investigation into addressing data imbalances in cancer incidence prediction.

پيش بيني مرگ , پيش‌بيني ابتلاء به سرطان , يادگيري ماشين , طبقه بندي , كوهورت

Mortality prediction , Cancer incidence prediction , Machine Learning , Classification , Cohort

Mahdi Alimohammadi

Dr. Ghousi - Dr. Aliha