Abstract:
Early and accurate cancer detection is crucial for improving patient outcomes from this
lethal disease, but traditional methods often lack sensitivity or specificity. Machine learning
algorithms, particularly Random Forests, offer promising tools for analyzing medical data
and achieving this goal. However, the performance of Random Forests is heavily dependent
on the appropriate configuration of hyperparameters, requiring an optimal configuration for
accurate prediction. This study proposes a novel multi-objective hyperparameter tuning
strategy to enhance the effectiveness of Random Forests for cancer detection. The research
starts off with single-goal hyperparameter tuning targeted on accuracy, observed by an
evaluation of performance the usage of five-fold move-validation. The results show a crossvalidated training accuracy of 0.96 and a test accuracy of 0.97, effectively addressing the
issue of overfitting. Subsequently, the methodology advances a state-of-the-art optimization
algorithm, inclusive of a multi-objective algorithm or a particle swarm optimization, to
explore the hyperparameter area efficiently. The proposed strategy aims to construct a
Random Forest model that not only delivers accuracy but also maintains equilibrium across
diverse performance aspects in cancer detection. To reap this, a multi-goal optimization
algorithm is integrated with the hyperparameter tuning technique, enabling the exploration
of numerous solutions across the Pareto front. This approach enhances the version's
potential to parent diffused patterns indicative of cancerous conditions whilst minimizing
false positives and fake negatives. The proposed multi-objective hyperparameter tuning
approach for Random Forest models is a significant development in the field of cancer
detection. In the end, it uses machine learning to improve healthcare outcomes by paving
the way for more precise, understandable, and clinically relevant cancer diagnoses. It also
highlights the significance of thorough hyperparameter optimization techniques in boosting
model efficacy.
