Exploring Machine Learning Models for Recurrence Prediction in Lung Cancer Patients

Abstract

Background: A proper assessment for the probability of recurrence in lung cancer is mandatory for a clinician to make an effective treatment-decision. Materials and Methods: Here, we employed machine learning algorithms to predict the lung cancer recurrence rate using the Caribbean and few white ethnicities populations. A 100 metastatic record with 15 predictor variables and 1 dependent variable was considered for model development. These models were evaluated using seven performance metrics, including accuracy and F1 score. Results: Our study results show that the decision tree outperformed the other models with the highest accuracy and F1 score of about 0.95 and 0.90, respectively. Of note, the p-value and correlation matrix show that the most significant features accounting for the tumor recurrence are cancer stage, ethnicity, tumor size, genome doubled and time to recurrence. Conclusion: Thus, our study provides insights into implementing machine learning algorithms to evaluate cancer outcomes in a clinical setting.