چكيده به لاتين
Abstract: Cancer is a paramount global healthcare concern, and breast cancer ranks among the most prevalent malignancies affecting women. Remarkable advancements in prevention, diagnosis, and treatment modalities have led to a growing number of breast cancer survivors. Consequently, individuals diagnosed with breast cancer face extended periods of coping with the disease and its repercussions, underscoring the critical importance of their quality of life. This study delves into comprehensive data from 1,114 female breast cancer patients post-treatment in Iran. This dataset spans one year and encompasses patients from four prominent hospitals: Imam Hussein, Imam Khomeini, Mahdieh, and Khatham al-Anbia. Following their treatment, patients completed three vital questionnaires: the EORTC_C30 and EORTC_BR23 questionnaires, assessing their quality of life, and the Schneider questionnaire, gauging their hope for life. The initial analysis involved a rigorous statistical examination of the relationship between hope for life and patients' quality of life. Subsequently, feature selection techniques identified pivotal factors. Revealing variables on general health status, physical condition, functionality, pain levels, sleep patterns, self-efficacy, economic circumstances, and patient-reported symptoms exerted the most significant influence on predicting the quality of life in breast cancer survivors. Given the continuous nature of the response variable, regression models were used during this research. Six distinct models were used, consisting of three fundamental models: linear regression, random forest regression, k-nearest neighbors regression, and decision tree classification and regression, alongside three advanced models: extreme gradient boosting, AdaBoost, and gradient boosting machines. Model fitting commenced, with an initial phase dedicated to determining optimal hyperparameters through an exhaustive search. Subsequently, models were fitted to the training dataset, and then validation and test errors were calculated. Comprehensive model evaluations hinged on critical criteria, including R-squared, Mean Squared Error (MSE), and Mean Absolute Error (MAE). Comparative analysis of results unequivocally underscores the efficacy of advanced models over their fundamental counterparts. Among these advanced models, gradient boosting machines demonstrated the most exceptional performance in this problem.
