PROJECT TITLE :

A heterogeneous ensemble learning method for neuroblastoma survival prediction

ABSTRACT:

Neuroblastoma is a form of childhood cancer that has a high fatality and incidence rate. An accurate prognosis of the length of survival for patients diagnosed with neuroblastoma is an essential component in the process of developing treatment plans. In this study, we proposed a heterogeneous ensemble learning method to predict the survival of neuroblastoma patients. Additionally, we extracted decision rules from the proposed method in order to assist medical professionals in making decisions. Following the completion of the data preprocessing phase, five different kinds of heterogeneous base learners were created. These learners included a decision tree, a random forest, a support vector machine based on a genetic algorithm, an extreme gradient boosting machine, and a light gradient boosting machine. Following this, a heterogeneous feature selection method was developed in order to obtain the optimal feature subset of each base learner. The optimal feature subset of each base learner then guided the construction of the base learners using their a priori knowledge. In addition, it was suggested to integrate the five different types of base learners by using an ensemble mechanism that was based on the area under the curve. In conclusion, the proposed method was evaluated alongside other mainstream machine learning methods using a variety of indicators, and valuable information was extracted from the proposed method by utilizing the partial dependency plot analysis method and the rule-extracted method. According to the findings of the experiments, the proposed method achieves a recall rate of 91.14%, an accuracy rate of 91.64%, and an area under the curve (AUC) of 91.35%, making it significantly superior to the conventional machine learning methods. In addition, the proposed method allows for the extraction of interpretable rules that have an accuracy of 0.900 or higher, as well as predicted responses. Our research has the potential to effectively improve the functioning of clinical decision support systems, which in turn can lead to an increase in the number of neuroblastoma patients who survive their disease.