Telecom Big Data assisted Identification Algorithm for Poverty Stricken Students in Colleges

Abstract

The identification and financial aid for poverty stricken students in colleges is significant for poverty alleviation and education equity while the traditional identification method based on voluntary reporting or subjective factors is not accurate enough. In this paper, we propose an identification architecture for poverty stricken students in colleges based on telecom big data and XGBoost (eXtreme Gradient Boosting) Algorithm. XGBoost is an ensemble learning algorithm while its ordinary parameters adjusting algorithm cannot improve the performance of this model significantly. Thus we propose an algorithm of parameters adjusting based on QBFO (Quantum Bacterial Foraging Optimization), called QBFO-XGBoost, improving the performance of XGBoost. The experimental results show that the proposed QBFO has advantages of both convergence accurate and convergence rate compared with other swarm intelligence algorithms. In addition, QBFO-XGBoost applied in identification for poverty stricken students in colleges proves higher recall and precision compared with XGBoost based on grid parameter adjustment method.