Research on Guangxi Yulin's Poverty Return Identification Based on Integrated Learning

2020 is the end of the poverty alleviation exit effectiveness inspection. To consolidate the achievements of poverty alleviation, promote the Rural Revitalization Strategy, and prevent poverty return will be its key means and important guarantee. At present, although many experts and scholars have studied the related fields of dynamic monitoring of poverty return, most of them focus on macro policy content, and the research on poverty return identification for the poor is still relatively lacking. At the same time, the in-depth mining and application of the accumulated large-scale targeted poverty alleviation data is insufficient. This paper finds relevant literature and combines China's achievements in precise poverty alleviation work and the construction of precise poverty alleviation systems, using Yulin's poor household record-keeping data as the research object. The three integrated learning algorithms, XGBoost, LightGBM and CatBoost, were used to identify poor households in three categories, out of poverty, poor and returning to poverty. The results show that the XGBoost algorithm model achieves optimal results, with an overall accuracy of 95.55% for three classification recognition, and can better identify the poor categories of the poor population. At the same time, the model's feature importance ranking can reveal the importance of poverty-causing characteristics of poor households in Fuping County. By analyzing the current work flow of helpers, it is clear that good dynamic monitoring of poverty return should achieve the needs of dynamic updating of data, dynamic feedback of results and dynamic adjustment of countermeasures.