Rare Potential Poor Household Identification With a Focus Embedded Logistic Regression

With the rapid development of poverty alleviation in China, multidimensional identification has always been challenging. This paper adopted a focus embedded logistic regression (FeLR) to solve two types difficulties–the rarity and hard-distinguishability, potential poor household (PPH) identification. The PPH was decomposed into subproblems–the re-poverty (PRPH) identification, unidentified (PUPH) FeLR focal loss deal with weighting technique address rarity. sample weight exponent extended negative values overlook hard samples. setting significantly improved recall PPHs, compared that using traditional regression. A few indicators were critical incidence PPH, especially income per capita, medical expenses for chronic diseases, house structure. Local policy makers are suggested pay more attention crucial against contrapuntally.