In the post-poverty alleviation era, accurately identifying and predicting the risk of returning to poverty of people who have been lifted out of poverty from multiple dimensions is a necessary measure to consolidate the effective connection between the achievements of poverty alleviation and rural revitalization. Based on the data of China Family Panel Studies ( CFPS ) from 2012 to 2018, a multidimensional poverty index is constructed, the multidimensional poverty and return to poverty in rural areas are measured based on the A-F double critical value method, and the contribution of different indicators to multidimensional poverty is discussed through the sum decomposition of indicators. On the basis, VMD and BiLSTM models are used to establish a risk prediction model for returning to poverty. The results show that the problem of multidimensional poverty return in rural China is basically concentrated in two-dimensional or three-dimensional poverty return. The dimension of social development ability has the highest contribution rate to the multidimensional poverty return index, with a contribution rate of 43.12%, followed by the health and education dimensions. In the prediction of the risk of returning to poverty, the overall prediction accuracy of BiLSTM is 88.9%, and the prediction accuracy for individuals at risk of returning to poverty is only 87.6%. The overall prediction accuracy of the AOA-VMD-BiLSTM model for the test set reaches 99.81%, of which the prediction accuracy for individuals with multidimensional poverty-returning risk is 99.6%, and the prediction accuracy for risk-free individuals reaches 100%, indicating that the model can accurately and stably predict the potential poverty-returning risk of multidimensional poverty groups and regions, and provide more accurate data support for poverty governance.
NUSSBAUMERP, BAZILIANM, MODIV. Measuring energy poverty: Focusing on what matters[J]. Renewable and Sustainable Energy Reviews, 2012, 16(1): 231-243.
[16]
ABBASK, LIS X, XUD Y, et al. Do socioeconomic factors determine household multidimensional energy poverty? Empirical evidence from South Asia[J]. Energy Policy, 2020, 146: 111754.
TANGB, LIUY Y, MATTESOND S. Predicting poverty with vegetation index[J]. Applied Economic Perspectives and Policy, 2022, 44(2): 930-945.
[21]
TIANF Y, WUB F, ZENGH W, et al. Identifying the links among poverty, hydroenergy and water use using data mining methods[J]. Water Resources Management, 2020, 34(5): 1725-1741.
[22]
PUURBALANTAR. A Clipped Gaussian Geo-Classification model for poverty mapping[J]. Journal of Applied Statistics, 2021, 48(10): 1882-1895.
[23]
SEN A. Development as freedom[M]. New York: Anchor books, 2000.
[24]
GRAVESA, SCHMIDHUBERJ. Framewise phoneme classification with bidirectional LSTM and other neural network architectures[J]. Neural Networks, 2005, 18(5-6): 602-610.