Frontiers in Business, Economics and Management ISSN: 2766-824X | Vol. 8, No. 2, 2023 153 Identification and Model Application of Household Income Poverty and Asset Poverty in China Caixu Guo, Xuelian Huang, Li Li Shandong Technology and Business University, Yantai 264000, China Abstract: In order to measure the poverty status of Chinese households, the households were first identified under the income and asset criteria, then the incidence and gap rate of income poverty and asset poverty were analyzed by using the information of grouping variables. Finally, the prediction effect was analyzed by comparing the traditional model and the decision tree model, and the generated decision tree structure was used to reflect the internal poverty-causing variables of households. It provides some reference for displaying the poverty status of Chinese families and predicting the poverty status in the future. Keywords: Poverty, Decision tree, Poverty prediction. 1. Introduction The year 2018 was a decisive year in China's poverty alleviation campaign, with 97 percent of the population lifted out of poverty by the end of the year. China's anti-poverty cause has achieved remarkable results[1]. Under the current income standard, although the total number of poor people in China is decreasing year by year, some family poverty problems, such as poverty due to illness or disability, are difficult to cure[2]. Therefore, some scholars have proposed to use asset standards to dig out the root causes of poverty in deeply poor families, so as to formulate corresponding policies, actively and quickly help the poor out of poverty, and effectively prevent the families from falling into poverty again after poverty alleviation. Accurate identification of poor families is the first step in combating poverty. Identifying poor families is mainly divided into the following aspects. The first is to identify the criteria, taking into account the factors of "no worries, three guarantees" and the family's ability to generate income[3-6]. The second is to identify the scope, take the families in a certain region as the research object, adhere to the principle of openness, justice and transparency to ensure that the phenomenon of wrong poverty, poverty leakage will not happen again, and analyze the poverty structure and characteristics of the families in this region according to local conditions. Then there is the identification method. Most scholars identify income poverty by comparing the per capita net income of a family with the official poverty line[7-9]. However, for the identification of asset poverty, it is difficult to obtain asset variables and the country has not published the official poverty line for assets. Finally, poverty prediction. Due to the lag of the official data, it is impossible to predict the poverty in the future. Therefore, how to accurately predict family poverty by using the known information is the further research direction of scholars[10]. Correctly predicting and analyzing the development trend of all poor families is of great significance for improving the relevant social security system and promoting social justice and harmony. 2. Methodology 2.1. Poverty identification method There are slight differences in poverty measurement results under different standards. In order to achieve the purpose of complementation and comprehensive analysis, this paper uses the two standards for identification at the same time. The identification methods are as follows: Income poverty is identified by comparing the per capita net income of resident households with the official poverty line. The calculation formula is as follows: Where, denotes the per capita net income of a household, denotes the official poverty line, if denotes non-income poor households, if denotes income poor households. The identification of asset poverty is to regression the ratio of family welfare level to poverty line through the characteristics of the head of the household such as the years of education, family size, etc. The asset poverty value is obtained by multiplying the calculated regression coefficient with characteristic variables, namely the asset poverty index. The calculation formula is as follows: Where, represents family welfare level, represents family background variable, is regression model coefficient, is the estimate of regression model coefficient, is family asset index, and the number of background variable is n. If 1 represents non-asset poor households, if 1 represents asset poor households. 2.2. Poverty forecasting method In this paper, the decision tree method of machine learning model is used to predict and classify family poverty. The basic principle of this method is to use the tree graph to represent the expected value of each decision, calculate the value of the objective function and finally optimize the classification of the decision method with maximum benefit and minimum cost. Figure 1 is the flow chart of the decision prediction process. 154 Figure 1. Decision tree prediction decision flow chart 3. Results and Discussion 3.1. Chinese family poverty identification Based on the data of China Household Tracking Survey in 2018, the income poverty and asset poverty status of Chinese households were calculated respectively, and the poverty status of resident households was analyzed by the poverty incidence rate and poverty gap rate. The specific results are shown in Table 1. Table 1. Comparison of household poverty identification in China Poverty situation Incidence rate of poverty(%) Notch rate of poverty(%) Rural area city Rural area city Income poverty 1.653 0.851 31.471 25.944 Asset poverty 3.858 2.447 74.283 56.942 As can be seen from the results of Table 1, the poverty situation of rural families in China is more serious than that of urban families, and the poverty incidence is 0.802% higher. However, for both rural and urban households, the poverty under asset standard is wider and deeper than that under income standard. The gap of asset poverty of rural households reaches 74.283%, indicating that many families can hardly get out of asset poverty by their own ability. 3.2. Different groups of Chinese family poverty status According to the background variable information, the families were analyzed according to the characteristics of the head of the household, the level of education and the size of the family, so as to analyze the degree of poverty faced by different families and the causes of poverty. The specific results are shown in Table 2. Through analysis, it can be concluded that the better the conditions of the household head, the less the burden of the family, the less likely to fall into poverty. Families with six or more people have the highest incidence of income poverty (4.952 percent), while families with poor health have the highest income poverty gap rate (51.178 percent), and families with an uneducated head have the highest asset poverty gap rate (90.323 percent). 3.3. Chinese household asset poverty prediction using different methods By outputting the prediction results of traditional regression model and machine learning model, it can be found that the performance of decision tree model is better, which can obtain higher accuracy and AUC, and can accurately predict the asset poverty situation of households with different characteristics, providing certain theoretical basis for the formulation of anti-poverty policy suggestions. begin same? choose yes generate leaf node end extract feature A no A=? yes occupy the largest proportion all feature?no yes best feature no End of traversal? yes no 155 Table 2. Income poverty and asset poverty of Chinese households discussed by groups Background variable Income poverty Asset poverty Incidence rate(%) Notch rate(%) Incidence rate(%) Notch rate(%) Gender of household head male 1.459 26.454 2.956 55.385 female 1.950 30.628 3.517 59.651 The education level of the head of the household uneducated 4.216 50.326 5.239 90.323 Primary school education 4.017 49.251 6.322 74.912 Junior high school education 2.574 38.304 3.227 68.219 High school education 2.003 24.831 2.353 60.288 University degree or above 0.608 21.368 1.219 41.547 Whether the head of the household has a job no 3.515 30.527 5.234 71.227 yes 1.263 23.662 1.982 56.283 Family size 3 people and below 1.325 16.249 2.004 49.616 Four to six people 3.831 23.065 3.876 56.328 6 people and above 4.952 34.218 5.578 67.594 Physical condition of family members good 0.823 9.624 0.831 47.221 general 1.076 19.63 1.925 53.964 bad 1.968 51.178 2.784 77.528 Home area Eastern part 1.153 18.725 1.552 49.291 Central part 1.697 26.932 2.264 57.609 Western part 2.607 33.651 3.335 73.554 Table 3. Comparison of models based on asset poverty prediction results Model comparison Accuracy rate AUC error Traditional model 51.83% 0.529 165.228 Decision tree model 69.551 0.783 97.297 Since the decision tree model has a good performance in predicting asset poverty, the poverty-causing structure of poor families can be more clearly reflected by drawing the tree structure, and the drawing results are shown in Figure 2. True False Figure 2. Poverty structure of asset-poor households in China 156 As can be seen from Figure 2, the poor physical condition of the head of household is affected by gender. Among them, when the size of male-headed households is three or less, the families with a high school education or less are asset-poor households, while the families with a college education or above are non-asset-poor households. In contrast, female- headed households in the East were all non-asset-poor, while households in the Midwest were asset-poor. 4. Conclusion In order to analyze the overall and regional poverty situation of Chinese resident families, this paper also measures and analyzes Chinese poor families from different perspectives, and uses more accurate methods to predict family poverty. The main contents of this paper include: First, the asset poverty of Chinese households is identified according to the availability of information, the appropriate method is selected and the asset poverty line is obtained. Second, it analyzes the results of poverty status of Chinese households under both income and asset criteria, and compares and analyzes the differences brought by different perspectives in groups with different characteristics and their sizes. Thirdly, machine learning model is used to predict the asset poverty of Chinese households, and the advantages and disadvantages of traditional prediction models are compared to mine the useful information that can be obtained in machine learning model. Through research, it can be found that the poverty situation of rural families in China is more serious than that of urban families. For both rural and urban families, the poverty under the asset standard is wider and deeper than that under the income standard. The better off the head of a household is and the less the burden on the family, the less likely he is to fall into poverty. The decision tree model has better prediction performance than the traditional model and can be used to analyze the poor structure. Acknowledgment The paper is coming to an end. The difficulties and obstacles encountered in the process of writing the paper have been passed with the help of my classmates and teachers. Thanks to the tutor for the careful guidance of this paper, thanks to the students' positive encouragement, thanks to their efforts.General Project Supported By National Social Science Foundation:Research on Statistical Monitoring and Policy Regulation of Household wealth Inequality in China under the goal of Common Prosperity(22BTJ038).General Project of Social Sciences in Shandong Province: Research on multi- level identification and long-term governance mechanism of relative poverty(21CTJJ03). References [1] Rist Carl. Wealth and Health: Exploring Asset Poverty as a Key Measure of Financial Security.[J]. North Carolina medical journal,2022,83(1). [2] Shuey Kim M,Willson Andrea E. Trajectories of Work Disability and Economic Insecurity Approaching Retirement.[J]. The journals of gerontology. Series B, Psychological sciences and social sciences, 2019, 74(7). [3] Hitomi Komatsu,Hazel Jean L. Malapit,Sophie Theis. Does women’s time in domestic work and agriculture affect women’s and children’s dietary diversity? Evidence from Bangladesh, Nepal, Cambodia, Ghana, and Mozambique[J]. Food Policy,2018,79. [4] YUANYUAN YANG,JUN-HONG CHEN,MINCHAO JIN. Who are the Asset-Poor in China: A Comprehensive Description and Policy Implications[J]. Journal of Social Policy, 2019, 48(4). [5] Muzindutsi P.-F.. A comparative analysis of income- and asset- based poverty measures of households in a township in South Africa[J]. International Journal of Economics and Finance Studies,2018,10(1). [6] Swati Dutta,Lakshmi Kumar. Is Poverty Stochastic or Structural in Nature? Evidence from Rural India[J]. Social Indicators Research,2016,128(3). [7] CATHERINE ROWEEN C. ALMADEN. Asset-based Determinants of Poverty Intensity: A Meso-level Application in the Philippines[J]. WSEAS Transactions on Business and Economics,2015,12. [8] Lee Youngra,Lee Sook Jong. Factors Influencing the Asset and the Income Poverty of the Elderly - Focusing on ‘Baby Boomers’ and ‘Liberation and Korean War Generation’ -[J]. Journal of Social Science, 2018,57(2). [9] Okunola Solomon Olufemi. Gender Dimension of Asset Poverty in the Near-Urban and Rural Households in Selected Local Government Areas of Oyo and Osun states, Nigeria[J]. Journal of Economics and Sustainable Development, 2016, 7(13). [10] Yunhee Chang, Swarn Chatterjee, Jinhee Kim. Household Finance and Food Insecurity [J]. Journal of Family and Economic Issues, 2014,35(4).