Prediction models and associated factors on the fertility behaviors of the floating population in China

The floating population has been growing rapidly in China, and their fertility behaviors do affect urban management and development. Based on the data set of the China Migrants Dynamic Survey in 2016, the logistic regression model and multiple linear regression model were used to explore the related factors of fertility behaviors among the floating populace. The artificial neural network model, the naive Bayes model, and the logistic regression model were used for prediction. The findings showed that age, gender, ethnic, household registration, education level, occupation, duration of residence, scope of migration, housing, economic conditions, and health services all affected the reproductive behavior of the floating population. Among them, the improvement duration of post-migration residence and family economic conditions positively impacted their fertility behavior. Non-agricultural new industry workers with college degrees or above living in first-tier cities were less likely to have children and more likely to delay childbearing. Among the prediction models, both the artificial neural network model and logistic regression model had better prediction effects. Improving the employment and income of new industry workers, and introducing preferential housing policies might improve their probability of bearing children. The artificial neural network and logistic regression model could predict individual fertility behavior and provide a scientific basis for the urban population management.

Introduction

The reform and opening-up policy implemented in 1979 promoted China's economic development, and the shifting population driven by the booming economy also expanded swiftly. The floating population referred to new industry workers without local household registration. It was a concept under the household registration system in China (1). According to a report on the development of the floating population in China, the total number reached 121 million in 2000, accounting for 10 percent of the whole country population at the time, and increased to 245 million in 2016 (2). Most new industry workers moved from rural areas to urban areas or from central and western regions to eastern coastal areas for better job opportunities and living conditions (3).

According to the 2010 census in China, 53.6 percent of the floating population was born in 1980 or later, indicating a high proportion of new industry workers in their reproductive age (15–49 years) (4). The urban fertility rate has been below the replacement level since 1990 in China (5). However, the fertility rate of the floating population was lower than that of those living in both rural and urban areas (4). In the context of low fertility in China, decreasing birth rates would lead to labor constraints (6), economic slowdown (7), lack of innovation, and population aging (8). The floating population was an important labor force in urbanization construction (9). Promoting their fertility behaviors can well-alleviate their poor psychological and social health (10), poor sense of belonging (11, 12), and poor understanding of reproductive health (13), which are closely related to the stability and development of cities. An analysis of married women in China between 1980 and 1992 showed that residence, education level, and coincident marriage affected the first birth interval (14). A study on willingness of the floating population to have a second child in Hunan Province found that the relevant factors of fertility willingness included gender, age, occupation, education level, and marital status (15). Logistic regression, neural networks, and other machine learning models had been used to predict the birth results of pregnant women (16) and live birth results of embryos (17). However, there was still a lack of model research used to predict the fertility behavior of the floating population.

The one-child policy was enacted in 1979 to slow population growth at a time when productivity in China was relatively low, and its population was growing too fast (18). Violators of the policy, which was mainly enforced in cities and densely populated rural areas, could be fined and forced to undergo abortions or sterilizations (19). For nearly 40 years, late marriage, late childbirth, and strict population control became the main tone of fertility policy in the long term. However, with the economic development, the fertility level of China continued to be low, resulting in the imbalance of gender ratio, the weakening of the demographic dividend (20), and the acceleration of population aging (21), which made the transition of fertility policy urgent. On 29 October 2015, China implemented the universal two-child policy (20). However, the response of young couples to the “two-child policy” was not positive, and their willingness to have a second child was not high (22). By 2018, the birth rate of China dropped to its lowest level in 7 decades (23). China might be entering an era of negative population growth, with serious demographic and economic consequences (24). So China introduced the three-child policy in 2021 (25). Population trends were usually defined by fertility rates, which continued to increase after reaching replacement fertility rates (26). In the context of the low fertility rate of China, encouraging marriage and childbearing could increase the fertility rate. The proportion of newborns would gradually increase, while the proportion of the elderly would correspondingly decrease, alleviating the degree of population aging (27). The increase in the fertility rate could provide support for future labor stock, and the goal of sustainable economic development would be achieved (28, 29). The demographic dividend brought by the large proportion of the working-age population could be extended (30, 31).

The Chinese government has been encouraging couples to have more children to curb negative population growth and the aging population, but the implementation of the “two-child policy and three-child policy” requires the cooperation of families and individuals. Given the floating population, a group with a low fertility rate, this study explored the factors affecting the fertility behavior of the floating population, which could be helpful for relevant departments to formulate corresponding policies and measures to promote their fertility behavior, increase the future labor population of the city, and accelerate its construction and development.

According to the 14th Five-Year Plan of the Communist Party of China Central Committee, it is necessary to strengthen the construction of the digital society and digital government and improve the digital intelligence of public services and social governance. Urban population management includes the management of the floating population, family planning, and the quantity and quality of citizens. It could further strengthen the digital construction by relying on the original population information management system. By comparing the effects of three kinds of mathematical models applied to the individualized prediction of the fertility behavior of the floating population, this study selected scientific models to help relevant departments predict the potential population increment brought by new industry workers after they settled down in the local area, identify the individuals with low fertility possibility of the floating population, and take corresponding measures.

Data and methods

Data source and sample

The data used in this study were obtained from the China Migrants Dynamic Survey in 2016, which used a stratified three-stage random sample proportional to the population and collected information in the form of anonymous questionnaires (32). It was a large-scale national sample survey of the floating population conducted by the National Health Commission of China, covering 31 provinces (autonomous regions and municipalities directly under the central government) and the Xinjiang Production and Construction Corps, where the floating population was highly concentrated, with a sample size of nearly 200,000 households per year. The data covered basic information about the floating population and family members, the extent and length of migration, employment and social security, income and expenditure, residence, basic public health services, marriage, and family planning services and management. This data set was the secondary data collected from the questionnaire survey of the floating population. After removing the samples with unreasonable data and blank data, 168,993 valid questionnaires were obtained. The analyses were in an anonymized form and, consequently, would not be offensive to any individual or community.

Dependent variable

In the research on the correlative factors of fertility behavior, since the data did not meet the conditions of the ordered multi-category logistic regression, it was divided into two binary logistic regressions. In logistic regression of all samples, the question “Do you have children?” was the outcome variable. Then, the samples with children were screened out, and a logistic regression analysis was conducted with the question “Do you have two or more children?” as the outcome variable. The dependent variable was a categorical variable, where “yes” was marked as “1” and “no” as “0.”

A multiple linear regression model was applied to study the related factors of the age of first childbearing and birth spacing of the floating population. The time of first birth and the birth interval were used as the dependent variables, which were the continuous variables.

Independent variable

The basic information of the respondents, such as gender and age, were generally included in the model as control variables (33). Some studies suggested that the education level (34) and occupation (35) might affect the fertility behavior of residents. The object of observation in this study was the floating population, so the scope of migration and the duration of residence were also worth noting. Studies showed that new industry workers could change the original fertility pattern and move closer to the fertility behavior of residents in the destination (36, 37). The precondition for new industry workers to settle down was to acquire sufficient material basis, which was closely related to new industry workers' occupation, income (35), and housing (38). In addition, some economists believed that the introduction of social insurance might reduce the population's fertility rate (39). The involvement of healthcare services was required during the reproductive process (40).

Therefore, the independent variables in this study were divided into four aspects: personal information, migration situation, economic conditions, and social services. The study encoded the relevant variables (Table 1). Personal information included gender, ethnic group, registered permanent residence, education level, and occupation. The migration situation included the duration of residence after migration and the migration range of the investigation object. Economic conditions were measured by the level of the city the respondents lived in, their monthly income in the past year, and their real estate, which was measured by whether they bought a house locally. Social services referred to those obtained by the subjects themselves, including insurance services and health services. The former included whether to participate in endowment insurance, unemployment insurance, industrial injury insurance, maternity insurance, and medical insurance. The latter referred to the establishment of residents' health records and whether they have received health education related to occupational diseases, infectious diseases, and mental diseases.

TABLE 1

Table 1. Coding of categorical variables.

Methods

This study grasped the overall distribution characteristic of the floating population based on the related statistical descriptions. For univariate analysis, logistic regression and multiple linear regression models were used to analyze the influencing factors of fertility behaviors.

In the univariate analysis, the sample was grouped according to whether or not they had children. For the continuous variables with a non-normal distribution and the ordered categorical variable, the rank-sum test was used for comparison between groups. If the independent variable was an unordered categorical variable, the chi-square test was used for comparison between groups.

The aforementioned statistically significant associated factors were incorporated into the logistic regression model for multivariate analysis of fertility behavior. Logistic regression was often used to analyze the related factors of dichotomous outcomes (41, 42):

$\begin{array}{l}log\left(\frac{Pr(y=1)}{1-Pr(y=1)}\right)={\alpha }_0+b_1{x}_1+b_2{x}_2+\cdots +b_n{x}_n\\ (1)\end{array}$

where y = 1 means “yes” and y = 0 means “no.” x₁, x₂, ⋯, x_n represent the n independent variables in this study; b₁, b₂, ⋯, b_n are the coefficients of each variable; and e^b is equal to the odds ratio (OR). The estimated effect was expressed by OR with 95% confidence interval (CI).

When studying the associated factors of the age of the first birth and birth interval, the multiple linear regression model was established (43–45) with the associated factors as independent variables, and the age of the first birth or birth interval as dependent variables.

According to the aforementioned factors, the artificial neural network and naive Bayes models could be established. The first M = 90,000 samples were selected as the training set, and the remaining samples as the test set. The correlation coefficients of each model were trained to the optimal by using the training set.

The artificial neural network (ANN) (46) could be regarded as the simulation of the human brain nervous system. Dendrites were responsible for receiving input signals, and neurons were responsible for processing input signals. Then, they were transmitted to the next layer of neurons through synapses and continued to output after processing. The ANN model constructed in this study included input, two-layer activation function (hyperbolic tangent S-shaped function and linear function), and output (Figure 1).

FIGURE 1

Figure 1. Structure of artificial neural network model (ANN).

The naive Bayes model (47, 48) was based on the Bayes theorem to calculate the possibility of each outcome in the case of fixed features to select the outcome with the highest possible as the predicted value. The logistic regression model (49) could estimate the probability that samples with various attribute values belonging to a certain category. Logistic regression used the likelihood function as the training function, and the maximum likelihood estimate obtained was the predicted value of model coefficients (50).

For each model with the best parameters obtained by training, feature vectors of the test set were inputted to output its prediction results, and the accuracy rate (ACC), precision rate (PRE), and recall rate (REC) of the model were calculated to measure the prediction effects of models to select the optimal model.

where TP is the number of true-positive cases, TN is the number of true-negative cases, FP is the number of false-positive cases, and FN is the number of false-negative cases.

Statistical analysis methods

Continuous data with normal distribution were described by the mean and standard deviation. Continuous data with non-normal distribution were represented by the median and inter-quartile range (IQR). Classified data were described by using relative numbers. The rank-sum test of independent samples was used to process univariate analysis of continuous data with the non-normal distribution. Logistic regression and multiple linear regression models were used to analyze the related factors of fertility behaviors. The univariate analysis and multivariate analysis were processed by IBM SPSS Statistics 24. The artificial neural network and naive Bayes models could be conducted by MATLAB R2020a. The P ≤ 0.05 was considered statistically significant.

Results

Basic information

A total of 168,993 valid questionnaires were obtained in this study. The average age of the subjects was 39 years, with an inter-quartile range of 15 years. The local average monthly income in the previous year was 5500 CNY, with an inter-quartile range of 4000 CNY. The average insurance points was 2, with an inter-quartile range of 1. The average health service points was 6, with an inter-quartile range of 6. Descriptive statistics about the geographic location and demographic characteristics showed that 82.19% of subjects were residents from rural zones; 52.12% were male; 83.07% were married; 91.78% were Han; 61.71% had a junior high school education or less; 73.86% of the new industry workers had been away for <10 years; 46.95% were employees; 48.40% of the new industry workers crossed provinces or nations; and 72.32% of new industry workers rent houses in their city of residence (Table 2).

TABLE 2

Table 2. Basic information of the floating population.

Univariate analysis

The findings showed that the distribution of the variables mentioned earlier was different between the populace without children and with children, and the difference was statistically significant, including age (u = −215.36, P < 0.05), monthly income (u = −78.39, P < 0.05), insurance services (u = −4.91, P < 0.05), and health services (u = −7.03, P < 0.05). The duration of settlement (u = −88.48, P < 0.05), scope of migration (u = −13.48, P < 0.05), settlement (χ² = 33.47, P < 0.05), ethnic (χ² = 199.40, P < 0.05), household (χ² = 374.25, P < 0.05), marital status (χ² = 106641.67, P < 0.05), education (u = −133.30, P < 0.05), occupation (u = −49.16, P < 0.05), and housing (χ² = 1718.00, P < 0.05) statistically correlated with fertility behaviors of the floating population (Table 3).

TABLE 3

Table 3. Univariate analysis on fertility behavior of the floating population.

Multivariate analysis

The number of biological children born in the floating population was taken as the outcome variable in this model. However, this model did not pass the test of parallel lines. Therefore, two binary logistic models were chosen to analyze the related factors. In the model of one-birth behavior, the sample range was all the respondents, and the model was established with whether they had biological children as the outcome variable. The survey scope in the model of the second-child fertility behavior was all the survey subjects who had children, and the model was established with whether they were to have a second child as the dependent variable.

Factors related to fertility behavior of the floating population include age (χ² = 2578.01, P < 0.05), gender (χ² = 62.07, P < 0.05), ethnic group (χ² = 27.22, P < 0.05), household registration (χ² = 156.61, P < 0.05), marital status (χ² = 15581.80, P < 0.05), education level (χ² = 1908.52, P < 0.05), occupation (χ² = 308.26, P < 0.05), duration of residence (χ² = 1355.19, P < 0.05), scope of migration (χ² = 91.13, P < 0.05), settlement (χ² = 107.82, P < 0.05), and monthly household income (χ² = 109.53, P < 0.05) (Table 4). In particular, female new industry workers were more likely to have children than men (OR = 0.85, 95% CI: 0.81–0.88). The odds of new industry workers having children increased with age (OR = 1.08, 95% CI: 1.077–1.084). The odds of non-Han new industry workers bearing children were less than those of Han new industry workers (OR = 0.83, 95% CI: 0.77–0.89). The odds of the new industry workers with non-agriculture household registration had less active reproductive behavior than those with agriculture household registration (OR = 0.72, 95% CI: 0.68–0.76). In addition, new industry workers with better economic conditions (OR = 1.24, 95% CI: 1.19–1.29) were more likely to have children. Fertility behavior and education level show an inverted U-shaped distribution. Under junior high school, the higher the education level, the more the childbirth. However, for the new industrial workers whose education level is junior high school or above, the higher the education level, the fewer the childbirth. New industrial workers with junior high school education had 6.07 times as many children as those with postgraduate (OR = 6.07, 95% CI: 5.06–7.29).There is no statistical difference in the fertility behavior of employee and blue-collar workers (OR = 1.08, 95% CI: 0.93–1.25), but the fertility behavior of employers (OR = 1.29, 95% CI: 1.09–1.52), self-employed workers (OR = 1.77, 95% CI: 1.51–2.07), and the unemployed (OR = 1.41, 95% CI: 1.21–1.65) is higher than that of blue-collar workers. The number of children born to new industrial workers living in non-first-tier cities was 1.4 times that of those living in first-tier cities (OR = 1.40, 95% CI: 1.31–1.49) (Table 4).

TABLE 4

Table 4. Associated factors of the one-birth fertility behaviors of floating population.

Further analysis findings showed that the migrant population with non-agricultural household registration has about half the number of second children as the migrant population with agricultural household registration (OR = 0.51, 95% CI: 0.49–0.53). New industry workers with lower education levels were more motivated to have a second child. Age (OR = 1.04, 95% CI: 1.040–1.043) and household income (OR = 1.07, 95% CI: 1.05–1.09) were positively correlated with the likelihood of having a second child among the floating population. Meanwhile, the odds of the non-Han floating population giving birth to a second child was 1.42 times that of the Han floating population (OR = 1.42, 95% CI: 1.36–1.49). New industry workers living in non-first-tier cities were more likely to have a second child than those dwelling in first-tier cities (OR = 1.12, 95% CI: 1.08–1.17) (Table 5).

TABLE 5

Table 5. Associated factors of the two-children fertility behaviors of floating population.

Related independent variables were included in the multiple linear regression model, and it was found that there was no statistical relationship between monthly income and outcome variables. The factors that were positively correlated with the age of the first childbearing were insurance, health service, age, education, and housing property (Table 6). The age of first birth increased by 0.98 (95% CI: 0.95–1.00) years on average for each rank of education. The duration of settlement after migration (b = −0.03, P < 0.05) and the migration scope (b = −0.07, P < 0.05) were negatively correlated with the age of the first childbearing significantly. The first childbearing age of new industry workers living in first-tier cities was 0.338 years later than that of non-first-tier cities on average. The initial childbearing age of agricultural accounts was 0.62 (95% CI: 0.56–0.67) years earlier than that of non-agricultural accounts. Han new industry workers had one child 0.35 (95% CI: 0.28–0.43) years earlier than non-Han new industry workers on average. The age of the first birth of the female floating population is 1.49 (95% CI: 1.45–1.53) years earlier than that of the male floating population.

TABLE 6

Table 6. Associated factors of age at first childbearing and birth interval of floating population.

Insurance, health service, the duration of settlement after migration, age, and education were positively correlated with birth interval. In addition, the interval between multiple births of the floating population living in first-tier cities was 0.19 (95% CI:0.09–0.29) years shorter than that living in non-first-tier cities on average. The range of migration was a significant negative correlation factor, and the birth interval decreased by 0.05 (95% CI: 0.02–0.09) years for every one unit of migration scope increase (Table 6).

Prediction model

The statistically significant factors mentioned previously were incorporated into the prediction models of fertility behavior of the floating population. A total of 90,000 samples were retained as training data sets to fit the models, and the remaining samples were used as validation data sets to measure the prediction accuracy of the models. The results showed that the accuracy of the naive Bayes model was slightly inferior to that of the artificial neural network and logistic regression models. The artificial neural network and logistic regression models had better prediction effects, with an accuracy of 93.3% and a recall rate higher than 92.0% (Table 7). Therefore, it was more accurate to predict the fertility behavior of the floating population by using the artificial neural network model and the logistic models, which included the independent variables of personal status, the duration of settlement after migration, migration scope, economic conditions, and social services.

TABLE 7

Table 7. Comparison of the prediction effect on three models.

Discussion

As the total fertility rate of China had been declining, the family planning policy was changed into a two-child policy and, subsequently, three-child policy, which has become a current hot topic in society (51). In addition, the fertility rate of the floating population was lower than that of residents, so it was necessary to pay attention to the fertility situation of the floating population. The birth of the floating population was related to the urban construction and development. However, at present, there are few research studies on the factors affecting the fertility of the floating population, and the corresponding prediction models are also relatively lacking.

This study showed that personal status, the duration of settlement, scope of migration, economic conditions, and social services all influence the reproductive behavior of the floating population. For details, Han new industry workers were more likely to give birth to one child and less likely to give birth to two children than non-Han new industry workers. Migrant farmers were more active in childbearing and have children earlier on average. People with junior high school education were the most likely to have a child, showing a U-shaped pattern that first increased and then decreased. However, in terms of having a second child, the less educated new industry workers were more motivated to give birth. Higher educational attainment was associated with a later age at first birth and a larger spacing between births. Employers were much more likely to have children than blue-collar workers.

New industry workers who had settled for more than 10 years after emigration were more active in their reproductive behavior. The improvement of family economic conditions had a positive influence on the fertility behavior of new industry workers. The influence of monthly income on the second child was less than that of the first child. The new industry workers in first-tier cities were less likely to give birth to a kid and more likely to delay childbearing. New industry workers who owned property locally were far less likely to have a second child. Improvements in insurance and health services might be associated with later age at first birth and longer intervals between births.

A study of women's health in Texas found that an increase in clinics around the house would lead to an increase in fertility (52). At the first International Symposium on West African Studies, experts pointed out that improving the current situation of maternal and child health service supply in China could improve the fertility desire of the population of childbearing age (53). Combined with these studies, it could be concluded that the fertility desire of residents could be improved by bettering social medical services.

The health insurance reform has reduced the cost of pregnancy, which might increase the fertility rate of married women aged 20–34 years by about 1% (54). Insurance services in this study did not have a statistically significant effect on the fertility rate of new industry workers. This might be related to the unsatisfactory social security coverage of Chinese new industry workers (55) and the geographical limitations of some medical insurance (56, 57). Household income correlated closely with the number of children in metropolitan areas of the United States (58). People with better personal economic conditions expected more children. Also, in countries and regions with high economic status, the fertility rate of local women was relatively higher (59). Therefore, it supported the result that the increase in family income could promote reproductive behavior. People with higher education would delay marriage to some extent, resulting in a lower fertility rate (60). The human capital theory suggested that investment in education might produce marriage market returns (61). However, the higher demand for marriage partners among highly educated people, coupled with the huge cost of marriage caused by soaring property prices in China, might have reduced the desire of this group to get married, thus lowering the fertility rate. Consistent with this conclusion, people with higher education backgrounds were less likely to get married than those with a high school diploma, according to the Chinese Family Group Study (62).

A study on the ex-pat effect of a Maya Population from rural Guatemala found that new industry workers had their first babies earlier but had lower fertility rates, which could be attributed in part to stress (63), which explained the negative correlation between the migration range and the age of the first childbearing in this study, to some extent. After settling down for more than 10 years after migration, the immigrants' reproductive behavior was more active. This might be related to their wealth accumulation and improved quality of life.

First-tier cities and high housing prices might be important factors in decreasing fertility rates and delaying childbirth (64, 65). New industry workers who had their own houses in first-tier cities had spent longer time accumulating wealth in the past, thus delaying their childbearing. A study on Korean couples found that families living in non-metropolitan areas and renting houses had more active fertility behavior, which might be related to the family's housing requirements and the length of time spent to meet these demands (65). It was also confirmed by the results of our study. More preferential policies for renting or buying property might provide economic stability for new industry workers' initial settlement and meet their housing needs to promote the fertility rate. The difference between the rural floating population and non-agricultural fertility behavior might be related to the one-child policy of China announced in 1979. The policy was first strictly carried out in Shanghai and other big cities, while the implementation strategy was relaxed in the rural population with certain flexibility (66). Moreover, the concept of “raising children for old age” was deeply rooted in the rural population, and its fertility desire was stronger than that in the urban population.

In terms of employment, employment opportunities in first-tier cities were more attractive to the floating population (67), and it was more necessary to protect the basic rights and interests of new industry workers, such as income, and maintain their employment stability by building harmonious labor relations (68). In addition, it was necessary to improve the affordability of urban housing (69) and bring more new industry workers into the security scope of public rental housing and the community service system. Moreover, welfare policies such as housing subsidies could promote the settlement of the floating population (70). It was also suggested that their enthusiasm be increased to participate in insurance by expanding the coverage of work-related injury insurance (71), endowment insurance (72), and medical insurance (73, 74). Referring to medical and health services, integrating the floating population into the community health services, strengthening the maternal healthcare system, and adjusting the number of subsidies could improve the fertility rate of the floating population (75).

According to the associated factors obtained by the regression models, neural network, naive Bayes, and Logistic regression models were applied to predict the fertility behavior of the floating population. It was found that artificial neural networks and logistic regression could predict marriage and childbearing behavior of the floating population more effectively. This might be related to the assumption that the naive Bayes model needed to satisfy the independence of each feature vector (76).

Logistic regression used the logic function of a linear combination of numerical features to model the logarithmic probability of each category (77). Neural networks had low requirements for data. An artificial neural network consisted of an input layer, a hidden layer, and an output layer, with each linked to an earlier layer and each layer linked to another layer. In this study, we specified a hidden layer, a hyperbolic tangent, as the activation function and the identity activation function of the output layer and determined the model when the optimal difference of fitting conditions between the training set and test set was obtained (78). Among them, the performance of the ANN was superior to other networks in the field of medical prediction tasks (79). Accurate prediction of population fertility could reveal the trend of urban population growth, facilitate urban population management and construction, and benefit social stability and prosperity. Therefore, based on the information on the floating population's identity, the duration of settlement and migration scope, economic conditions, and social services, it was suggested that an artificial neural network and logistic regression be applied to predict fertility behavior, and the model coefficients be updated in time according to real-time data.

The study also had some limitations. The data set of Floating Population Dynamic Monitoring Survey of China in 2016 needed to be further supplemented by longitudinal follow-up data. In the analysis of related factors, regression analysis was used in this study, focusing on the dependence between variables. The causal relationship should be further explored to guide practical application.

In conclusion, the factors related to the reproductive behavior of the floating population were complex, such as social health services, family income, and urban living burden. We recommend the expansion of social health and insurance services, the promotion of employment and income levels of new industry workers, and the introduction of preferential policies for settling down. Furthermore, we should not blindly stimulate marriage and childbearing for the sake of urban population development. Due to the promotion of eugenics and the improvement of social construction such as insurance, people would no longer emphasize the number and speed of birth. Instead, they might pay more attention to the education and cultivation of the next generation. By incorporating the multi-factor analysis, the statistically significant correlation factors of personal status, the duration of settlement after migration and migration scope, economic conditions, and social services could be obtained. The artificial neural network model and logistic model with better performance might be used to make individual predictions. The prediction model of the population's childbearing behavior with high accuracy could help relevant departments to better predict and intervene in the development of the floating population, screen the population with low fertility possibility, and improve their fertility rate, ultimately to alleviate population aging and promote economic development.

Data availability statement

The data analyzed in this study is subject to the following licenses/restrictions: Application to data provider is required. Requests to access these datasets should be directed to Floating Population Service Center of China National Health Commission, China's Floating Population Dynamic Monitoring Survey Data Set (2016), http://hdl.handle.net/20.500.12291/10227.

Author contributions

XL and XZ designed the research study. XZ performed the research and wrote the manuscript. XL, XZ, ZZ, LG, LC, YZ, CH, JX, and JL offered help and advice on data collection and analysis. All authors have contributed to editorial changes in the manuscript, read, and approved the final manuscript version.

Acknowledgments

We acknowledge to the Floating Population Service Center of China National Health Commission for collecting and providing the data employed in this study. We thank all the peer reviewers and editors for their opinions and suggestions.

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

References

1. Wu XG, Treiman DJ. The household registration system and social stratification in China: 1955-1996. Demography. (2004) 41:363–84. doi: 10.1353/dem.2004.0010

PubMed Abstract | CrossRef Full Text | Google Scholar

2. Guan M. Associations between geodemographic factors and access to public health services among Chinese floating population. Front Public Health. (2020) 8:563180. doi: 10.3389/fpubh.2020.563180

PubMed Abstract | CrossRef Full Text | Google Scholar

3. Wang J-W, Cui Z-T, Ding N, Zhang C-G, Usagawa T, Berry HL, et al. A qualitative study of smoking behavior among the floating population in Shanghai, China. BMC Public Health. (2014) 14:1138. doi: 10.1186/1471-2458-14-1138

PubMed Abstract | CrossRef Full Text | Google Scholar

4. Guo W, Tan Y, Yin X, Sun Z. Impact of Pm2.5 on second birth intentions of china's floating population in a low fertility context. Int J Environ Res Public Health. (2019) 16:4293. doi: 10.3390/ijerph16214293

PubMed Abstract | CrossRef Full Text | Google Scholar

5. Morgan SP, Guo Z, Hayford SR. China's below-replacement fertility: recent trends and future prospects. Popul Dev Rev. (2009) 35:605–29. doi: 10.1111/j.1728-4457.2009.00298.x

PubMed Abstract | CrossRef Full Text | Google Scholar

6. Peng X. China's demographic history and future challenges. Science. (2011) 333:581–7. doi: 10.1126/science.1209396

PubMed Abstract | CrossRef Full Text | Google Scholar

7. Golley J, Tyers R, Zhou Y. Fertility and savings contractions in china: long-run global implications. World Economy. (2018) 41:3194–220. doi: 10.1111/twec.12602

CrossRef Full Text | Google Scholar

8. Skakkebaek NE, Jorgensen N, Andersson A-M, Juul A, Main KM, Jensen TK, et al. Populations, decreasing fertility, and reproductive health. Lancet. (2019) 393:1500–1. doi: 10.1016/S0140-6736(19)30690-7

PubMed Abstract | CrossRef Full Text | Google Scholar

9. You Z, Yang H, Fu M. Settlement intention characteristics and determinants in floating populations in Chinese border cities. Sustain Cities Soc. (2018) 39:476–86. doi: 10.1016/j.scs.2018.02.021

CrossRef Full Text | Google Scholar

10. Duan J-J, Wang D, Nie J. Analysis of self-rated health status of the floating population in a district of Guangzhou. Nan Fang Yi Ke Da Xue Xue Bao. (2008) 28:998–1000.

PubMed Abstract | Google Scholar

11. Yan QL, Yang Y, Gao YY, Hou RF, Zhang X, Dai MX, et al. Mental health status and its influencing factors in the floating population in Chengdu. Zhongguo yi xue ke xue yuan xue bao Acta Academiae Medicinae Sinicae. (2019) 41:729–36. doi: 10.3881/j.issn.1000-503X.11196

PubMed Abstract | CrossRef Full Text | Google Scholar

12. Wang X, Wu J, Li Y, Zhou Y, Li Y, Zhao R, et al. Changes in the prevalence of induced abortion in the floating population in major cities of China 2007-2014. Int J Environ Res Public Health. (2019) 16:3305. doi: 10.3390/ijerph16183305

PubMed Abstract | CrossRef Full Text | Google Scholar

13. Zhou Y, Wang T, Fu J, Chen M, Meng Y, Luo Y. Access to reproductive health services among the female floating population of childbearing age: a cross-sectional study in Changsha, China. BMC Health Serv Res. (2019) 19:540. doi: 10.1186/s12913-019-4334-4

PubMed Abstract | CrossRef Full Text | Google Scholar

14. Zheng ZZ. Social-demographic influence on first birth interval in China, 1980-1992. J Biosoc Sci. (2000) 32:315–27. doi: 10.1017/S0021932000003151

PubMed Abstract | CrossRef Full Text | Google Scholar

15. Wang Y, Ye W, Mao G, Xie H. Fertility desire and its influencing factors in hunan province under the policy of comprehensive two children. Chin J Reprod. Contracept. (2017) 37:738–42.

16. Sun YT, Zheng WW, Zhang L, Zhao HJ, Li X, Zhang C, et al. Quantifying the impacts of pre- and post-conception Tsh Levels on birth outcomes: an examination of different machine learning models. Front Endocrinol. (2021) 12:755364. doi: 10.3389/fendo.2021.755364

PubMed Abstract | CrossRef Full Text | Google Scholar

17. Huang B, Zheng SY, Ma BX, Yang YL, Zhang SP, Jin L. Using deep learning to predict the outcome of live birth from more than 10,000 embryo data. BMC Pregnancy and Childbirth. (2022) 22:36. doi: 10.1186/s12884-021-04373-5

PubMed Abstract | CrossRef Full Text | Google Scholar

18. Zhang JS. The evolution of china's one-child policy and its effects on family outcomes. J Econ Perspect. (2017) 31:141–60. doi: 10.1257/jep.31.1.141

CrossRef Full Text | Google Scholar

19. Hesketh T, Lu L, Xing ZW. The effect of china's one-child family policy after 25 years. N Engl J Med. (2005) 353:1171–6. doi: 10.1056/NEJMhpr051833

PubMed Abstract | CrossRef Full Text | Google Scholar

20. Jiang QB, Liu YX. Low fertility and concurrent birth control policy in China. Hist Fam. (2016) 21:551–77. doi: 10.1080/1081602X.2016.1213179

CrossRef Full Text | Google Scholar

21. Wang F, Zhao LQ, Zhao Z. China's family planning policies and their labor market consequences. J Popul Econ. (2017) 30:31–68. doi: 10.1007/s00148-016-0613-0

CrossRef Full Text | Google Scholar

22. Liu J, Liu M, Zhang SK, Ma QY, Wang QM. Intent to have a second child among chinese women of childbearing age following china's new universal two-child policy: a cross-sectional study. BMJ Sex Reprod Health. (2020) 46:59–66. doi: 10.1136/bmjsrh-2018-200197

PubMed Abstract | CrossRef Full Text | Google Scholar

23. Su-Russell C, Sanner C. Chinese childbearing decision-making in mainland China in the post-one-child-policy era. Fam Process. doi: 10.1111/famp.12772. [Epub ahead of print].

PubMed Abstract | CrossRef Full Text | Google Scholar

24. Feng W, Cai Y, Gu B. Population, policy, and politics: how will history judge china's one-child policy? Popul Dev Rev. (2013) 38:115–29. doi: 10.1111/j.1728-4457.2013.00555.x

CrossRef Full Text | Google Scholar

25. Zhang Y, Lin H, Jiang WB, Lu LX, Li YM, Lv BH, et al. Third birth intention of the childbearing-age population in mainland China and sociodemographic differences: a cross-sectional survey. BMC Public Health. (2021) 21:2280. doi: 10.1186/s12889-021-12338-8

PubMed Abstract | CrossRef Full Text | Google Scholar

26. Gendell M. The dynamics of population momentum. Popul Today. (1984) 12:6–7.

Google Scholar

27. Qi M, Dai M, Zheng Y. Discuss the impact and tendency of the universal two-child policy on China's birth rate fluctuation. China Popul Resour Environ. (2016) 26:1–10.

28. Alam MM, Murad MW, Molla RI, Rahman KM, Khondaker TR. Global population stabilisation policy and declining work-age population: a threat to global economic sustainability. Int J Environ Sustain Dev. (2019) 18:369–86. doi: 10.1504/IJESD.2019.103469

CrossRef Full Text | Google Scholar

29. Liu X, Wang S. Estimating china's labor supply in the sub age group during 2018-2025 under the background of universal two-child policy. Sci Technol Dev. (2018) 14:17–22.

30. Fang C. Population dividend and economic growth in China, 1978-2018. China Econ J. (2018) 11:243–58. doi: 10.1080/17538963.2018.1509529

CrossRef Full Text | Google Scholar

31. Liu Z, Fang Y, Ma L. A study on the impact of population age structure change on economic growth in China. Sustainability. (2022) 14:3711. doi: 10.3390/su14073711

CrossRef Full Text | Google Scholar

32. Commission FPSCoCNH. China migrants dynamic survey (2016).

Google Scholar

33. Liu P, Cao J, Nie W, Wang X, Tian Y, Ma C. The influence of internet usage frequency on women's fertility intentions-the mediating effects of gender role attitudes. Int J Environ Res Public Health. (2021) 18:4784. doi: 10.3390/ijerph18094784

PubMed Abstract | CrossRef Full Text | Google Scholar

34. Khanna T, Chandra M, Singh A, Mehra S. Why ethnicity and gender matters for fertility intention among married young people: a baseline evaluation from a gender transformative intervention in rural India. Reprod Health. (2018) 15:63. doi: 10.1186/s12978-018-0500-0

PubMed Abstract | CrossRef Full Text | Google Scholar

35. Ting TF. Resources, fertility, and parental investment in Mao's China. Popul Environ. (2004) 25:281–97. doi: 10.1023/B:POEN.0000036481.07682.7e

CrossRef Full Text | Google Scholar

36. Choi KH. Fertility in the context of Mexican migration to the United States: a case for incorporating the pre-migration fertility of immigrants. Demogr Res. (2014) 30:703–37. doi: 10.4054/DemRes.2014.30.24

CrossRef Full Text | Google Scholar

37. Nie W, Baizan P. Does Emancipation matter? fertility of Chinese international migrants to the United States and nonmigrants during china's one-child policy period. Int Migr Rev. (2021) 55:1029–60. doi: 10.1177/0197918321994789

CrossRef Full Text | Google Scholar

38. Vignoli D, Rinesi F, Mussino E. A home to plan the first child? fertility intentions and housing conditions in Italy. Popul Space Place. (2013) 19:60–71. doi: 10.1002/psp.1716

CrossRef Full Text | Google Scholar

39. Guinnane TW, Streb J. The introduction of bismarck's social security system and its effects on marriage and fertility in Prussia. Popul Dev Rev. (2021) 47:749–80. doi: 10.1111/padr.12426

CrossRef Full Text | Google Scholar

40. Kelly MM, Tobias J. Recommendations to optimize life-long health and wellbeing for people born preterm. Early Hum Dev. (2021) 162:105458. doi: 10.1016/j.earlhumdev.2021.105458

PubMed Abstract | CrossRef Full Text | Google Scholar

41. Austin PC. Absolute risk reductions, relative risks, relative risk reductions, and numbers needed to treat can be obtained from a logistic regression model. J Clin Epidemiol. (2010) 63:2–6. doi: 10.1016/j.jclinepi.2008.11.004

PubMed Abstract | CrossRef Full Text | Google Scholar

42. El Sanharawi M, Naudet F. Understanding logistic regression. J Fr Ophtalmol. (2013) 36:710–5. doi: 10.1016/j.jfo.2013.05.008

PubMed Abstract | CrossRef Full Text | Google Scholar

43. Wang H, Sui W, Xue W, Wu J, Chen J, Dai Y. Univariate and multiple linear regression analyses for 23 single nucleotide polymorphisms in 14 genes predisposing to chronic glomerular diseases and Iga nephropathy in han Chinese. Saudi J Kidney Dis Transpl. (2014) 25:992–7. doi: 10.4103/1319-2442.139882

PubMed Abstract | CrossRef Full Text | Google Scholar

44. Tsuji H, Tetsunaga T, Tetsunaga T, Misawa H, Nishida K, Ozaki T. Cognitive factors associated with locomotive syndrome in chronic pain patients: a retrospective study. J Orthop Sci. (2021) 26:896–901. doi: 10.1016/j.jos.2020.08.007

PubMed Abstract | CrossRef Full Text | Google Scholar

45. Baxter LA, Finch SJ, Lipfert FW, Yu Q. Comparing estimates of the effects of air pollution on human mortality obtained using different regression methodologies. Risk Anal. (1997) 17:273–8. doi: 10.1111/j.1539-6924.1997.tb00865.x

PubMed Abstract | CrossRef Full Text | Google Scholar

46. Alamsyah A, Permana MF. Artificial neural network for predicting Indonesian economic growth using macroeconomics indicators. Int J Adv Intell Informatics. (2018) 15–19. doi: 10.1109/SAIN.2018.8673347

CrossRef Full Text | Google Scholar

47. Kefei C, Cong Z. Naive Bayesian classifiers using feature weighting. Comput Simul. (2006) 23:92–94150.

PubMed Abstract | Google Scholar

48. Shree SRB, Sheshadri HS. Diagnosis of Alzheimer's disease using naive Bayesian classifier. Neural Comput Appl. (2018) 29:123–32. doi: 10.1007/s00521-016-2416-3

CrossRef Full Text | Google Scholar

49. Ksiazek W, Gandor M, Plawiak P. Comparison of various approaches to combine logistic regression with genetic algorithms in survival prediction of hepatocellular carcinoma. Comput Biol Med. (2021) 134:104431. doi: 10.1016/j.compbiomed.2021.104431

PubMed Abstract | CrossRef Full Text | Google Scholar

50. Gao SJ, Hui SL. Logistic regression models with missing covariate values for complex survey data. Stat Med. (1997) 16:2419–28.

PubMed Abstract | Google Scholar

51. Yang S, Jiang Q, Sanchez-Barricarte JJ. China's fertility change: an analysis with multiple measures. Popul Health Metr. (2022) 20:12. doi: 10.1186/s12963-022-00290-7

PubMed Abstract | CrossRef Full Text | Google Scholar

52. Lu Y, Slusky DJG. The impact of women's health clinic closures on fertility. Am J Health Econ. (2019) 5:334–59. doi: 10.1162/ajhe_a_00123

CrossRef Full Text | Google Scholar

53. Zhou M-D, Duan C, Wu H-X. An empirical study on the fertility and the supply of maternal and child health care service in China. In: 12th International Conference on Public Administration/1st International Symposium on West African Studies. Ghana (2017). p. 226–32.

54. Apostolova-Mihaylova M, Yelowitz A. Health insurance, fertility, and the wantedness of pregnancies: evidence from massachusetts. Contemp Econ Policy. (2018) 36:59–72. doi: 10.1111/coep.12235

CrossRef Full Text | Google Scholar

55. Liu Y, Shuai C, Zhou H. How to identify poor immigrants?–an empirical study of the three gorges reservoir in China. China Econ Rev. (2017) 44:311–26. doi: 10.1016/j.chieco.2017.05.004

CrossRef Full Text | Google Scholar

56. Peng B-l, Ling L. Association between rural-to-urban migrants' social medical insurance, social integration and their medical return in China: a nationally representative cross-sectional data analysis. BMC Public Health. (2019) 19:86. doi: 10.1186/s12889-019-6416-y

PubMed Abstract | CrossRef Full Text | Google Scholar

57. Shi X. Locked out? China's health insurance scheme and internal migration. Labour Econ. (2020) 67:101931. doi: 10.1016/j.labeco.2020.101931

CrossRef Full Text | Google Scholar

58. Freedman R, Coombs L. Economic considerations in family growth decisions. Popul Stud. (1966) 20:197–222. doi: 10.1080/00324728.1966.10406094

PubMed Abstract | CrossRef Full Text | Google Scholar

59. Davalos E, Fabio Morales L. Economic crisis promotes fertility decline in poor areas: evidence from Colombia. Demogr Res. (2017) 37:867–88. doi: 10.4054/DemRes.2017.37.27

CrossRef Full Text | Google Scholar

60. Nitsche N, Hayford SR. Preferences, partners, and parenthood: linking early fertility desires, marriage timing, and achieved fertility. Demography. (2020) 57:1975–2001. doi: 10.1007/s13524-020-00927-y

PubMed Abstract | CrossRef Full Text | Google Scholar

61. Courtioux P, Lignon V. A Good career or a good marriage: the returns of higher education in France. Econ Model. (2016) 57:221–37. doi: 10.1016/j.econmod.2016.04.011

CrossRef Full Text | Google Scholar

62. Hu M, Sun Z. Too educated to be married? an investigation into the relationship between education and marriage in China. J Family Stud. (2021) 1–19. doi: 10.1080/13229400.2021.1927152

CrossRef Full Text | Google Scholar

63. McKerracher L, Collard M, Altman R, Richards M, Nepomnaschy P. The ex-pat effect: presence of recent western immigrants is associated with changes in age at first birth and birth rate in a maya population from rural guatemala. Ann Hum Biol. (2017) 44:441–53. doi: 10.1080/03014460.2017.1343385

PubMed Abstract | CrossRef Full Text | Google Scholar

64. Hwang J. . Housing price and the level and timing of fertility in Korea: an empirical analysis of 16 cities and provinces. Health Social Welfare Rev. (2016) 36:118–42. doi: 10.15709/hswr.2016.36.1.118

CrossRef Full Text

65. Jeon S, Lee M, Kim S. Factors influencing fertility intentions of newlyweds in south korea: focus on demographics, socioeconomics, housing situation, residential satisfaction, and housing expectation. Sustainability. (2021) 13:1534. doi: 10.3390/su13031534

CrossRef Full Text | Google Scholar

66. Kane R, Choi CY. China's one child family policy. Br Med J. (1999) 319:992–4. doi: 10.1136/bmj.319.7215.992

PubMed Abstract | CrossRef Full Text | Google Scholar

67. Zheng Y, Zhang X, Dai Q, Zhang X. To float or not to float? internal migration of skilled laborers in China. Int J Environ Res Public Health. (2020) 17:9075. doi: 10.3390/ijerph17239075

PubMed Abstract | CrossRef Full Text | Google Scholar

68. Guan H. Empirical analysis of the impact of the employment stability on income of the floating populations in the pearl river delta region. In: 5th International Conference on Management, Computer and Education Informatization (MCEI) Shenyang (2015).

Google Scholar

69. Zhou M, Guo W. Fertility intentions of having a second child among the floating population in China: effects of socioeconomic factors and home ownership. Popul Space Place. (2020) 26:e2289. doi: 10.1002/psp.2289

CrossRef Full Text | Google Scholar

70. Wu Y, Luo J, Peng Y. An optimization-based framework for housing subsidy policy in china: theory and practice of housing vouchers. Land Use Policy. (2020) 94:104526. doi: 10.1016/j.landusepol.2020.104526

CrossRef Full Text | Google Scholar

71. Tang S, Long C, Wang R, Liu Q, Feng D, Feng Z. Improving the utilization of essential public health services by chinese elderly migrants: strategies and policy implication. J Global Health. (2020) 10:010807. doi: 10.7189/jogh.10.010807

PubMed Abstract | CrossRef Full Text | Google Scholar

72. Guo H, Luo Q. Analysis of China's population flow between urban and rural areas and the reform of public health old-age insurance system under the background of sustainable ecological environment. J Environ Public Health. (2022) 2022:1–8. doi: 10.1155/2022/9752913

PubMed Abstract | CrossRef Full Text | Google Scholar

73. Zhao Y, Kang B, Liu Y, Li Y, Shi G, Shen T, et al. Health insurance coverage and its impact on medical cost: observations from the floating population in China. PLoS ONE. (2014) 9:e111555. doi: 10.1371/journal.pone.0111555

PubMed Abstract | CrossRef Full Text | Google Scholar

74. Liu L. Does family migration affect access to public health insurance? Medical insurance participation in the context of Chinese family migration flows. Front Public Health. (2021) 9:724185. doi: 10.3389/fpubh.2021.724185

PubMed Abstract | CrossRef Full Text | Google Scholar

75. Cai Y-Y, Shi L-L, Jiang X-Q, Nortey GN, Mohammed IS, Ma J, et al. Dynamic analysis for health service dilemma of pregnant woman in floating population. In: 4th Conference on Systems Science, Management Science and System Dynamics. Shanghai: Donghua Univ (2011).

76. Fleuret F. Fast binary feature selection with conditional mutual information. J Mach Learn Res. (2004) 5:1531–55.

Google Scholar

77. Miyagi Y, Habara T, Hirata R, Hayashi N. Feasibility of artificial intelligence for predicting live birth without aneuploidy from a blastocyst image. Reprod Med Biol. (2019) 18:204–11. doi: 10.1002/rmb2.12267

PubMed Abstract | CrossRef Full Text | Google Scholar

78. Celik S, Eyduran E, Karadas K, Tariq MM. Comparison of predictive performance of data mining algorithms in predicting body weight in Mengali rams of Pakistan. Revista Brasileira De Zootecnia. (2017) 46:863–72. doi: 10.1590/s1806-92902017001100005

CrossRef Full Text | Google Scholar

79. Paydar K, Kalhori SRN, Akbarian M, Sheikhtaheri A. A clinical decision support system for prediction of pregnancy outcome in pregnant women with systemic lupus erythematosus. Int J Med Inform. (2017) 97:239–46. doi: 10.1016/j.ijmedinf.2016.10.018

PubMed Abstract | CrossRef Full Text | Google Scholar