Factors And Prediction Of The Residing Preference Of '90s' Migrant Residence

In this paper,based on the long-term residing preference of '90s' Migrant Residence in the inflow area,considering the complex nonlinear relationship of the influencing factors,a combined prediction method,Group Lasso-Adaboost model,is presented.In recent years,the '90s' Migrant Residence has shown a strong growth trend and is the main force for the continuous growth of the new generation of floating population.At the same time,the '90s' Migrant Residence has a greater change in the residence time of the inflow area.Their residing preference is need to study deeply.Therefore,this article takes the '90s' Migrant Residence as the research object.There may be many factors affecting a willingness to stay,such as intergenerational mobility,public services,medical insurance,health status,housing security,social integration,economic factors,etc.The main factor for correct selection is to improve model interpretability.And predictive premise.Considering that there are many factors affecting the willingness of migrants to stay in the national mobile population dynamic monitoring data used in this paper,and there are multiple collinearity problems among some influencing factors,this paper uses Group Lasso method to screen the long-term residence willingness of floating population after 90 s.The main factor.It can be seen from the parameter estimation coefficient of Group Lasso that its influence on residence will be affected by demographic,economic,social and psychological factors.The specific performance is: gender,education level,marital status,mobility and flow reasons,5 individual oral indicators,monthly housing expenditure,total family monthly expenditure,weekly working hours,employment unit nature,personal monthly income,income change and family month Income: 7 economic indicators,type of household registration,nature of housing,location of medical insurance,and whether to apply for temporary social residence,4 social indicators,as well as flow time,frequency of movement,number of years of non-going home,parental mobility,social type and community integration index.Then based on the selection of variables,the KNN model,Adaboost model and BP neural network model are constructed respectively.50% of the samples are selected to train the non-linear relationship between the long-term residing preference of '90s' Migrant Residence and the main influencing factors.Finally,the test set is predicted by the trained prediction model.By comparing the prediction results,the Adaboost modelhas the best prediction result,and the accuracy rate is 79.97%.This verifies the feasibility and effectiveness of the Group Lasso-Adaboost combination forecasting model in studying the behavior of migrants' residing preference.