| CPC G06Q 50/26 (2013.01) [G06Q 10/06375 (2013.01)] | 4 Claims |
|
1. A method for civil administration in a smart city based on Internet of Things (IoT), which is applied in a processing device of a civil administration platform, comprising:
obtaining target resident information of a target area from a civil service platform through a network, wherein the target resident information is obtained by the civil service platform after processing resident information of a target user in response to a civil service request sent by the target user through a terminal used by the target user of a user platform through the network; and
determining, based on the target resident information, an increased population of the target area within a target period, wherein the determining the increased population of the target area within the target period based on the target resident information comprises:
obtaining social development features of the target area;
determining increased population features of the target area within the target period by processing, based on a feature prediction model, the social development features; wherein the feature prediction model is machine learning model, including Long Short-Term Memory (LSTM), the feature prediction model is obtained by training an initial feature similarity determination model based on first training samples and first labels, the initial feature similarity determination model includes an initial feature prediction model, another initial feature prediction model, and an initial similarity determination model, the first training samples include a first historic social development feature and a second historic social development feature, the first labels include a historical increased population features similarity between a target area corresponding to the feature prediction model and a target area corresponding to the another initial feature prediction model, when the feature prediction model is LSTM,
inputting the social development features of the target area at social development features of a plurality of continuous times to the feature prediction model;
determining impact of the social development features of the plurality of continuous times on the increased population based on the changes of the social development features over time;
wherein the feature prediction model is trained based on training data which is determined based on historical data, the training data includes third training samples and third labels, the training of the feature prediction model includes:
inputting the third training samples to an initial feature prediction model and obtaining an output of the initial feature prediction model;
constructing a loss function based on the output of the initial feature prediction model and the third labels; and
updating parameters of the initial feature prediction model based on the loss function, until meeting preset conditions;
obtaining the feature prediction model; and
the training of the feature similarity determination model includes:
inputting the first historical social development feature to the initial feature prediction model and obtaining the output of the initial feature prediction model;
inputting the second historical social development feature to the another initial feature prediction model and obtaining the output of the another initial feature prediction model;
inputting the output of the initial feature prediction model and the output of the another initial feature prediction model to the initial similarity determination model and obtaining the output of the initial similarity determination model;
constructing a loss function based on the output of the initial similarity determination model and the first labels; and
updating parameters of the initial feature prediction model, the another initial feature prediction model, and the initial similarity determination model iteratively at the same time based on the loss function, until meeting preset conditions;
obtaining the feature prediction model, another feature prediction model, and a similarity determination model; and
determining the increased population by processing, based on a population prediction model, the target resident information and the increased population features; wherein the population prediction model is machine learning model, the population prediction model is obtained by training based on second training samples and second labels, the second training samples include historical target resident information within a plurality of sample periods and historical increased population features, and the second labels include historical increased population corresponding to each sample period;
the method further comprises:
obtaining a plurality of candidate solutions used for encouraging fertility when the increased population is smaller than a preset population threshold;
for each of the plurality of candidate solutions, determining a score of the candidate solution by processing, based on a scoring model, the candidate solution, the target resident information, and increased population features; wherein the scoring model is machine learning model, the scoring model is obtained by training based on third training samples and third labels, the third training samples include candidate solutions performed at different time points in the history, target resident information at that time, and historical increased population features within each historical target period, and the third labels are determined based on the increased population within a target time period after performing the candidate solutions, the training of the scoring model includes:
inputting the third training samples to an initial scoring model and obtaining an output of the initial scoring model;
constructing a loss function based on the output of the initial scoring model and the third labels;
updating parameters of the initial scoring model, until meeting preset conditions; and
obtaining the scoring model; and
determining, based on scores of the candidate solutions, a target solution of the target area within the target period.
|