Prediction program, prediction method, and prediction device

1 . A non-transitory computer-readable storage medium storing a prediction program that causes at least one computer to execute a process, the process comprising:

generating a first simulation group including a plurality of simulations of a flow of people, each of the simulations being based on a plurality of hypothetical behavior intentions of people;

acquiring observation data of an actual flow of people at a first time point, the observation data being in a form of a first density map;

for each of the simulations in the first simulation group, generating a second density map from the simulated flow of people at the first time point;

evaluating each of the simulations by calculating a likelihood of the flow of people based on an error between the first density map and the respective second density map; and

estimating, based on the likelihoods, a behavior intention that is a basis of a simulation evaluated as being close to the observation data;

constructing a second simulation group for a subsequent time point by selecting and updating the simulations based on the behavior intention; and

predicting a future flow of people at the subsequent time point by executing the second simulation group.

2 . A prediction method for a computer to execute a process comprising:

generating a first simulation group including a plurality of simulations of a flow of people, each of the simulations being based on a plurality of hypothetical behavior intentions of people;

acquiring observation data of an actual flow of people at a first time point, the observation data being in a form of a first density map;

for each of the simulations in the first simulation group, generating a second density map from the simulated flow of people at the first time point;

evaluating each of the simulations by calculating a likelihood of the flow of people based on an error between the first density map and the respective second density map;

estimating, based on the likelihoods, a behavior intention that is a basis of a simulation evaluated as being close to the observation data;

constructing a second simulation group for a subsequent time point by selecting and updating the simulations based on the behavior intention; and

predicting a future flow of people at the subsequent time point by executing the second simulation group.

3 . A prediction device comprising:

one or more memories; and

one or more processors coupled to the one or more memories and the one or more processors configured to:

generate a first simulation group including a plurality of simulations of a flow of people, each of the simulations being based on a plurality of hypothetical behavior intentions of people,

acquiring observation data of an actual flow of people at a first time point, the observation data being in a form of a first density map;

for each of the simulations in the first simulation group, generating a second density map from the simulated flow of people at the first time point;

evaluate each of the based on an error between the first density map and the respective second density map,

estimate, based on the likelihoods, a behavior intention that is a basis of a simulation evaluated as being close to the observation data,

construct a second simulation group for a subsequent time point by selecting and updating the simulations based on the behavior intention, and

predict a future flow of people at the subsequent time point by executing the second simulation group.