TRANSITION PROBABILITY LEARNING DEVICE, BEHAVIOR SERIES GENERATION DEVICE, AND PROGRAM

摘要

PROBLEM TO BE SOLVED: To achieve learning of a transition probability of a behavior for precisely calculating behavior series of a person.SOLUTION: A learning part 26 of behavior series probability for each time estimates a transition probability from the behavior of one of a pair to the behavior of the other for each of the behavior pair as a first transition probability for each time, and estimates the transition probability from the first behavior and the second behavior of a combination to the third behavior as a second transition probability for each of combinations of three behaviors aligned in time series for each time. A behavior series generation part 44 generates a behavior series for each of agents. The learning part 26 of behavior series probability for each time calculates a ratio of persons who perform behavior in a day for each behavior, compares data on the ratio of persons who perform behavior in a day and a ratio of persons who perform behavior in a calculated day, changes parameters and until the error reaches a predetermined range, and at the same time repeats estimation of a first transition probability, estimation of a second transition probability, and calculation of the ratio of persons who perform behavior in a day.