To make it possible to highly precisely learn a prediction model which is used to predict the number of people who passes at a prediction time.SOLUTION: A learning unit 300 learns parameters of a prediction model on the basis of learning data, which contains a previous time at each of plural observation points and the number of people passing the observation point at the previous time, so that the number of people, who pass an observation point at a prediction time will coincide with the number of people passing at a time corresponding to a prediction time contained in the learning data. Herein, the number of people who pass the observation point at the prediction time is predicted using the prediction model having the parameters which include a first weight and a third weight. Here, the first weight represents a degree of influence exerted by the number of people, who pass another observation point, obtained based on a distance to the another observation point from each of plural observation points. The third weight is different from a second weight representing a degree of influence which is exerted by the number of people passing the another observation point along a moving route and represents a degree of influence exerted by the number of people passing the another observation point.SELECTED DRAWING: Figure 4
【課題】精度よく、予測時刻における通過人数の予測を行うための予測モデルを学習することができるようにする。【解決手段】学習部300が、複数の観測点の各々について、当該観測点の過去の時刻と、当該過去の時刻における当該観測点の通過人数とを含む学習データに基づいて、複数の観測点の各々について、他の観測点との距離に基づく当該他の観測点の通過人数からの影響度を示す第1の重み、及び移動ルートに基づく当該他の観測点の通過人数からの影響度を示す第2の重みとは異なる、当該他の観測点の通過人数からの影響度を示す第3の重みとを含むパラメータを有する予測モデルを用いて予測される、予測時刻の当該観測点の通過人数と、当該学習データに含まれる予測時刻に対応する時刻の通過人数とが一致するように、予測モデルのパラメータを学習する。【選択図】図4
TIME-SERIES LEARNING SYSTEM, TIME-SERIES LEARNING METHOD, TIME-SERIES PREDICTION SYSTEM, TIME-SERIES PREDICTION METHOD, AND PROGRAM
時系列学習装置、時系列学習方法、時系列予測装置、時系列予測方法、及びプログラム
2020-01-16
Patent
Electronic Resource
Japanese
European Patent Office | 2020
|European Patent Office | 2021
|Adaptive prediction of time series
NTRS | 1967
|Deep Learning-Based Flight Trajectory Prediction Using Time Series Decomposition
Springer Verlag | 2025
|The adaptive prediction of time series.
NTRS | 1966
|