| 摘要 |
A method and a device for predicting inventory in an AD(ADvertisement) transfer system are provided to predict data for a next period by estimating a model from page view data of the previous period and using the model, and predict an objective result by estimating weight based on a pattern of the page view data according to each condition. A data extractor(110) extracts the page view data for the previous period. An estimator(130) estimates the model by using the extracted page view data. A predictor(140) predicts inventory data for the next period by using the estimated model. An abnormal data filter(120) filters abnormal data among the data extracted in the data extractor. The estimator includes a model estimator(131) estimating a prediction model by utilizing filtered weekly page view data, and a daily weight estimator(132) estimating daily weight of a week from daily page view data when the extracted data is the daily page view data. The predictor includes a weekly predictor(141) predicting weekly page view data from the estimated model and a daily weight predictor(142) predicting daily view data by using the estimated daily weight.
|
