DATA PROCESSING APPARATUS, DATA PROCESSING METHOD, AND PROGRAM

To provide a data processing apparatus which effectively utilizes a data group with a defect.SOLUTION: A series of learning data including defects are acquired to calculate, from the series of learning data, a representative value of the data and an effectiveness rate indicating a rate of existence of the effective data with respect to a predetermined collection unit. An estimation model is learnt so that a difference between an output obtained by inputting the representative value and the effectiveness rate into the estimation model and the representative value can be minimized. A series of estimation data including defects are acquired to calculate, from the series of estimation data, a representative value of the data and an effectiveness rate indicating a rate of existence of the effective data with respect to a predetermined collection unit. A feature amount of the series of estimation data is acquired or data estimation is carried out by inputting the representative value and the effectiveness rate into the learnt estimation model.SELECTED DRAWING: Figure 1 【課題】欠損を含むデータ群を有効に活用するデータ処理装置を提供する。【解決手段】欠損を含む一連の学習用データを取得し、当該一連の学習用データから、あらかじめ定められた集計単位ごとに、データの代表値と、有効なデータが存在する割合を表す有効率とを算出し、前記代表値および前記有効率を推定モデルに入力して得られる出力と、前記代表値との差に基づく誤差を最小化するように、前記推定モデルを学習する。また、欠損を含む一連の推定用データを取得し、当該一連の推定用データから、あらかじめ定められた集計単位ごとに、データの代表値と、有効なデータが存在する割合を表す有効率とを算出し、前記代表値および前記有効率を学習済みの推定モデルに入力して、一連の推定用データについて特徴量の取得またはデータ推定を行う。【選択図】図１