PROBLEM TO BE SOLVED: To accurately estimate the reproductive ability of livestock by using a probabilistic model without using a judgment of a professional veterinarian or using an expensive measuring device. A livestock breeding ability estimation model generation device using a plurality of estimation indexes, and a learning measurement value acquisition unit that acquires a learning measurement value for each of a plurality of estimation indexes including a common index and a learning index. , An inter-index structure specifying unit that specifies an inter-index structure indicating a parent-child relationship between a plurality of estimated indices, a conditional probability table generation unit that generates a conditional probability table for each of a plurality of estimated indices, and a plurality of estimated indices Actual value for evaluation acquisition unit for acquiring the actual measured value for each evaluation, simple estimated value based on the actual measured value for evaluation for each common index, actual measured value for evaluation for each common index, and actual measured value for evaluation for each learning index Based on the simple estimated value and the complete estimated value, and a storage section for storing the conditional probability table as a livestock breeding ability estimation model. [Selection diagram] Figure 2【課題】専門的な獣医師の判断や高価な測定器を用いることなく、確率モデルを用いて家畜の繁殖能力を精度よく推定する。【解決手段】複数の推定指標を用いた家畜繁殖能力推定モデル生成装置であって、共通指標と学習指標とを含む複数の推定指標ごとの学習用実測値を取得する学習用実測値取得部と、複数の推定指標の間の親子関係を示す指標間構造を特定する指標間構造特定部と、複数の推定指標ごとの条件付確率表を生成する条件付確率表生成部と、複数の推定指標ごとの評価用実測値を取得する評価用実測値取得部と、共通指標ごとの評価用実測値に基づく簡易推定値と、共通指標ごとの評価用実測値と学習指標ごとの評価用実測値とに基づく完全推定値と、を算出する評価用推定値算出部と、簡易推定値と完全推定値とに基づいて、条件付確率表を家畜繁殖能力推定モデルとして記憶する記憶部と、を有する。【選択図】図２
