近年，道路などのインフラの老朽化は深刻な社会問題となっており，補修工事が日々行われている．そのような工事を行うときにNTTの設備を傷をつけてしまうことを防ぐため，各工事にNTTの社員が立ち会うようにしている．そのため，各工事にどの社員を割り当てるかを定める必要がある．その際，難易度の高い工事が1人の社員に集中していないかなどを考慮している．また，各社員が担当する工事をすべて巡回したときの総移動時間を小さくしたいという要望や，スキルの低い人に難しい工事を割り当てると事故が起こりやすくなるため各工事を適切なスキルを持つ人に割り当てたいという要望など，様々な要望を考慮する必要がある．そのため，実社会の工事の割当ての意思決定には高度な技能が必要になる．本稿では，高度な技能を有している手配者と呼ばれる専門家により意思決定が行われている工事手配業務に対し，実用的な手配結果を算出可能な数理モデルを提示する．また，手配者がコメントした点を改善するように数理モデルを修正するに従い，算出される解と手配者が手作業で作成した手配結果との相違度が減少することを数値実験により示す．

In recent years, the aging of infrastructure such as roads has become a serious social problem, and maintenance works are being carried out daily. To avoid damaging NTT's facilities during such maintenance works, NTT employees attend them. For this, it is necessary to decide which employee is assigned to each maintenance work site considering various constraints and requirements. For example, each employee should not be assigned to too many sites with difficult maintenance works. It is preferable to reduce the total travel time of each employee to travel through all the maintenance work sites he or she is assigned to. It is also desirable to assign each maintenance work site to an employee with appropriate skills because otherwise accidents will become more likely to happen. With such requirements from the real world, it is not a simple task to find a good assignment, and the decision-making process is carried out by highly skilled professionals called arrangers. We present a mathematical programming model that obtains practically reasonable assignments. We also compare the assignments obtained by the several mathematical programming models that we have examined during the phase of improving the model through discussions with arrangers until the obtained assignments became acceptable for the arrangers. The results indicate that such assignments become closer to manually created ones as the model is improved.