JP4398672B2 - Allocation table creation system and allocation table creation program - Google Patents

Description

割付条件ごとの評価関数Ｅ_type（Ｘ）の定義は、割付条件の種類ごとに異なるので、以下に割付条件ごとの評価値の計算方法の例を示す。
【００５３】
まず、日ごとの割付条件については、勤務表中のある１日分のマスだけを（縦方向に）見て、その日について割付条件の違反があるかどうかが判断される。ここでは、その日ごとの違反度合いがペナルティ値となる。全ての日ごとの割付条件に対して、このペナルティ値を勤務表全体で日ごとに計算してその和をとれば、日ごとの割付条件のペナルティ合計が求められる。以下では、日ごとの割付条件である需要条件および組合せ条件について、ペナルティ合計の計算例を示す。
【００５４】
需要条件の例として、例えば「助産師に対して日勤を下限２名以上とする（重み：１００）」という条件があるとき、勤務表中のある日の助産師の日勤者が１名だったとすると、条件で指定された人数を１名下回っているのでペナルティ値は１×１００＝１００となる。同様の計算を全ての需要条件、全ての日に対して行い、ペナルティ値を合計したものが評価関数（需要条件用）の値となる。
【００５５】
なお、割付条件が「上限○名」の場合は、ペナルティ値としては（条件で指定した人数を上回った人数）×（重み）の値とし、それが「目標○名」の場合は（条件で指定した人数との差）×（重み）の値を用いればよい。
【００５６】
組合せ条件の例として、例えば「職員Ａは準夜勤務で職員Ｂ、Ｃ、Ｄとは組まない（重み：１００）」という条件があるとき、勤務表中のある日に職員Ａに準夜勤務が割り付けられていたとする。そのとき、同じ日に職員Ｂ、Ｃ、Ｄのうちで２名に準夜勤務が割り付けられていたとすると、当該条件に対して２名分違反していることになるので、ペナルティ値は２×１００＝２００となる。このような計算を全ての組合せ条件について行って、それにより得られる全ペナルティ値を合計したものが評価関数（組合せ条件用）の値となる。
【００５７】
次に、職員ごとの割付条件については、勤務表中のある職員１名分のマスだけを（横方向に）見て、その職員に対する割付条件の違反があるかどうかが判断される。ここではその職員ごとの違反度合いをペナルティ値とする。このペナルティ値を勤務表全体で職員ごとに計算し、その和をとれば、職員ごとの割付条件のペナルティ合計が求められる。以下では、職員ごとの割付条件である供給条件、パターン条件および間隔条件について、ペナルティ合計の計算例を示す。
【００５８】
供給条件の例として、例えば、「休みの回数は１ヶ月間で下限８回とする（重み：２０）」という条件があるとき、ある職員に対する１ヶ月間の休みの回数合計が６回だったとすると、条件で指定された回数を２回下回っているので、この供給条件に対するペナルティ値は２×２０＝４０となる。同様の計算を全ての供給条件、全ての職員に対して行い、ペナルティ値を合計したものが評価関数（供給条件用）の値となる。
【００５９】
なお、条件が「上限○回」の場合は、ペナルティ値としては（条件で指定した回数を下回った回数）×（重み）の値を用いればよく、当該条件が「目標○回」の場合は（条件で指定した回数との差）×（重み）の値を用いればよい。
【００６０】
パターン条件の例としては、例えば、「準夜−深夜の並びは回避する（重み：３００）」という条件があるとき、ある職員に対する１ヶ月間に「準夜−深夜」の並びが３回出現していたとすると、このパターン条件に対するペナルティ値は３×３００＝９００となる。同様の計算を全てのパターン条件、全ての職員に対して行い、ペナルティ値を合計したものが評価関数（パターン条件用）の値となる。なお、前の月から引き続く勤務表を作成する場合には、前月とまたがって出現する「準夜−深夜」の並びも１回として数えてよい。
【００６１】
間隔条件の例としては、例えば、「休みと休みの間隔を上限５日とする（重み：１０）」という条件があるとき、ある職員に対する１ヶ月間に休みと休みの間隔が５日を超える個所が２箇所出現していたとすると、この間隔条件に対するペナルティ値は２×１０＝２０となる。同様の計算を全ての間隔条件、全ての職員に対して行い、ペナルティ値を合計したものが評価関数（間隔条件用）の値となる。なお、前の月から引き続く勤務表を作成する場合には、前月とまたがって休みと休みの間隔が５日を超える場合も１箇所として数えてよい。
【００６２】
ちなみに、上記説明では、重みが数値である場合を例に説明したが、重みが「絶対」である場合には、ペナルティ値の計算において擬似的に重みとして十分大きな数（１００００など）を用いて計算することで、上記と同じ扱いとすることができる。また、職員ごとの条件におけるペナルティ値の計算方法としては、条件違反の度合いが大きいほど大きな値をとる方法であればどのような方法でもよく、上記の例に限定されるものではない。すなわち、前述の供給条件を例にとると、条件で指定された回数を２回下回っている場合のペナルティ値として、下回った回数の２乗と重みの積、すなわち２²×２０＝８０という値を用いてもよい。また、職員ごとの条件の種類については、上記の例以外にも多様な種類が考えられる。例えば、「３連続の休みを月１回取得する」というような条件は、職員毎の条件として表現しうる割付条件の一例である。この場合も、条件に違反した度合いが大きいほど値が大きくなるようなペナルティ値の計算方法を定義すれば、上記例と同様に取り扱うことができる。
【００６３】
なお、最初の段階では、通常、空欄が多くしかもその内容も適当でない場合が多いために、あまりよい全体評価値とはならない。しかし、勤務表全体に対する割付の自動的な繰り返し（本実施形態では部分割付のスキャンの繰り返し）により、勤務表の内容は段階的に良好になり、最終的には高品質の勤務表が得られる。
【００６４】
図７において、Ｓ１２０では、上記のように設定された勤務表作成期間の先頭日を日付Ｄとする。そして、Ｓ１２５では、その日付Ｄから始まるｄ日間の、全職員に対する全ての勤務予定の割り付け（部分割付）を実行し、記憶された勤務表の内容を部分的に更新する。この具体例が図８に示されており、これについては後に詳述する。
【００６５】
Ｓ１３０では、日付Ｄを翌日に移して、部分割付を繰り返す。ただし、Ｓ１３５において、日付Ｄ＋(ｄ−１)が作成期間の最終日を越えるならば、反復を終了し次のステップに移る。なお、上記では自動割付の対象となる期間の先頭の日から順番に割付を行う例を示したが、順番はこれに限定されるものではなく、その期間の後ろの日から前にさかのぼったり、あるいは適当な順番で対象日を選んでいってもよい。また、ここではｄ日間の連続した日を単位として割付を行っていく例を示したが、単日（ｄ＝１）もしくは勤務表作成期間の中の飛び飛びの複数日でも同様に取り扱うことができる。更に、一度に割付を行う日数ｄは勤務表全体を通じて一定である必要はなく、部分割付けを繰り返すごとに異なっても本質的に同様に取り扱うことができる。
【００６６】
Ｓ１４０では、以上のような反復処理が終了すると、図２に示した勤務表記憶部４８に保存されている勤務表に基づいて、評価値計算部４０により全体評価値を計算し、それを変数Ｅに格納する。続いてＳ１４５では、評価値計算部４０内の変数Ｅの値とＳ１１５で算出した変数Ｅ０の値とを比較する。変数Ｅの値が変数Ｅ０の値よりも小さいならば、直前のＳ１２５〜１３５の反復により、勤務表の内容が改善されており、さらなる反復で、より勤務表を改善できる可能性があるので、Ｓ１５０で、変数Ｅの値を新たに変数Ｅ０に格納しなおして、Ｓ１２０以降の各工程を繰り返す。一方、変数Ｅの値が変数Ｅ０と等しいあるいは大きいならば、それ以上勤務表の改善が見込めないものとして、Ｓ１６０で、その時に保存されている勤務表を表示部２８で表示し、自動割付処理を終了する。また、Ｓ１２５の直後に、その都度、表示部２８に表示させることも可能である。
【００６７】
なお、ここでは勤務表の最終日まで割付が終わった段階で全体評価値の計算を行って反復終了を判定しているが、終了判定の方法は、反復回数や経過時間の上限による方法、あるいはＳ１２５の直後に毎回Ｓ１４５等と同等の処理を行って全体評価値の変化を調べ、Ｓ１２５〜１３５の反復回数が一定回数に達するまでの間に、全体評価値の改善が見られなければ終了とすることもできる。
【００６８】
次に、図８を用いて、図７のＳ１２５に示した工程の具体的内容について詳述する。
【００６９】
Ｓ２００では、個別評価マトリクスを構成するために、ｄ日分について勤務パターンのリストを用意する。既に示したように、図３には、勤務が「休み」、「日勤」、「準夜」、「深夜」の４種類で、ｄ＝３の場合における各種の勤務パターンが示されている。なお、先に述べた通り、ｄ日は必ずしも連続している必要はない。すなわち、部分割付の対象として４月２日、４月４日、４月６日のように飛び飛びの日を選んだ場合、図３で示す勤務パターンにおける各勤務がそれぞれ４月２日、４月４日、４月６日に対応するものと考えれば、以降は同様に取り扱うことができる。
【００７０】
Ｓ２０５では、任意の１名の職員（職員ｓとする）を選択し、Ｓ２１０では、一連の勤務パターンの中から一つのパターンを選択する（これをパターンｉとする）。続いて、Ｓ２１５で、勤務表における職員ｓのセル列のうち、日付Ｄから始まるｄ日間分（図５の符号５６を参照）を仮にパターンｉで置き換えたとみなして、複数の第１割付条件について、ペナルティ値を計算する。そして、Ｓ２２５では、それらのペナルティ値を合計してその加算値を個別評価値とし（図５における符号６８参照）、その個別評価値を個別評価マトリクスにおける対応セルへ記入する。
【００７１】
個別評価マトリクスは、既に説明したように、（職員数）×（勤務パターンの数）のサイズをもったテーブルであり、その内で、上記の個別評価値は、（ｓ，ｉ）の位置に格納される（図４参照）。
【００７２】
Ｓ２３０では、すべての勤務パターンを仮割り付けしたかが判断され、未了の場合には、Ｓ２７０で勤務パターンが次に変更されて、上記Ｓ２１５からの各工程が繰り返し実行される。更に、Ｓ２３５では、すべての職員について上記の工程を実行したか否かが判断され、それが終了していなければ、Ｓ２６５で次の職員が選択されて、Ｓ２１０からの各工程が繰り返し実行される。このような過程を経ると、個別評価マトリクスにおける全セルに個別評価値が記入されることになる。
【００７３】
Ｓ２４０では、制約式生成部３４により、まず職員ｓに対して勤務パターンｉを割り当てるかどうかを定める０−１変数ｘ_si、および、以下で述べる補助変数を用いて以下のような制約式を生成する。
【００７４】
１）一人の職員ｓに対して、勤務パターンを１つしか割り当てないことを表す制約式は以下の通りである。各職員について同様の制約式を生成する。
【数２】

【００７５】
２）職員ごとの希望の勤務や休みなど、割付可能な勤務が限定されていることを表す制約式は以下の通りである。各職員について、日付Ｄから始まるｄ日間のうちで勤務が限定されている各日について同様の制約式を生成する。
【数３】

【００７６】
３）日付Ｄから始まるｄ日間に関する日ごとの条件（第２割付条件）のうち、重みが絶対のものに対応する制約式は、以下の通りである。
（例１）需要条件▲１▼「職員全員に対して日勤をちょうど１０名とする（重み：絶対）」に対する制約式
日付Ｄから始まるｄ日間の各日について、同様の制約式を生成する。
【数４】

（例２）需要条件▲２▼「「ベテラン」の職員に対して準夜勤務を下限１名とする（重み：絶対）」に対する制約式
日付Ｄから始まるｄ日間の各日について、同様の制約式を生成する。
【数５】

（例３）組合せ条件▲１▼「職員Eは日勤で職員F、G、H、Iのいずれかと組む（重み：絶対）」に対する制約式
ただし、Ｃは十分大きな数で、日付Ｄから始まるｄ日間の各日について、同様の制約式を生成する。
【数６】

【００７７】
４）日付Ｄから始まるｄ日間に関する日毎の条件（第２割付条件）のうち、重みが絶対でないものに対応する制約式は、以下の通りである。
（例１）需要条件▲３▼「助産師に対して日勤を下限２名とする（重み：１００）」に対する制約式
日付Ｄから始まるｄ日間の各日について、同様の制約式を生成する。
なお、以下の補助変数についても日毎に別のものを用いる。
【数７】

ただし、
【数８】

は補助変数で、以下の条件を満たす。
【数９】

なお、補助変数
【数１０】

は以下に説明する目的関数に対しても組み込まれる。
【００７８】
なお、上記の２）において、職員毎の希望休み等が指定されている場合の制約式を示したが、このような制約式を用いる代わりに、他の制約式や目的関数で使用する勤務パターンを制限してもよい。つまり、例えば職員Ａの日付Ｄに相当する日に「休」を希望として休みが設定されている場合、日付Ｄの勤務は必ず「休」となるため、勤務パターンのうち日付Ｄが「休」でないものについては、職員Ａに関する制約式を生成する際に勤務パターンの選択肢から除く、つまり対応する変数を制約式および目的関数に含めないものとしてもよい。
【００７９】
図８において、Ｓ２４５では目的関数が生成される。目的関数は以下のように表される。
【数１１】

ここで、右辺の各項について以下の通りである。
【００８０】
ａ）変数とペナルティマトリクス上の（s,i）の位置の値（ａ_siとする）の積に対する和（職員毎の条件に対するペナルティ値の総和を表す）
【数１２】

ｂ）重みが絶対でない日毎の条件のうち、上限を示す条件に関連する項
【数１３】

ｃ）重みが絶対でない日毎の条件のうち、目標を示す条件に関連する項
【数１４】

ｄ）重みが絶対でない日毎の条件のうち、下限を示す条件に関連する項
【数１５】

【００８１】
図８において、Ｓ２５０では、上記の各制約式を満たし、且つ、目的関数の値を最小化するように、全ての変数ｘ_siの０−１値および補助変数の値を決定する。ここで、決定される変数ｘ_siの０−１値は、後述のように、各職員ごとの最適な勤務パターンを特定するものである。
【００８２】
上記の制約式および目的関数は整数計画問題として定式化されているので、変数値の決定には整数計画法として既知の手法（Branch&Bound法など）を利用すればよい。これにより、絶対条件を満足し、且つ、それ以外の条件をできる限り満足できる理想的な勤務パターンのセットを生成することができる。
【００８３】
ただし、各種制約式の内容如何によっては、解を求められない（解なしとなる）場合がある。例えば日ごとの条件で、「ベテラン」資格を持つ職員が２名しかいない場合に２名とも同一日に希望の休みが設定されていると、需要条件「「ベテラン」の職員に対して準夜勤務を下限１名とする（重み：絶対）」は決して満足することができず、解なしとなってしまう。
【００８４】
従って、解を求める処理の後、Ｓ２６０で実際に解が得られているかどうかを判定し、解が得られなかった場合は、次に述べる勤務表の更新を行わないようにする。
【００８５】
なお、整数計画法においては、変数の個数が多いと計算時間が増加することが知られている。よって、高速化するには、変数の個数をなるべく減らすことが望ましい。そこで変数ｘ_siのうち、対応する個別評価マトリクス上の（s,i）の位置の値（ａ_siとする）がある閾値以上である場合には、変数ｘ_siを制約式および目的関数に含めないものとすることもできる。すなわちａ_siが大きい場合、対応する変数ｘ_siを１とすると目的関数値が大きくなることから、そもそもそのような変数ｘ_siが１となる可能性は低い。よって、あらかじめそのような変数を用いないこととし、高速化を図ることができる。
【００８６】
図８のＳ２６０では、上記において解が求まった場合、値が１となる変数が各職員ｓに対して１つずつ、かつ、全職員分同時に定まっている。よって、勤務表中の日付Ｄから始まるｄ日間について、個々の職員ｓごとに、値１に対応する勤務パターンを割り付ける操作を行う。これにより、部分期間に対して部分割付が完了する。
【００８７】
そして、図７で示したように、このような部分割付が順次実行されると、勤務表全体に対する割付（記入あるいは更新）が完了する。そして、勤務表全体の品質が一定条件を満たすまで、勤務表全体に対する割付が繰り返される。
【００８８】
上述したように、複雑かつ多様な割付条件を未分離状態で一括して取り扱うと、且つ、一定期間の全体を同時割付処理しようとすると、どうしても演算量が複雑かつ膨大となり、ひいては割付精度の低下を招きやすい。これに対し、上記の実施形態によれば、一度に割り付ける対象として部分範囲を設定し、それについて、時間軸方向の第１割付条件群について、各割付候補を暫定的に個別評価しておき（その評価は品質を考慮して割付表全体を対象）、次に、その集約された個別評価の結果及び人的構成に関する第２割付条件群を用いて候補の最適組合せを決定しようというものである。換言すれば、解の導出演算に当たって、範囲限定及び事前集約によって演算条件を削減し、迅速な演算を達成するものである。そして、必要に応じて、満足できる品質になるまで、部分割付の全体展開が繰り返し行われ、勤務表全体の品質が優良化される。よって、実用性及び有用性の高い勤務表作成（支援）システムを提供できるという利点がある。
【００８９】
なお、変形例としては、最初に第２割付条件群を利用して個別評価マトリクスを生成し、その個別評価マトリクスと第１割付条件群とを用いて部分割付内容を決定してもよい。また、上記のシステムは、勤務表以外の仕事割当表など他の割当表の作成にも応用することができる。
【００９０】
【発明の効果】
以上説明したように、本発明によれば、質のよい内容をもった割付表を迅速に作成できる。
【図面の簡単な説明】
【図１】 本発明に係るシステムによって作成される勤務表の一例を示す説明図である。
【図２】 本発明に係るシステムの好適な実施形態を示すブロック図である。
【図３】 パターン（候補）番号によって特定される勤務パターン（割付候補）の種類を示す図である。
【図４】 個別評価マトリクスを示す説明図である。
【図５】 第１割付条件群及び第２割付条件群に基づくペナルティの演算を説明するための図である。
【図６】 整数計画問題の解法に基づいて解（部分割付内容）を決定する流れを説明する図である。
【図７】 図２に示したシステムの動作を説明するためのフローチャートである。
【図８】 図７のＳ１２５の具体的内容を説明するためのフローチャートである。
【符号の説明】
２２ 演算処理部、２４ 勤務表作成プログラム（割付実行部）、３２ ペナルティ計算部、３４ 制約式生成部、３６ 目的関数生成部、３８ 解導出部、４０ 評価値算出部、４２ 終了条件判定部、４４ マスタ情報記憶部、４６ 割付条件記憶部、４８ 勤務表記憶部、５０ 個別評価マトリクス（個別評価結果記憶部）。[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an allocation table creation system, and more particularly to creation of an allocation table in which work schedules such as nurses and guards are allocated.
[0002]
[Prior art]
The conventional technology will be described taking a hospital as an example. In a hospital, a nurse chief generally creates a work schedule for one month of each nurse (a kind of allocation table) in each ward. There are nurses' daily work schedules such as “day shift”, “semi-night shift”, “night shift”, “rest”, etc. The entire work schedule must be prepared so as to satisfy many conditions, such as the number of nurses required for each work during the day. However, since there are countless combinations of work styles that can be taken in creating the work schedule, it is an exaggeration to say that it is extremely difficult or impossible to create a work schedule that sufficiently satisfies all the allocation conditions. is not. Therefore, while taking into consideration the magnitude of problems given to the work schedule when each allocation condition cannot be met, we are striving to create a work schedule so that a high-quality work schedule can be obtained through repeated trial and error. In this way, a great deal of labor and time is spent in preparing the work schedule by the nurse chief.
[0003]
In order to reduce the amount of work by creating the work schedule as described above, several methods for automatically creating the work schedule have been considered. As a method of automatically creating a work schedule, first search for all combinations of work styles that can be taken in the work schedule using a computer, check how much the various assignment conditions are satisfied for each work schedule, One way is to adopt a high-quality work schedule. However, there are innumerable combinations of work styles that can be taken in the work schedule, and it takes an enormous amount of time to evaluate all the combinations, so it is not possible to create a work schedule in practical time.
[0004]
Next, in order to create a work schedule at a practical time, a method of creating a work schedule by setting an upper limit on the number of iterations and processing time and randomly assigning a work schedule to the work schedule can be considered. However, because there are countless combinations of work schedules that can be taken in the work schedule as described above, it is not always possible to obtain a high-quality work schedule in a given iteration. The number of types of work styles and the number of nurses As the conditions for creating such work schedules increase, it becomes more difficult to create a high quality work schedule.
[0005]
Therefore, as a method for solving such a problem, various methods have been proposed in which the allocation process is devised, such as allocation while considering the conditions for creating the work schedule. For example, it is a method as disclosed in Non-Patent Document 1 below. This method pays attention to the fact that the conditions relating to “night shift” make it difficult to create a work schedule when creating a work schedule with two shifts. In this method, first, for each nurse, horizontal conditions related to “night shift”, such as the conditions for the number of “night shifts” for one month and the conditions for continuous work for “night shift” (which should be considered for each nurse). All work patterns that satisfy only (condition) are listed. Next, minimizing the degree of not satisfying the vertical condition related to “night shift” (conditions to be considered for each day) is given as an objective function, and this objective function is solved from the combinations of nurses' working patterns listed. It is. This determines the work of all nurses for “night shift”. After that, “day shift” and “vacation” are allocated to the unallocated portion of the work schedule. Since the conditions for allocating “day shift” and “vacation” in the two shift system are not strict, the allocation is completed by repeating the replacement of work that satisfies the conditions. As described above, the assignment for the entire work schedule is completed, and the work schedule is completed. According to such a method, the quality of the created work schedule can be guaranteed to some extent by allocating from the severe work, and the pattern of the created work schedule can be narrowed down to some extent, so that it is relatively short. Work schedule can be created in time.
[0006]
The following Patent Document 1 discloses a system for creating a work schedule, and particularly discloses a system having various judgment classifications.
[0007]
[Non-Patent Document 1]
Ikegami, Niwa, “An effective approach to nurse scheduling
Realization in a two-shift algorithm-"OR Japan Society of Papers,
Vol. 41, December 1998, P572
[Patent Document 1]
JP 2000-99569 A
[0008]
[Problems to be solved by the invention]
However, according to the method described in Non-Patent Document 1 above, at the stage of assigning “night shift”, it is not known whether or not the condition regarding the connection relationship with “day shift” or “vacation” can be satisfied. It can be pointed out that a high work schedule cannot always be created. In addition, a work schedule can be created in a relatively short time in the case of two shift work where the type of work is “day shift”, “night shift”, or “vacation”, but in the case of a shift work system of three shifts or more, “ Since there are other types of work such as “night shift”, the number of work combinations that the work schedule can take is very large, and it is not possible to create a work schedule within practical hours. Therefore, with the above method, it is not always possible to create a high-quality work schedule within a practical time. Moreover, generally speaking, when creating an allocation table such as a work schedule, it is desired to improve the allocation quality and enable quick calculation.
[0009]
An object of the present invention is to make it possible to quickly create an allocation table having high-quality contents.
[0010]
[Means for Solving the Problems]
(1) The present invention is a system for creating an allocation table by allocating work schedules to each unit of time within a certain period for a plurality of members constituting the organization, and comprising one or more time units within the certain period A partial range setting means for setting a partial range; and a division attaching means for executing a division with a plan to work for the set partial range, wherein the part attaching means includes a plurality of members for each member within the partial range. Temporary allocation means for tentatively assigning allocation candidates, individual evaluation means for performing individual evaluation on the entire allocation candidate for which the provisional allocation has been performed, with the entire fixed period being evaluated, and individual allocation candidates by the individual evaluation means Based on the evaluation result, an optimal combination of a plurality of allocation candidates for the plurality of members is determined and allocated to the partial range. It characterized by having a a optimal combination allocating unit.
[0011]
According to the above configuration, with respect to the partial range set artificially or automatically in the allocation table, a plurality of allocation candidates are assigned to each member on a trial basis, and an individual evaluation is performed on each temporarily allocated candidate. The In that case, in order to raise evaluation accuracy, the whole fixed period in an allocation table is made into evaluation object. And based on the individual evaluation result of each allocation candidate, the optimal combination of the some allocation candidate allocated about a some member is determined.
[0012]
Therefore, since the assignment target at one time is limited to a partial range (that is, with partial division), and even with such partial division, the entire assignment table is subject to evaluation, so it is quick and good. Can be assigned by quality. Of course, if necessary, the position of the partial range may be changed stepwise to expand the partial assignment to the entire assignment table.
[0013]
In the above, the allocation candidate corresponds to the type of work schedule or a series of types. Further, the optimum combination does not mean only the optimum combination in a strict sense, but means the combination that is determined to be optimum in terms of calculation (for example, by various known methods for obtaining an optimum solution). The member is typically an individual such as a staff member, a person in charge, or an employee, but may be a group (for example, a staff group). Moreover, although a time unit is usually a day, it is not necessarily limited thereto. The assigned work schedule may be a work form (day shift, night shift, etc.), or work content.
[0014]
Preferably, a storage unit in which a first allocation condition group relating to a series of work configurations of each member and a second allocation condition group relating to each hourly member configuration are registered, and the individual evaluation means includes the first evaluation unit. The individual assignment candidates are individually evaluated based on the assignment condition group, and the optimum combination assignment means determines the optimum combination based on the individual evaluation results of the assignment candidates and the second assignment condition group.
[0015]
According to the above configuration, the individual evaluation is performed on each allocation candidate based on the first allocation condition group while targeting the entire evaluation table. Then, the optimal combination is determined based on the individual evaluation result of each allocation candidate and the second allocation condition group.
[0016]
Therefore, by distinguishing between the allocation condition (first allocation condition) for the work composition in the time axis direction for each member and the allocation condition (second allocation condition) for the member structure for each unit time, As a whole, the allocation quality can be improved while reducing the amount of calculation in stages.
[0017]
Desirably, the plurality of first allocation conditions configuring the first allocation condition group and the plurality of second allocation conditions configuring the second allocation condition group, respectively, attribute information indicating the priority of the allocation condition And the individual evaluation means calculates an individual evaluation value for each allocation candidate for each member using the attribute information given to each first allocation condition, whereby each individual member for each member is calculated. An individual evaluation matrix composed of individual evaluation values of allocation candidates is constructed, and the optimum combination allocation means determines the optimal combination based on the contents of the individual evaluation matrix and attribute information given for each of the second allocation conditions To do.
[0018]
Each allocation condition has a different importance, priority, and influence due to violation, and the attribute information is used to reflect such differences in the allocation calculation. The attribute information is used for calculation of individual evaluation values, determination of an optimal combination, and the like.
[0019]
Preferably, the partial range setting means repeatedly sets the partial range while changing the setting position within the fixed period, and the partial division attaching unit executes the partial division attaching at each of the repeatedly designated setting positions. Thus, there is provided an end determination means for completing the entire allocation for the allocation table and repeatedly executing the entire allocation until a predetermined end determination condition is satisfied.
[0020]
According to the above configuration, the overall allocation is achieved by the stepwise progress of the division, and the overall allocation is repeated until a certain overall allocation quality is obtained. By repeating this, the quality of the allocation table is improved step by step. Therefore, it is desirable to appropriately set the end condition in consideration of the required quality and the required processing time.
[0021]
In particular, as the termination determination unit, a unit that calculates a total evaluation value for the entire allocation table, a unit that determines whether or not to repeatedly execute the total allocation based on the total evaluation value, It is desirable to provide. In that case, the overall evaluation value is no longer improved, the overall evaluation value has reached a certain value, the number of repetitions has reached a certain number, and a certain time has elapsed since the start of a series of allocation processes And so on.
[0022]
(2)PreferablySystem that creates an assignment table by assigning work schedules to each unit of time within a certain period for multiple members that make up the organizationBut,Means for setting a partial range consisting of one or a plurality of time units within the predetermined period, a first allocation condition group regarding a series of work configurations of each member, and a second allocation condition group regarding the member configuration of each time unit; Based on one of the first allocation condition group and the second allocation condition group, a preliminary evaluation is performed on each allocation candidate that can be allocated to the partial range, and an individual evaluation matrix is generated. Means for determining the optimum allocation content for the partial range based on the individual evaluation matrix and the other of the first allocation condition group and the second allocation condition group, thereby executing partial allocation; , IncludingMu
[0023]
According to the above configuration, a large number of allocation conditions are distinguished from each other in terms of the first allocation condition group and the second allocation condition group, and one of them is first used for preliminary evaluation for partial allocation, The other of them is used to determine the content with partial division. That is, each allocation condition group is used in stages. If you try to determine the optimal assignment contents using all the assignment conditions at once, the calculation contents will inevitably become complicated and the amount of calculation will be extremely large. Preliminary use of condition groups and the results of prior evaluations are aggregated into individual evaluation matrices, and the optimal assignment content can be derived from the results (individual evaluation matrix) and the other assignment condition group, so the derivation of solutions is efficient Can be done. In addition, the non-patent document 1 also discloses stepwise application of allocation conditions in a sense, but it gives priority to “night shift” allocation, and the classification of the first and second allocation condition groups as described above and It is different from handling.
[0024]
Preferably, the first allocation condition group is used first, and then the second allocation condition group is used. However, depending on the contents and conditions of the allocation table, the same principle can be applied even if the reverse is true. The following effects can be obtained. In the above configuration, when individual allocation candidates are individually evaluated, it is desirable that the entire allocation table be the evaluation target in consideration of the allocation quality. In the above configuration, each allocation candidate is a work schedule pattern of each member when the first allocation condition group is used in advance, and each second allocation condition group is each when the second allocation condition group is used in advance. It becomes a member composition of time unit (for example, day).
[0025]
(3) Further, the present invention is an allocation table creation program that is executed by being read by a computer and creates an allocation table by allocating work schedules for each time unit within a certain period for a plurality of members constituting an organization. The allocation table creation program includes a function of setting a partial range consisting of one or a plurality of time units within the predetermined period, and a function of executing a division of work scheduled for the set partial range, The function of executing the partial allocation is a function of temporarily assigning a plurality of allocation candidates to each member within the partial range, and an individual evaluation with the entire fixed period as an evaluation target for each allocation candidate that has been temporarily allocated And a plurality of assignments for the plurality of members based on the individual evaluation result of each assignment candidate by the individual evaluation means. Determining the optimum combination of candidates, and having a function of allocating it to the subranges, the.
[0026]
In the above, the allocation table creation program may be application software that is read and executed by an information processing apparatus such as a personal computer or a workstation.
[0027]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
[0028]
First, the allocation table created by the allocation table creation system according to the present invention will be described with reference to the work table shown in FIG.
[0029]
FIG. 1 shows an example of a work schedule displayed on the display unit of the computer. In this example, the work schedule is a schedule in which one month's work schedule is assigned to a plurality of nurses working in a specific ward of a hospital (a symbol indicating a work style is entered in each cell). In FIG. 1, the vertical axis direction (y direction described later) indicates a plurality of staff members (nurses) 10, and the horizontal axis direction (x direction described later) indicates each day in a certain period 12. Reference numeral 14 indicates a month for which allocation is completed or allocation is performed, and reference numeral 16 (shaded portion) indicates several days at the end of the previous month for reference. Since the work schedule is often viewed from the previous month, the work schedule for the previous month is partially displayed. Reference numerals 18 and 20 respectively indicate the totaling columns in the vertical axis direction and the horizontal axis direction (this is disclosed in the above-mentioned Patent Document 1). Note that the layout of the work schedule shown in FIG. 1 is an example, and other layouts may be adopted. For example, the date may be associated with the vertical axis direction and the staff may be associated with the horizontal axis direction.
[0030]
FIG. 2 is a functional block diagram showing the configuration of the system for creating the work schedule shown in FIG. This system is broadly divided into an arithmetic processing unit 22 configured by a CPU and a work schedule creation program 24, an input unit 26 and a display unit 28 connected to the arithmetic processing unit 22, and an arithmetic processing unit 22. Information storage unit 30. The input unit 26 includes a pointing device such as a keyboard and a mouse. Using this input unit 26, it is possible to input the desired holiday date of each staff member, input calendar information and master information described later, and set allocation conditions and weights described later. You can also. Furthermore, an assignment start instruction and an assignment end instruction can be input using the input unit 26. The display unit 28 is a display device such as a liquid crystal display. The display section 28 displays the work schedule shown in FIG. The information storage unit 30 may be an external storage device or may be constructed on the main memory. The work schedule creation program 24 functions as an allocation execution unit, and data generated along with the execution of the program is stored in the memory inside the arithmetic processing unit 22 or stored in the information storage unit 30. Note that necessary information can be exchanged by communicating with an external device via the network 52.
[0031]
Each typical function of the work schedule creation program 24 is shown as a block in FIG. The work schedule creation program 24 includes a penalty calculation unit 32, a constraint expression generation unit 34, an objective function generation unit 36, a solution derivation unit 38, an evaluation value calculation unit 40, an end condition determination unit 42, and the like. The information storage unit 30 includes a master information storage unit 44, an assignment condition storage unit 46, a work table storage unit 48. An individual evaluation matrix (individual evaluation value table) 50 is provided.
[0032]
Each configuration will be described below with reference to other drawings as appropriate.
[0033]
In the master information storage unit 44, calendar information, staff information, work schedules to be fixedly assigned, and other information, which are master information for creating a work schedule, are registered.
[0034]
The allocation condition storage unit 46 stores various allocation conditions necessary for creating a work schedule and “weights (weight values)” for each allocation condition in association with each other. The weight is an index indicating how much importance is given to the assignment condition associated therewith in the assignment process, and the weight is given as a label “absolute” or a numerical value (degree). The multiple allocation conditions are the allocation conditions (hereinafter referred to as the allocation conditions for each staff or the first allocation conditions) group regarding the work composition for each staff in the time axis direction, and the allocation conditions for the staff structure for each day (hereinafter, It is broadly divided into a group and a daily allocation condition or a second allocation condition. Specific examples thereof will be described later in detail, but the allocation conditions for each staff may be allocation conditions common to each staff or non-common allocation conditions. Similarly, the allocation condition for each day may be an allocation condition common to each day or each day of the week, or may be an allocation condition that is not common.
[0035]
In the work schedule storage unit 48, there is a matrix (work schedule) in which the date is taken on the horizontal axis and the staff is taken on the vertical axis, and information indicating the work schedule is entered in each cell of the matrix.
[0036]
As will be described later in detail, the individual evaluation matrix 50 is created prior to determining the optimal combination of allocation candidates for a certain period (part) in the work schedule. In this embodiment, the individual evaluation matrix 50 is configured as a table in which the type of work pattern is associated with one axis and the staff is associated with the other axis, and each allocation for each staff is included in each cell. Individual evaluation values for candidates are entered. FIG. 3 shows, as a list, all possible work patterns in the case where, for example, three days in one month are set as targets for division. For example, pattern (candidate) number # 1 is an arrangement of vacation-vacation-vacation work schedule, and pattern number # 2 is an arrangement of vacation-vacation-day work schedule. Each of these work patterns is positioned as a candidate for assignment for each staff member.
[0037]
FIG. 4 illustrates the contents of the individual evaluation matrix. The horizontal axis shows the work pattern shown in FIG. 3, and the vertical axis shows the staff. As will be described in detail later for each staff member, individual evaluation is performed on the basis of the entire work schedule, and the individual evaluation value 69 calculated thereby is entered in the corresponding cell. In this embodiment, the individual evaluation value is defined as the sum of a plurality of penalty values calculated based on a plurality of first allocation conditions. The individual evaluation of allocation candidates will be described in detail later.
[0038]
The penalty calculation unit 32 calculates a penalty value for each allocation condition on the premise of obtaining the overall evaluation value of the entire work schedule and the individual evaluation value. This will be described with reference to FIG. FIG. 5 schematically shows a work schedule 54. This work schedule 54 is in the middle of allocation (there may be some blanks). At the present time, the part indicated by reference numeral 56 is a target with partial division. The code | symbol 56 has shown the allocation condition (1st allocation condition) group for every staff. The code | symbol 58 has shown the allocation condition (2nd allocation condition) group for every day. Penalty calculation part 32 judges the degree of whether the condition is violated for every allocation condition for the whole work schedule. And when it violates, the weight matched with the condition is made into a penalty value. In FIG. 5, the penalty value for each allocation condition is abstractly represented by a symbol P surrounded by a circle symbol. For example, for the staff member A, a penalty value is calculated for each first allocation condition x1, x2, x3,. Similarly, a penalty value is similarly calculated for each first allocation condition for other staff members. A set of penalty values for each first allocation condition is indicated at 60. On the other hand, for each day, a penalty value is calculated for each second allocation condition y1, y2, y3,. A set of penalty values is indicated by reference numeral 62.
[0039]
In the present embodiment, an overall evaluation value 64 serving as an index for evaluating the entire work table is calculated by adding the set of penalty values indicated by the reference numerals 60 and 62 as a whole. The evaluation value calculation unit 40 shown in FIG. The overall evaluation value 64 indicates that the smaller the value, the better the evaluation. In the embodiment, the overall evaluation value 64 is used as a criterion for end determination when a series of division with the entire work schedule is repeatedly performed. Here, the end determination is made by the end condition determination unit 42 shown in FIG. Moreover, in this embodiment, the individual evaluation value 68 of the said allocation candidate is calculated | required by adding the penalty value for each 1st allocation condition about the said staff when the allocation candidate is temporarily allocated about each staff. Yes. The calculation is also performed by the evaluation value calculation unit 40 in this embodiment. The individual evaluation value 68 is entered in the individual evaluation matrix 50 as described above, and is used in selecting an optimal combination of allocation candidates.
[0040]
FIG. 6 conceptually shows a method of performing division with respect to d days as a partial period (the specific contents will be described later with reference to FIGS. 7 and 8). In the present embodiment, a known integer programming problem solution is adopted in order to determine an optimal combination of allocation candidates (a set of work patterns) to be allocated to a partial period. In addition to this, a solution such as a genetic algorithm can be used.
[0041]
In FIG. 6, a plurality of first allocation conditions 70 and a plurality of second allocation conditions 72 are registered in the allocation condition storage unit 46 as already described. As described above, the individual evaluation matrix 50 is generated using the plurality of first allocation conditions 70. On the other hand, from the plurality of second allocation conditions 72, a plurality of constraint expressions 74 necessary for executing the above integer programming problem solution are generated. In addition, one or a plurality of constraint expressions 74 are generated from other conditions (for example, fixed values). The objective function 78 includes a second allocation condition (a second allocation condition in which a numerical value as a non-absolute weight is given) other than those assigned “absolute” as a weight among the plurality of second allocation conditions 72 Are generated using the individual evaluation matrix 50. In this case, there are constraint equations having auxiliary variables (described later) in the plurality of constraint equations 74, but these auxiliary variables are also incorporated in the objective function 78 (in FIG. To show). As indicated by reference numeral 80, an integer programming problem determined by the constraint equation 74 and the objective function 78 is assumed to satisfy all the generated constraint equations 74 and minimize the value of the objective function 78. Is determined. The solution identifies the optimal combination of each employee's work pattern in the partial period.
[0042]
In FIG. 2, the constraint expression generation unit 34 generates the plurality of constraint expressions 74 shown in FIG. The objective function generator 36 generates the objective function 78 shown in FIG. Further, the solution derivation unit 38 determines an optimal combination using a plurality of constraint equations 74 and an objective function 78. In addition, although penalty calculation part 32 and evaluation value calculation part 40 were shown as another structure in FIG. 2, you may integrate these functions. The same applies to other configurations.
[0043]
In the following, each item described above will be described in detail using a specific example.
[0044]
First, a specific example of allocation conditions will be described. As mentioned above, the allocation conditions are as follows: allocation conditions for the work composition for each staff (assignment conditions for each staff, first allocation conditions), and allocation conditions for the staff composition for each day (daily allocation conditions, second allocation) And conditions).
[0045]
The daily allocation condition (second allocation condition) is a vertical allocation condition in the work schedule, that is, an allocation condition in which it can be determined whether the condition is satisfied every day in the work schedule. For example, there are demand conditions and combination conditions as follows.
(Example of demand conditions)
Conditions that determine the upper limit, lower limit, and target number of staff per day for each staff qualification and duty
・ The day shift for all staff members is just 10 (weight: absolute)
・ Minimum two days for midwives (weight: 100)
・ Semi-night duty is set to one lower limit for “Veteran” staff (weight: absolute)
In the above example, the day to which the allocation condition is applied is not limited. However, the condition to be applied only on a specific day or a specific day of the week can be similarly defined.
(Example of combination conditions)
Conditions that define the staff members who want or do not want to work together
-Staff A works semi-night and does not pair with staff B, C, D (weight: 100)
・ Staff E is a day shift and works with staff F, G, H, or I (weight: absolute)
(In the case of a new employee E who is in charge of the guidance staff)
[0046]
Next, the allocation condition for each staff member (first allocation condition) is the horizontal allocation condition in the work schedule, that is, the allocation condition that can determine whether or not the condition is satisfied for each staff member in the work schedule. is there. For example, there are a supply condition, a pattern condition, and an interval condition as follows.
(Example of supply conditions)
Conditions for the upper and lower limits of the number of working hours of employees within a predetermined period (for example, one month) and the target number of times
・ The minimum number of holidays is 8 times per month (weight: 20)
・ The maximum number of night shifts for staff B is six per month (weight: 100)
(Example of pattern conditions)
Conditions that define the work you want to avoid when you focus on one employee
・ Avoid the arrangement of midnight and midnight (weight: 300)
・ Non-breaking nights (weight: 50)
(Example of interval condition)
Conditions that set the upper and lower limits and goals of the interval between work when paying attention to one employee
・ Maximum interval between holidays is 5 days (weight: 10)
[0047]
7 and 8 are flowcharts showing an example of the operation of the system shown in FIG. FIG. 7 is a diagram showing the overall operation, and FIG. 8 is a diagram showing a specific processing example of S125 in FIG. It is assumed that master information such as calendar information, staff information, and type of work to be assigned is input in advance using the input unit 26 shown in FIG. Further, it is assumed that each allocation condition and its weight necessary for creating the work schedule are also input by the input unit 26 and stored in the allocation condition storage unit 46.
[0048]
In S <b> 105, the user inputs a period to be automatically assigned to the work schedule by using the input unit 26. For example, when a work schedule for April is to be created, the period from April 1 to April 30 is input. When the work schedule for April 15 is already created and the work schedule for April 15 and thereafter is recreated, the period from April 15 to April 30 is input. The period input here is stored in the information storage unit 30. In the following, a case where the entire work schedule is a period for automatic allocation will be described as an example.
[0049]
In S110, the user inputs a desired holiday and a desired work style for each staff member within the creation period using the input unit 26. It is stored as a fixed value in the corresponding cell in the work schedule storage unit 48 of FIG. This is to respect the person's wishes and applications. However, the desired holidays and desired working modes for each staff member may be handled with weights in the same manner as other allocation conditions described below.
[0050]
In S115, when an instruction to start allocation is input using the input unit 26, the evaluation value calculation unit 40 calculates the overall evaluation value for the entire work table and records it in the variable E0 in the evaluation value calculation unit 40. . Specifically, as shown in FIG. 5, the penalty calculation unit 32 is based on the contents of the current work schedule (there may be a blank) for the allocation conditions for each staff member and the allocation conditions for each day. The violation degree is specified as a penalty value. The evaluation value calculation unit 40 in FIG. 2 sums these penalty values to obtain an overall evaluation value for the entire work schedule.
[0051]
Here, the overall evaluation value will be specifically described. The overall evaluation value is a numerical value determined for the work schedule X in a state where some work schedule is assigned throughout the work schedule, or in a state where some blanks remain in the work schedule, The smaller the value, the better the content. Assuming that an evaluation function for obtaining an overall evaluation value for the work schedule X is E (X), E (X) is an evaluation function E for each allocation condition for the work schedule X as shown below._typeIt is defined as the sum of (X).
[0052]
[Expression 1]

Evaluation function E for each allocation condition_typeSince the definition of (X) differs for each type of allocation condition, an example of a method for calculating an evaluation value for each allocation condition is shown below.
[0053]
First, regarding the allocation conditions for each day, it is determined whether or not there is a violation of the allocation conditions for that day by looking only at a square for one day in the work schedule (in the vertical direction). Here, the degree of violation for each day is a penalty value. If the penalty value is calculated for each day in the entire work schedule and the sum is calculated for all the daily allocation conditions, the total penalty for the daily allocation conditions is obtained. Below, the example of calculation of a penalty total is shown about the demand conditions and combination conditions which are the allocation conditions for every day.
[0054]
As an example of the demand condition, for example, when there is a condition that “the minimum day shift is 2 or more for midwives (weight: 100)”, there is one midwife on one day in the work schedule. Then, since the number of persons specified in the condition is one person less, the penalty value is 1 × 100 = 100. The same calculation is performed for all demand conditions and all days, and the sum of the penalty values is the value of the evaluation function (for demand conditions).
[0055]
If the allocation condition is “upper limit ○ name”, the penalty value is (number of people exceeding the number of people specified in the condition) × (weight) value, and if it is “target ○ name” A value of (difference from specified number of people) × (weight) may be used.
[0056]
As an example of the combination condition, for example, when there is a condition that “employee A works semi-night and does not associate with staff B, C, D (weight: 100)”, employee A works semi-night on a certain day in the work schedule. Is assigned. At that time, if two employees of staff B, C, and D are assigned to work on a semi-night basis on the same day, the penalty value is 2 × because two people are violating the condition. 100 = 200. Such calculation is performed for all combination conditions, and the total of all penalty values obtained thereby is the value of the evaluation function (for combination conditions).
[0057]
Next, regarding the allocation conditions for each staff member, only the square for one staff member in the work schedule is viewed (in the horizontal direction) to determine whether there is a violation of the allocation conditions for that staff member. Here, the degree of violation for each employee is taken as a penalty value. If this penalty value is calculated for each staff member in the entire work schedule and the sum is taken, the total penalty of the allocation conditions for each staff member is obtained. The following shows an example of calculating the penalty total for the supply conditions, pattern conditions, and interval conditions that are the assignment conditions for each staff member.
[0058]
As an example of supply conditions, for example, when there is a condition that “the number of holidays is a lower limit of 8 times per month (weight: 20)”, the total number of holidays per month for a certain employee is 6 times. Then, since the number of times specified in the condition is twice, the penalty value for this supply condition is 2 × 20 = 40. The same calculation is performed for all supply conditions and all staff, and the sum of the penalty values is the value of the evaluation function (for supply conditions).
[0059]
When the condition is “upper limit ○ times”, the penalty value may be (number of times less than the number of times specified in the condition) × (weight). If the condition is “target ○ times” A value of (difference from the number of times specified by the condition) × (weight) may be used.
[0060]
As an example of the pattern condition, for example, when there is a condition of “avoid quasi-night-midnight arrangement (weight: 300)”, the “quasi-night-midnight” arrangement appears three times in a month for a certain employee. If so, the penalty value for this pattern condition is 3 × 300 = 900. The same calculation is performed for all pattern conditions and all staff, and the sum of the penalty values is the value of the evaluation function (for pattern conditions). In addition, when creating a work schedule that continues from the previous month, the “quasi-night-midnight” sequence that appears across the previous month may be counted as one time.
[0061]
As an example of the interval condition, for example, when there is a condition that “the interval between the holidays is 5 days (upper limit: 10)”, the interval between the holidays for a certain employee exceeds 5 days in one month. If two locations appear, the penalty value for this interval condition is 2 × 10 = 20. The same calculation is performed for all interval conditions and all staff, and the sum of the penalty values is the value of the evaluation function (for the interval condition). In addition, when creating a work schedule that continues from the previous month, it may be counted as one place even when the interval between holidays and holidays exceeds 5 days.
[0062]
By the way, in the above description, the case where the weight is a numerical value has been described as an example. However, when the weight is “absolute”, a sufficiently large number (such as 10,000) is used as the weight in the calculation of the penalty value. By calculating, the same treatment as above can be performed. Moreover, as a calculation method of the penalty value in the conditions for each staff member, any method may be used as long as the degree of condition violation is larger, and the method is not limited to the above example. That is, taking the above supply condition as an example, as a penalty value when the number of times specified by the condition is twice, the product of the square of the number of times below and the weight, ie, 2²A value of × 20 = 80 may be used. In addition to the above examples, there are various types of conditions for each staff member. For example, a condition such as “acquire three consecutive holidays once a month” is an example of an allocation condition that can be expressed as a condition for each staff member. In this case as well, if a penalty value calculation method is defined such that the value increases as the degree of violation of the condition increases, it can be handled in the same manner as in the above example.
[0063]
In the first stage, since there are usually many blanks and the contents are often not appropriate, the overall evaluation value is not very good. However, due to automatic repetition of assignment to the entire work schedule (in this embodiment, repetition of scanning with division), the contents of the work schedule become better in stages, and finally a high-quality work schedule is obtained. .
[0064]
In FIG. 7, in S120, the first day of the work schedule creation period set as described above is set as the date D. In S125, allocation of all work schedules (with division) for all staff for d days starting from the date D is executed, and the contents of the stored work schedule are partially updated. An example of this is shown in FIG. 8, which will be described in detail later.
[0065]
In S130, the date D is moved to the next day and the division is repeated. However, if the date D + (d−1) exceeds the last day of the creation period in S135, the iteration is terminated and the process proceeds to the next step. In addition, in the above, an example is shown in which the allocation is performed in order from the first day of the period subject to automatic allocation, but the order is not limited to this, going back from the day after the period, Alternatively, the target days may be selected in an appropriate order. In addition, here, an example is shown in which allocation is performed in units of d consecutive days. However, a single day (d = 1) or multiple days in the work schedule creation period can be handled in the same manner. . Furthermore, the number of days d to be allocated at a time does not need to be constant throughout the work schedule, and can be handled essentially in the same manner even if it differs every time divisional allocation is repeated.
[0066]
In S140, when the above iterative process ends, the evaluation value calculation unit 40 calculates the overall evaluation value based on the work schedule stored in the work table storage unit 48 shown in FIG. Store in E. Subsequently, in S145, the value of the variable E in the evaluation value calculation unit 40 is compared with the value of the variable E0 calculated in S115. If the value of the variable E is smaller than the value of the variable E0, the content of the work table is improved by the previous iteration of S125 to 135, and there is a possibility that the work table can be further improved by further iterations. In S150, the value of variable E is newly stored in variable E0, and each step after S120 is repeated. On the other hand, if the value of the variable E is equal to or greater than the variable E0, the work schedule stored at that time is displayed on the display unit 28 in S160, and the automatic allocation process is performed. Exit. It is also possible to display on the display unit 28 immediately after S125.
[0067]
In this case, the end of the iteration is determined by calculating the overall evaluation value at the stage when the assignment is completed until the last day of the work schedule, but the method for determining the end is a method based on the upper limit of the number of iterations or elapsed time, or Immediately after S125, the same process as S145 is performed every time to check the change in the overall evaluation value, and if the improvement of the overall evaluation value is not seen until the number of iterations of S125 to 135 reaches a certain number of times, the process ends. You can also
[0068]
Next, specific contents of the process shown in S125 of FIG. 7 will be described in detail with reference to FIG.
[0069]
In S200, in order to construct an individual evaluation matrix, a list of work patterns is prepared for d days. As described above, FIG. 3 shows various work patterns in the case of d = 3 with four types of work, “day off”, “day shift”, “semi-night”, and “late night”. As described above, the d days do not necessarily have to be continuous. In other words, if a skipped day such as April 2, April 4, or April 6 is selected as the target of division, each work in the work pattern shown in FIG. If it is considered to correspond to April 4 and April 6, it can be handled in the same manner.
[0070]
In S205, one arbitrary staff member (referred to as staff member s) is selected, and in S210, one pattern is selected from a series of work patterns (this pattern is referred to as pattern i). Subsequently, in S215, it is assumed that d days (see reference numeral 56 in FIG. 5) starting from the date D in the cell column of the staff s in the work schedule are replaced with the pattern i, and the plurality of first allocation conditions Calculate the penalty value. In S225, the penalty values are summed up, and the added value is used as the individual evaluation value (see reference numeral 68 in FIG. 5), and the individual evaluation value is entered in the corresponding cell in the individual evaluation matrix.
[0071]
As described above, the individual evaluation matrix is a table having a size of (the number of staff) × (the number of work patterns), and the individual evaluation value is in the position of (s, i). Stored (see FIG. 4).
[0072]
In S230, it is determined whether or not all work patterns have been provisionally allocated. If not, the work pattern is changed in S270, and the processes from S215 are repeated. Further, in S235, it is determined whether or not the above steps have been executed for all staff members. If not completed, the next staff member is selected in S265, and each process from S210 is repeatedly executed. . Through this process, individual evaluation values are entered in all cells in the individual evaluation matrix.
[0073]
In S240, the constraint expression generator 34 first determines whether to assign the work pattern i to the staff s 0-1 variable x_siAnd the following constraint expression is generated using the auxiliary variables described below.
[0074]
1) A constraint expression indicating that only one work pattern is assigned to one employee s is as follows. A similar constraint formula is generated for each staff member.
[Expression 2]

[0075]
2) The constraint formulas indicating that work that can be assigned is limited, such as desired work and holidays for each staff member, are as follows. For each staff member, a similar constraint expression is generated for each day for which work is limited within d days starting from date D.
[Equation 3]

[0076]
3) A constraint equation corresponding to an absolute weight among the conditions (second allocation condition) for each day regarding d days starting from the date D is as follows.
(Example 1) Requirement condition (1) Constraint formula for "10 employees should have 10 day shifts (weight: absolute)"
A similar constraint expression is generated for each day of d days starting from the date D.
[Expression 4]

(Example 2) Requirement condition (2) Constraint formula for “Low-time work is set to one lower limit for“ Veteran ”staff (weight: absolute)”
A similar constraint expression is generated for each day of d days starting from the date D.
[Equation 5]

(Example 3) Combination condition (1) Constraint formula for "Staff E is a day shift and is paired with one of staff F, G, H, or I (weight: absolute)"
However, C is a sufficiently large number, and a similar constraint expression is generated for each day of d days starting from the date D.
[Formula 6]

[0077]
4) A constraint equation corresponding to a non-absolute weight among conditions (second allocation condition) for each day regarding d days starting from the date D is as follows.
(Example 1) Requirement condition (3) Constraint formula for “Midwives have a minimum of two day shifts (weight: 100)”
A similar constraint expression is generated for each day of d days starting from the date D.
The following auxiliary variables are also used for each day.
[Expression 7]

However,
[Equation 8]

Is an auxiliary variable that satisfies the following conditions.
[Equation 9]

Auxiliary variables
[Expression 10]

Is also incorporated for the objective function described below.
[0078]
In addition, in 2) above, the constraint equation when the desired leave for each staff member is specified is shown. Instead of using such a constraint equation, work patterns used in other constraint equations and objective functions May be restricted. In other words, for example, when a day off corresponding to the date D of the staff member A is set as a “day off” request, the work on the date D is always “day off”, so the date D in the work pattern is “day off”. For those that are not, it is possible to exclude them from the choice of work pattern when generating the constraint equation for the staff A, that is, do not include the corresponding variable in the constraint equation and the objective function.
[0079]
In FIG. 8, an objective function is generated in S245. The objective function is expressed as follows.
## EQU11 ##

Here, each term on the right side is as follows.
[0080]
a) The value of the position of (s, i) on the variable and penalty matrix (a_siSum of products (represents the sum of penalty values for each employee's condition)
[Expression 12]

b) Terms related to the condition indicating the upper limit among the conditions for each day whose weight is not absolute
[Formula 13]

c) A term related to a condition indicating a target among conditions for each day whose weight is not absolute.
[Expression 14]

d) Terms related to the condition indicating the lower limit among the daily conditions where the weight is not absolute.
[Expression 15]

[0081]
In FIG. 8, in S250, all the variables x are satisfied so as to satisfy the above constraint expressions and to minimize the value of the objective function._si0-1 values and auxiliary variable values are determined. Where the variable x to be determined_siThe 0-1 value specifies the optimum work pattern for each staff member, as will be described later.
[0082]
Since the above constraint equation and objective function are formulated as an integer programming problem, a known method (Branch & Bound method or the like) may be used as an integer programming method for determining variable values. This makes it possible to generate an ideal work pattern set that satisfies the absolute conditions and satisfies the other conditions as much as possible.
[0083]
However, depending on the contents of various constraint equations, there is a case where a solution cannot be obtained (there is no solution). For example, if there are only two employees who have “Veteran” qualifications on a daily basis, and if both of them have the desired day off on the same day, it will be quasi-night for employees with the demand condition “Veteran” “Working as one lower limit (weight: absolute)” can never be satisfied, and there is no solution.
[0084]
Therefore, after the process of obtaining the solution, it is determined whether or not the solution is actually obtained in S260, and if the solution is not obtained, the work schedule described below is not updated.
[0085]
In integer programming, it is known that the calculation time increases when the number of variables is large. Therefore, to increase the speed, it is desirable to reduce the number of variables as much as possible. So variable x_siOf the position (s, i) on the corresponding individual evaluation matrix (a_siVariable) x is greater than a certain threshold_siCan be excluded from the constraint equation and the objective function. Ie a_siIs large, the corresponding variable x_siIf 1 is set to 1, the objective function value increases, so in the first place such a variable x_siIs unlikely to be 1. Therefore, it is possible to increase the speed by not using such a variable in advance.
[0086]
In S260 of FIG. 8, when the solution is obtained in the above, one variable is set for each staff member s, and the number of staff members is determined simultaneously. Therefore, for d days starting from the date D in the work schedule, an operation for assigning a work pattern corresponding to the value 1 is performed for each staff member s. Thereby, the division with the partial period is completed.
[0087]
Then, as shown in FIG. 7, when such division is sequentially performed, the assignment (entry or update) to the entire work schedule is completed. The assignment for the entire work schedule is repeated until the quality of the entire work schedule satisfies a certain condition.
[0088]
As described above, if complicated and diverse allocation conditions are handled together in an unseparated state, and if an attempt is made to perform allocation processing for the entire fixed period at the same time, the amount of computation will inevitably become complicated and enormous, resulting in a decrease in allocation accuracy. It is easy to invite. On the other hand, according to the above-described embodiment, a partial range is set as an object to be allocated at a time, and each allocation candidate is provisionally individually evaluated for the first allocation condition group in the time axis direction ( The evaluation is for the entire allocation table in consideration of quality), and then the optimum combination of candidates is determined using the result of the aggregated individual evaluation and the second allocation condition group regarding the human composition. . In other words, in the solution derivation calculation, the calculation conditions are reduced by range limitation and pre-aggregation to achieve quick calculation. Then, if necessary, the entire development with division is repeated until the quality is satisfactory, and the quality of the entire work schedule is improved. Therefore, there is an advantage that a work schedule creation (support) system having high practicality and usefulness can be provided.
[0089]
As a modification, first, an individual evaluation matrix may be generated using the second allocation condition group, and the content with partial division may be determined using the individual evaluation matrix and the first allocation condition group. The above system can also be applied to the creation of other assignment tables such as work assignment tables other than work schedules.
[0090]
【The invention's effect】
As described above, according to the present invention, it is possible to quickly create an allocation table having high-quality contents.
[Brief description of the drawings]
FIG. 1 is an explanatory diagram showing an example of a work schedule created by a system according to the present invention.
FIG. 2 is a block diagram showing a preferred embodiment of the system according to the present invention.
FIG. 3 is a diagram showing types of work patterns (assignment candidates) specified by pattern (candidate) numbers.
FIG. 4 is an explanatory diagram showing an individual evaluation matrix.
FIG. 5 is a diagram for explaining a penalty calculation based on a first allocation condition group and a second allocation condition group.
FIG. 6 is a diagram illustrating a flow of determining a solution (content with partial division) based on an integer programming problem solution.
7 is a flowchart for explaining the operation of the system shown in FIG. 2;
FIG. 8 is a flowchart for explaining specific contents of S125 of FIG.
[Explanation of symbols]
22 arithmetic processing units, 24 work schedule creation program (assignment execution unit), 32 penalty calculation unit, 34 constraint expression generation unit, 36 objective function generation unit, 38 solution derivation unit, 40 evaluation value calculation unit, 42 end condition determination unit, 44 master information storage unit, 46 allocation condition storage unit, 48 work schedule storage unit, 50 individual evaluation matrix (individual evaluation result storage unit).

Claims (4)

In a system for creating an allocation table having a first axis representing the certain period and a second axis representing the plurality of members by allocating work schedules to each unit of time within a certain period for a plurality of members constituting the organization ,
A partial range setting means for setting a partial region consisting of multiple time units Te said predetermined period in smell,
A first allocation condition group consisting of a plurality of allocation conditions relating to a series of work configurations in the direction of the first axis, and a second allocation condition group consisting of a plurality of allocation conditions relating to member configurations of each time unit in the direction of the second axis And an assignment condition storage unit for storing
Using the first assignment condition group and the second assignment condition group stored in the assignment condition storage unit, a part attaching unit that performs part assignment with scheduled work for the set partial range;
A matrix storage unit for storing an individual evaluation matrix generated in the partial division;
Including
The parting means is
Within the partial range, and the tentative allocation means for each member, be provisionally allocated a plurality of allocation candidates as a sequence of work scheduled to be taken,
Based on the first allocation condition groups stored in the assignment condition storage unit, individual I rows individual evaluation as an evaluation for the entire of the predetermined period for each allocation candidate for each member the temporary allocation is made An individual evaluation unit that calculates an evaluation value and thereby generates an individual evaluation matrix in which the individual evaluation value is written on the matrix storage unit;
Based on the individual evaluation matrix stored in the matrix storage unit and the second allocation condition group stored in the allocation condition storage unit, an optimal combination of a plurality of allocation candidates for the plurality of members is determined, and Optimal combination assignment means for assigning to the partial range;
An allocation table creation system characterized by comprising: The system of claim 1, wherein
The individual evaluation value is a summation of penalty values indicating the degree of violation of each allocation condition of the first allocation condition group . The system of claim 1, wherein
The partial range setting means repeatedly sets the partial range while changing the setting position within the fixed period,
The part dividing unit executes the parting at each setting position that is repeatedly specified, thereby completing the whole assignment for the assignment table,
An allocation table creating system comprising: an end determination unit that repeatedly executes the entire allocation until a predetermined end determination condition is satisfied. The first axis representing the certain period and the second axis representing the plurality of members are assigned to a plurality of members that are read and executed by a computer, and assign work schedules to each unit of time within the certain period. An allocation table creation program for creating an allocation table having
The allocation table creation program is
A function of setting a partial region consisting of multiple time units Te said predetermined period in smell,
A first allocation condition group consisting of a plurality of allocation conditions relating to a series of work configurations in the direction of the first axis, and a second allocation condition group consisting of a plurality of allocation conditions relating to member configurations of each time unit in the direction of the second axis And a function for storing in the allocation condition storage unit,
A function of performing division of work scheduled for the set partial range using the first allocation condition group and the second allocation condition group stored in the allocation condition storage unit ;
A function of storing an individual evaluation matrix generated in the division with a matrix storage unit;
Including
The function to execute the division is as follows:
Within the partial range, a function for each member, be provisionally allocated a plurality of allocation candidates as a sequence of work scheduled to be taken,
Based on the first allocation condition groups stored in the assignment condition storage unit, individual I rows individual evaluation as an evaluation for the entire of the predetermined period for each allocation candidate for each member the temporary allocation is made A function of calculating an evaluation value, thereby generating the individual evaluation matrix in which the individual evaluation value is entered on the matrix storage unit;
Based on the individual evaluation matrix stored in the matrix storage unit and the second allocation condition group stored in the allocation condition storage unit, an optimal combination of a plurality of allocation candidates for the plurality of members is determined, and The ability to assign to partial ranges,
An allocation table creation program characterized by comprising:

Images