JP2004227301A - Shipment amount predicting device and method for product and computer readable recording medium - Google Patents

Shipment amount predicting device and method for product and computer readable recording medium Download PDF

Info

Publication number
JP2004227301A
JP2004227301A JP2003014532A JP2003014532A JP2004227301A JP 2004227301 A JP2004227301 A JP 2004227301A JP 2003014532 A JP2003014532 A JP 2003014532A JP 2003014532 A JP2003014532 A JP 2003014532A JP 2004227301 A JP2004227301 A JP 2004227301A
Authority
JP
Japan
Prior art keywords
contract
information
shipment
similar
result
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
JP2003014532A
Other languages
Japanese (ja)
Other versions
JP4268408B2 (en
Inventor
Kiyoshi Wajima
潔 和嶋
Kenji Sugiyama
賢司 杉山
Hiroaki Miyoshi
浩昭 三好
Naohiro Tomita
尚宏 冨田
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nippon Steel Corp
Original Assignee
Nippon Steel Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Nippon Steel Corp filed Critical Nippon Steel Corp
Priority to JP2003014532A priority Critical patent/JP4268408B2/en
Publication of JP2004227301A publication Critical patent/JP2004227301A/en
Application granted granted Critical
Publication of JP4268408B2 publication Critical patent/JP4268408B2/en
Anticipated expiration legal-status Critical
Expired - Fee Related legal-status Critical Current

Links

Images

Classifications

    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Landscapes

  • General Factory Administration (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

<P>PROBLEM TO BE SOLVED: To estimate the shipment amount of products when the several number of times of shipping of products may be required for one order reception contract from the past shipping result information. <P>SOLUTION: The similarity of contract examples is evaluated by using numerical information and non-numerical information in a contract to be predicted, and a plurality of similarity contracts are extracted. A timing which is the most similar to a timing to be predicted is extracted in the shipping patterns of the similar contracts, and predicted shipment is estimated form the similar contracts. The predicted shipment from the plurality of similar contracts is multiplied by linear connection coefficients calculated from the similarity so that the predicted shipment amount can be calculated. <P>COPYRIGHT: (C)2004,JPO&NCIPI

Description

【０００１】
【発明の属する技術分野】
本発明は、製品の出荷量予測装置、方法及びコンピュータ読み取り可能な記憶媒体に関し、特に鉄鋼製品のような、一つの受注契約に対して製品の出荷が複数回にわたる場合があるような製品の出荷量を契約情報から予測するために用いて好適な技術に関する。
【０００２】
【従来の技術】
製品の出荷量予測は、生産計画の立案や、製品在庫の不足や過剰の防止、顧客からの注文に対する納期予測、更には数ヵ月後の収益予測を行う為に重要な指標である。通常、この出荷量予測は、生産管理担当者や販売担当者が、各自が担当する製品を購入する顧客の需要動向、さらには受注量と生産に必要な工期を考慮して、経験的に予測を行っていた。
また、特許文献１に開示された手法では、過去の製品出荷履歴と日付の実績データを用いて、多層神経回路網（ｍｕｌｔｉ ｌａｙｅｒ ｎｅｕｒａｌ ｎｅｔｗｏｒｋ）による出荷量推定装置を学習させ、この製品出荷量推定装置に、予測対象期間直前の出荷量履歴と日付情報を入力し、予測出荷量推定値を算出することで、過去の実績に基づく予測を可能としている。
【０００３】
【特許文献１】
特開平６−１４９８４９号公報
【０００４】
【発明が解決しようとする課題】
しかしながら、製品毎の出荷量予測を人の経験に頼る従来の方法では、担当者の個人差によって予測精度バラツキが有ることに加えて、多数の製品や顧客を抱える場合には、製造工期や需要動向を詳細に考慮することが困難である為、勘に基づいて予測を行わざるを得ず、精度が良くない問題があった。
【０００５】
また、特許文献１に開示された手法では、過去の製品出荷履歴とカレンダー情報の実績データを用いて、多層神経回路網（ｍｕｌｔｉ ｌａｙｅｒ ｎｅｕｒａｌ ｎｅｔｗｏｒｋ）に基づく出荷量推定装置を学習させ、予測を行っている。しかしながら、実際の出荷予測対象製品においては、多層神経回路網の学習に十分なデータ数が得られるほど、常に安定して注文がある製品ばかりではなく、出荷が全く無い期間が実績データに多く含まれるような製品については、データ量が不十分である為に出荷量推定装置の学習が困難である問題があった。更に、出荷量予測の判断材料となる因子の多くは、製品の種類、需要家、生産工場、流通経路、輸送手段など一般にはコードで表現される非数値情報であり、これらに加えて、総出荷量などの数値情報も考慮して予測を行わねばならないが、多層神経回路網は数値情報を入力して、数値情報を出力する推定装置である為、非数値で表現されている因子の情報を有効に予測に使用できず、必ずしも十分に高い精度が得られない、という問題もあった。
【０００６】
本発明は、上記のような点に鑑みてなされたものであり、一つの受注契約に対して製品の出荷が複数回にわたる場合があるような製品の出荷量を、過去の出荷実績に基づいて予測できるようにすることを目的とする。特に、過去に同様の出荷実績事例が少ないような製品においても、需要家や流通経路など非数値情報を有効に活用して類似の出荷事例を探索し、これら類似事例に基づく適切な予測ができるようにすることを目的とする。
【０００７】
【課題を解決するための手段】
本発明の製品の出荷量予測装置は、一つの受注契約に対して製品の出荷が複数回にわたる場合がある製品の過去の出荷実績情報、及び前記過去の出荷実績情報に対する契約情報を使用して、予測対象期間における予測対象契約からの出荷量を予測する「製品の出荷量予測装置」であって、過去の契約実績情報とそれに対応する複数回にわたる製品出荷実績パターン情報を入力する「実績データ入力手段」と、前記実績データ入力手段から入力された実績情報を蓄積する「実績データ蓄積手段」と、予測対象となる契約情報と当該契約から既に発生した出荷パターン情報を入力する「予測対象情報入力手段」と、前記予測対象となる契約情報と前記過去の契約実績情報の類似度を計算する「契約類似度計算手段」と、前記契約類似度計算手段で算出した類似度に基づき所定の類似度を有する複数の過去の類似契約実績を選び出す「類似契約実績抽出手段」と、前記予測対象契約における既に発生した出荷パターン情報と選び出した前記類似契約実績の出荷実績パターンから、予測対象の出荷タイミングに最も類似した前記類似契約実績における出荷タイミングを抽出し、前記類似契約実績における最類似出荷タイミングの出荷量を抽出する「最類似タイミング出荷量抽出手段」と、前記複数の類似契約事例それぞれに対して抽出された最類似タイミング出荷量に、前記類似度に基づいて重み付けを行って、前記予測対象情報に対する出荷量を演算する「出荷量演算手段」と、前記出荷量演算手段から出力された予測出荷量を表示する「予測結果出力手段」を備えた点に特徴を有する。
【０００８】
本発明の製品の出荷量予測装置の他の特徴とするところは、前記契約類似度計算手段が、契約情報における出荷量のような数値属性情報と、製品の種類や出荷先のような非数値属性情報の両者を用いて、前記予測対象となる契約情報の各項目と対応する前記過去の契約実績情報の各項目とのノルムを評価することにより、類似度計算処理するようにした点にある。
また、本発明の製品の出荷量予測装置の他の特徴とするところは、前記契約類似度計算手段が、契約情報における複数の項目を用いて類似度計算処理する場合に、前記複数の項目それぞれに対する重要度を個別に設定して類似度処理するようにした点にある。
また、本発明の製品の出荷量予測装置の他の特徴とするところは、前記契約類似度計算手段が、契約情報における数値属性情報の項目を用いて類似度計算処理する場合に、前記数値属性情報項目の数値の範囲に基づいて決定される重み付けを行って類似度処理するようにした点にある。
また、本発明の製品の出荷量予測装置の他の特徴とするところは、前記最類似タイミング出荷量抽出手段が、契約総出荷量から予測対象期間前の累積出荷量を差し引いた契約残量を指標として最類似出荷タイミング量を抽出するようにした点にある。
また、本発明の製品の出荷量予測装置の他の特徴とするところは、前記過去の契約実績情報に、製品の種類情報、総出荷量情報、需要家情報、生産工場情報、流通経路情報、輸送手段情報のいずれか一つ、或いは複数を含むようにした点にある。
また、本発明の製品の出荷量予測装置の他の特徴とするところは、鉄鋼製品の出荷量予測に適用され、前記過去の出荷実績情報は、薄板、厚板、条鋼などの鉄鋼製品受注契約に対する月別の出荷量情報である点にある。
【０００９】
本発明における製品の出荷量予測方法は、一つの受注契約に対して製品の出荷が複数回にわたる場合がある製品の過去の出荷実績情報、及び前記過去の出荷実績情報に対する契約情報を使用して、予測対象期間における予測対象契約からの出荷量を予測する「製品の出荷量予測方法」であって、過去の契約実績情報とそれに対応する複数回にわたる製品出荷実績パターン情報を入力する「実績データ入力工程」と、前記実績データ入力工程から入力された実績情報を蓄積する「実績データ蓄積工程」と、予測対象となる契約情報と当該契約から既に発生した出荷パターン情報を入力する「予測対象情報入力工程」と、前記予測対象となる契約情報と前記過去の契約実績情報の類似度を計算する「契約類似度計算工程」と、前記契約類似度計算工程で算出した類似度に基づき所定の類似度を有する複数の過去の類似契約実績を選び出す「類似契約実績抽出工程」と、前記予測対象契約における既に発生した出荷パターン情報と選び出した前記類似契約実績の出荷実績パターンから、予測対象の出荷タイミングに最も類似した前記類似契約実績における出荷タイミングを抽出し、前記類似契約実績における最類似出荷タイミングの出荷量を抽出する「最類似タイミング出荷量抽出工程」と、前記複数の類似契約事例それぞれに対して抽出された最類似タイミング出荷量に、前記類似度に基づいて重み付けを行って、前記予測対象情報に対する出荷量を演算する「出荷量演算工程」と、前記出荷量演算工程から出力された予測出荷量を表示する「予測結果出力工程」を備えた点に特徴を有する。
【００１０】
本発明のコンピュータ読み取り可能な記憶媒体は、上記製品の出荷量予測装置における各手段として、コンピュータを機能させるプログラムを記録した点に特徴を有する。
本発明のコンピュータ読み取り可能な記憶媒体の他の特徴とするところは、上記製品の出荷量予測方法をコンピュータに実行させるためのプログラムを記録した点に特徴を有する。
【００１１】
【発明の実施の形態】
以下に、図面を参照して、本発明の製品の出荷量予測装置、方法及びコンピュータ読み取り可能な記憶媒体について説明する。
図１は、本実施の形態による製品の出荷量予測装置の構成を示す図である。同図の１０１は、実績データ入力部であり、出荷量予測装置には、過去の契約情報と、その契約に対応する出荷パターン情報が、例えばコンピュータシステムのキーボード入力装置を用いてインプットされる。上記契約情報は、例えば鉄鋼製品の受注契約に係わる製品コード、需要家コード、上記製品を生産するミルのコード、製品の受け渡し場所及び受け渡し条件コード、輸送手段コードなどの非数値情報と、契約総出荷トン数などの数値情報で構成されているのが、一般的である。また、上記出荷パターンは、受注契約後の期間を、例えば１ヶ月等の単位期間に分割し、上記契約から各単位期間にそれぞれ何トンの出荷実績が発生したかを示すものである。図２に、鉄鋼薄板製品の契約情報と一ヶ月を単位期間とした出荷パターン情報の例を示す。
【００１２】
１０２は、実績データ蓄積部であり、前記実績データ入力部１０１よりインプットされた契約情報と出荷パターン実績情報を蓄積し、かつ参照の指示が入力された場合は、蓄積された情報を随時、出力するものである。データ蓄積部を実現する手段としては、例えば、データベースソフトウェアと計算機を用いる方法がある。
【００１３】
１０３は、予測対象情報入力部であり、出荷量を予測したい受注契約に係わる契約情報、例えば鉄鋼製品の受注契約における製品コード、需要家コード、ミルのコード、製品の受け渡し場所及び受け渡し条件コード、輸送手段コード、及び契約総出荷トン数を入力する。更に上記予測したい受注契約から、既に出荷が発生している場合には、上記契約情報に加えて、既に発生した出荷パターン情報を入力する。
【００１４】
１０４は、契約類似度計算部であり、例えばＣＰＵ、メモリ、及び計算アルゴリズムを記憶したハードディスク等を備えたコンピュータを用いて実現される。予測対象情報入力部１０３で入力した予測対象の契約情報と、実績データ蓄積部１０２に蓄積された実績契約情報を比較し、全ての契約実績について、予測対象契約との類似度評価を行う。コード等の非数値情報と、数値情報の両者を用いて類似度を評価する方法としては、例えば式（１）に示すように、数値情報の項目についてはユークリッド距離等、幾何学的に評価される距離（ノルム）を用い、非数値情報の項目については、比較すべき両者のコードが同一コードの場合ノルム０、異なるコードの場合ノルム１と定義して、全ての項目のノルムを加算したものを類似度とする方法がある。
【数１】

Figure 2004227301
ここで、非数値情報項目のノルムが０或いは１の値を取ることに対応させる為、数値情報についても、ａｋ ｉ及びａｋ ｊの正規化処理を行い、ノルムが０から１の範囲となるようにしている。式（１）より、類似性が高い場合Ｄｉｊは小さい値となり、契約ｉとｊが全く同一条件を有する場合には、Ｄｉｊ＝０と評価されることが判る。
また、式（１）では、契約情報における全ての項目が類似度に対して同一の寄与率として評価しているが、ある特定の項目が、契約間の類似性を評価する上で、他の項目よりも寄与率が高いことが判明している場合、式（２）に示すように項目それぞれに対して重要度を設定して類似度処理を行うことが出来る。
【数２】
Figure 2004227301
重み係数が大きく設定された項目は、他の項目に比べてノルムの差が大きく評価される為、結果的に、重み係数が大きく設定された項目が類似した契約を、類似性が高いと評価する。
【００１５】
更に、式（１）及び（２）において、数値情報項目の数値が広いレンジに分布しており、かつ契約の類似性を評価する上で、数値が存在する範囲情報の寄与率が高いと判明している場合、式（３）に示すように、２つの契約それぞれの数値情報の範囲に応じて、重み付けを行い、類似度処理を行っても良い。
【数３】
Figure 2004227301
ここで、ｗｋ ｉ′ｊ′は、対角要素の値が１、ｎ行ｎ列の対称行列
【数４】
Figure 2004227301
の要素である。ここでＷｋ ｉｊの行及び列のインデックスは、ａｋの値の範囲と例えば以下のように対応している。
【数５】
Figure 2004227301
ａｋ ｉ及びａｋ ｊそれぞれについて、値に応じて行と列のインデックスを決定し、このインデックスに対応するＷｋ ｉｊの要素の値を重み係数として使用する。このため、ａｋ ｉ及びａｋ ｊ同一の範囲に存在する場合は、重み１として評価されるが、両者が異なる範囲にある場合は、式（４）に応じた重み付けを行って、類似度処理が行われる。
【００１６】
１０５は類似契約契約実績抽出部であり、契約類似度契約部１０４にて算出した類似度を用いて、予測対象契約に類似した契約実績を抽出する。類似契約の抽出方法としては、契約実績を類似性の高い順番にソートし、別途装置に入力された抽出契約個数分だけ類似性の高い契約から選択する方法がある。また、これとは別に、別途装置に入力された類似度限界指標を用いて、前記類似度限界指標よりも高い類似性を持っている契約実績を全て類似契約とする方法でも良い。
【００１７】
１０６は、最類似出荷タイミング出荷量抽出部である。類似契約実績抽出部１０５で得られた類似契約実績の出荷パターンと、予測対象情報入力部１０３で入力された既に発生した出荷パターン情報に基づいて、類似契約実績の出荷タイミングの中で、予測対象契約の出荷タイミングに最も類似したタイミングを抽出し、更に類似契約実績からの予測出荷量を抽出する。最類似タイミングの抽出方法としては、例えば、単位期間毎に契約総出荷量から予測対象期間前の累積出荷量を差し引き、更に契約総出荷量で除した正規化契約残量を求めて、予測対象契約と類似契約の正規化契約残量を比較し、類似契約の正規化契約残量パターンにおいて最も予測対象契約の正規化契約残量に近いタイミングを最類似出荷タイミングとする方法がある。この方法を以下、例に従って説明する。
【００１８】
図３に予測対象契約から既に発生した出荷パターン例、図４に類似契約事例の出荷パターンを、月毎の表形式にした例を示す。この例では、予測対象契約におけるＮ＋２月の出荷量を予測することが目的であり、また類似契約事例は３件抽出されている。図３、図４のそれぞれにおいて、上記に述べた手順に従って正規化契約残量を月毎に算出し、表形式にしたものを、図５及び図６に示す。図５におけるＮ＋１月の正規化契約残量は５０％であり、図６における類似契約の出荷パターンにおいて、これに最も近い正規化契約残量を有しているタイミングを抽出すれば、Ｎｏ１の類似契約事例ではＮ＋２月、Ｎｏ２の類似契約事例ではＮ＋１月、Ｎｏ３の類似契約事例ではＮ＋２月が最類似出荷タイミングと評価される。更に図４の類似契約からの出荷パターンと、上記の手順で求められた最類似出荷タイミング情報を用いて、予測対象契約のＮ＋２月出荷量に対応する推定出荷量は、図７のように求められる。
【００１９】
１０７は、出荷量演算部であり、最類似出荷タイミング出荷量抽出部１０６で得られた類似契約毎の予測出荷量と、契約類似度計算部１０４で算出した類似契約の類似度に基づいて、予測出荷量を算出する。予測出荷量の計算方法としては、例えば式（１）〜（３）のいずれかに基づいて計算された類似度Ｄｉｊの逆数を用いて各類似契約に対する重み係数を算出し、各類似契約からの予測出荷量に重み係数を乗じて、線形和を求める方法がある。
【００２０】
１０８は、出荷量予測結果表示部であり、対象契約の出荷量予測結果、及びその推定の根拠となった類似契約事例と出荷パターン、最類似出荷タイミングなど、予測結果の評価に有効な情報を、例えばコンピュータシステムにおけるＣＲＴや、或いはプリンタ等の印刷手段によって出力する。
【００２１】
以上に述べた本実施の形態による製品の出荷量予測装置によれば、予測対象の契約情報に類似した過去の契約実績を探索する際に、数値情報にのみならず、顧客や輸送手段情報など出荷量推定に影響の高い非数値情報も考慮する為、高い精度で予測を行うことが可能である。また、過去に出荷実績が少ない契約事例についても、存在する契約実績の中で、最も類似した契約を抽出し、その出荷実績に基づいた予測を行うことができる。
【００２２】
【実施例】
以下では、鉄鋼薄板製品の出荷量を本手法で予測した実施例について説明する。
今回の実施例では、５年間における約１５万件の契約情報、及び各契約に対する月毎の出荷実績情報を用いて、出荷量予測を行った。契約情報のうち、最初の４年分を前記実績データ蓄積部に保存し、残り１年分の出荷量を予測することとした。予測出荷量については、更に実績値と比較することで予測精度を評価している。
【００２３】
契約の類似性を評価する為の項目としては、数値情報である契約総出荷トン数に加え、非数値情報である品名コード、ユーザーコード、商社コード、営業部門コード、製品の受け渡し条件コード、受け渡し場所コード、輸送形態コード、輸送機関コードを使用した。
また、類似度指標の計算においては、薄板製品の出荷パターンが、特に総契約出荷トン数に大きく影響を受けることに着目し、非数値情報項目に対する重み係数は全て１に設定した状態で、総契約出荷トン数に対する重み係数Ｑを種々変更して、予測対象契約と抽出された類似契約の出荷パターンを比較した。その結果、Ｑ＝３において、最も類似性の高い出荷パターンが多く抽出されると評価された為、以降、この条件で出荷予測を行っている。
【００２４】
また、上記総出荷トン数と出荷パターンの関連を調査する過程で、総出荷トン数の範囲によって出荷パターンが概ねグループ化できることが明らかになったため、総契約トン数の範囲に応じたグルーピングと、各グループ間の重み係数を種々変更して、予測対象契約と抽出された類似契約の出荷パターンを比較した。その結果、総契約出荷トン数を、グループ１：総契約出荷トン数０〜１０トン、グループ２：総契約出荷トン数１０〜１００トン、グループ３：総契約出荷トン数１００〜１０００トン、グループ４：総契約出荷トン数１０００トン以上の４つに分類し、式（４）におけるＷｉｊ行列を以下のように設定した場合に、最も類似性の高い出荷パターンが抽出されると評価された為、以降、この条件で予測を行っている。
【数６】
Figure 2004227301
ここで、行列のインデックス１から４は上記グループの１〜４に対応するものである。
【００２５】
次に、類似契約事例の抽出については、契約実績の中で類似性が高いものからＰ個の事例を抽出するものとし、Ｐの値を種々変更して予測出荷量を算出、出荷実績と比較した精度評価で最も、誤差が小さくなる条件を探索した。その結果、Ｐ＝３とした場合に、最も誤差が小さかった為、以下、この条件で予測を行っている。
【００２６】
図８は、予測対象契約の情報と、本手法によって抽出された３つの類似契約の情報、及びそれぞれの出荷パターンである。各類似契約には、予測対象との類似度も記載している。予測対象契約からは、Ｎ＋１月分までの出荷が既に発生しており、Ｎ＋２月分の出荷量を予測することが目的である。本手法によって、類似契約における最類似出荷タイミングを抽出した結果、いずれの類似契約の場合もＮ＋２月と抽出された。類似契約の類似度指標から逆数を求め、更に線形和係数の合計が１となるように正規化したものを図９に示す。この線形和係数と推定出荷量を乗じて、３つの類似契約について加えたものは、３９７．５トンである。予測対象事例のＮ＋２月における出荷実績は、４１２トンであった為、この契約事例に関する予測誤差は３．５％と、比較的良好であった。
【００２７】
図１０は、月毎に全ての契約から、特定の鋼種に対する出荷量を予測し、実績出荷量と併記してプロットしたものである。本手法による平均予測誤差は、約２．５％であり、例えば従来の人手による予測よりも、安定して高い精度で出荷量を予測できている。
この結果を利用して、各製鉄所に対する生産量の配分計画や、製品在庫の推移予測を従来よりも精度よく行うことが可能となる効果を得ることができた。
【００２８】
なお、今回の実施例では、コンピュータ上のプログラムとして出荷量予測装置を実現したが、演算装置、メモリ等を組み合わせたハードウェアによって構成されるものであっても良い。
また、本発明の製品の出荷量予測装置は、複数の機器から構成されるものであっても、一つの機器から構成されるものであっても良い。
また、上述した実施の形態は、コンピュータのＣＰＵ或いはＭＰＵ、ＲＡＭ、ＲＯＭ等で構成されるものであり、ＲＡＭやＲＯＭに記録されたプログラムが動作することで実施される。したがって、前記実施の形態の機能を実現するためのソフトウェアのプログラムコードをコンピュータに供給するための手段、例えばかかるプログラムコードを格納した記憶媒体は本発明の範疇に含まれる。
【００２９】
【発明の効果】
以上に述べたように、本発明によれば、契約情報における数値情報と、非数値情報の両者を活用して類似契約事例を抽出し、更に出荷パターン情報を用いて、類似した出荷タイミングから出荷量を推定することにより、より精度の高い出荷量予測を可能としている。
また、過去に同様の事例が少ない為に、出荷予測モデルを学習によって得ることが困難であったような場合でも、非数値情報を活用して類似事例を探索する為、これら類似事例に基づく適切な予測が可能である。
【図面の簡単な説明】
【図１】本発明に係る実施形態の製品の出荷量予測装置の構成を示す図である。
【図２】製品の契約実績データと出荷パターンデータを示す図である。
【図３】予測対象契約から既に発生した出荷のパターンデータを示す図である。
【図４】類似契約における出荷パターンデータを示す説明図である。
【図５】予測対象契約の正規化契約残量パターンデータを示す図である。
【図６】類似契約の出荷パターンから導出された正規化契残量パターンデータを示す図である。
【図７】予測対象事例の出荷タイミングに最も類似した類似契約の出荷タイミングと、その翌月の出荷量データを示す説明図である。
【図８】予測対象情報と類似契約の契約情報と出荷パターンデータを示す図である。
【図９】類似契約の類似度指標と線形結合係数を示す図である。
【図１０】１年間に渡る月毎の総出荷量の予測値と実績値を示す図である。
【符号の説明】
１０１：実績データ入力部
１０２：実績データ蓄積部
１０３：予測対象情報部
１０４：契約類似度計算部
１０５：類似契約実績抽出部
１０６：最類似出荷タイミング出荷量抽出部
１０７：出荷量演算部
１０８：予測出荷量表示部[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to an apparatus, a method, and a computer-readable storage medium for predicting the shipment amount of a product, and particularly relates to the shipment of a product such as a steel product in which the shipment of a product may be performed more than once for one order contract. A preferred technique for use in predicting quantities from contract information.
[0002]
[Prior art]
Product shipment volume prediction is an important index for drafting production plans, preventing shortage or excess of product inventory, predicting delivery times for customer orders, and predicting profits several months later. Usually, this shipment forecast is empirically estimated by production managers and sales staff, taking into account the demand trends of customers who purchase their products, as well as the amount of orders received and the time required for production. Had gone.
Further, according to the method disclosed in Patent Document 1, a shipping amount estimating apparatus using a multi-layer neural network is learned using past product shipping history and actual data of dates, and the product shipping amount is estimated. By inputting the shipment volume history and date information immediately before the prediction target period to the device and calculating the predicted shipment volume estimation value, it is possible to make predictions based on past results.
[0003]
[Patent Document 1]
JP-A-6-149849
[Problems to be solved by the invention]
However, in the conventional method that relies on human experience to predict the shipment volume of each product, the prediction accuracy varies due to individual differences among persons in charge, and in addition, when there are many products and customers, the Since it is difficult to consider trends in detail, it is necessary to make predictions based on intuition, and there has been a problem of poor accuracy.
[0005]
Further, according to the method disclosed in Patent Literature 1, using a past product shipping history and actual data of calendar information, a shipping amount estimating device based on a multi-layer neural network is learned and prediction is performed. ing. However, in the actual shipment forecast target products, as the number of data sufficient for learning of the multilayer neural network is obtained, not only products with stable orders but also long periods during which no shipment is included in the actual data For such products, there is a problem that it is difficult to learn the shipping amount estimation device because the data amount is insufficient. In addition, many of the factors that are factors in determining shipment volume forecasts are non-numerical information generally represented by codes, such as product types, consumers, production factories, distribution routes, and transportation means. It is necessary to make predictions in consideration of numerical information such as shipment volume, but since the multilayer neural network is an estimation device that inputs numerical information and outputs numerical information, information on factors expressed as non-numeric values Cannot be used effectively for prediction, and a sufficiently high accuracy cannot always be obtained.
[0006]
The present invention has been made in view of the above points, and based on past shipping results, the amount of product shipping such that product shipping may be performed multiple times for one order contract. The aim is to be predictable. In particular, even for products for which there have been few cases of similar shipping results in the past, it is possible to search for similar shipping cases by effectively utilizing non-numerical information such as consumers and distribution channels, and to make appropriate predictions based on these similar cases. The purpose is to be.
[0007]
[Means for Solving the Problems]
The product shipment quantity prediction device of the present invention uses past shipment result information of a product in which product shipment may be performed a plurality of times for one order contract, and contract information for the past shipment result information. A "product shipment quantity prediction device" for predicting the shipment volume from the forecast target contract in the forecast target period, and inputting past contract performance information and corresponding product shipment performance pattern information for a plurality of times. "Input means", "result data storage means" for storing the result information input from the result data input means, and "prediction target information" for inputting contract information to be predicted and shipping pattern information already generated from the contract. An input unit, a contract similarity calculating unit that calculates a similarity between the contract information to be predicted and the past contract actual information, and a contract similarity calculating unit. "Similar contract record extracting means" for selecting a plurality of past similar contract records having a predetermined similarity based on the issued similarity, and shipping pattern information already generated in the forecast target contract and shipping of the selected similar contract record. `` Most similar timing shipment amount extraction means '' for extracting a shipment timing in the similar contract result most similar to the predicted shipment timing from the result pattern, and extracting a shipment amount of the most similar shipment timing in the similar contract result, `` Shipping amount calculation means '' for performing weighting on the most similar timing shipment amount extracted for each of the plurality of similar contract cases based on the similarity to calculate the shipment amount for the prediction target information, The present invention is characterized in that a "prediction result output means" for displaying the predicted shipment quantity output from the shipment quantity calculation means is provided.
[0008]
Another feature of the device for predicting the shipment amount of a product according to the present invention is that the contract similarity calculation means includes numerical attribute information such as the shipment amount in the contract information, and non-numerical information such as the product type and the shipping destination. The similarity calculation process is performed by evaluating the norm between each item of the contract information to be predicted and the corresponding item of the past contract result information using both of the attribute information. .
Another feature of the device for predicting the shipment amount of a product according to the present invention is that, when the contract similarity calculating means performs the similarity calculating process using a plurality of items in the contract information, Is set individually and the similarity processing is performed.
Another feature of the device for predicting the shipment amount of a product according to the present invention is that, when the contract similarity calculation means performs similarity calculation processing using an item of numerical attribute information in contract information, The point is that similarity processing is performed by performing weighting determined based on the numerical value range of the information item.
Another feature of the product shipment amount prediction device of the present invention is that the most similar timing shipment amount extracting means subtracts the contract remaining amount obtained by subtracting the cumulative shipment amount before the forecast target period from the contract total shipment amount. The point is that the most similar shipment timing amount is extracted as an index.
Another feature of the device for predicting the shipment amount of a product according to the present invention is that the past contract result information includes product type information, total shipment amount information, customer information, production factory information, distribution route information, The point is that any one or a plurality of transportation means information is included.
Another feature of the device for predicting the shipment amount of products of the present invention is that it is applied to the estimation of the shipment amount of steel products, and the past shipment result information is used for contracting orders for steel products such as thin plates, thick plates, and steel bars. Is the monthly shipment amount information for
[0009]
The method for predicting the shipment amount of a product according to the present invention uses past shipment result information of a product in which shipment of a product may be performed a plurality of times for one order contract, and contract information for the past shipment result information. A "product shipment quantity forecasting method" for predicting the shipment quantity from the contract to be forecast in the forecast target period, wherein "actual data" is used to input past contract result information and corresponding product shipment result pattern information for a plurality of times. "Input step", "result data storage step" for storing the result information input from the result data input step, and "prediction target information" for inputting contract information to be predicted and shipping pattern information already generated from the contract. An "inputting step", a "contract similarity calculating step" for calculating the similarity between the contract information to be predicted and the past contract actual information, and the contract similarity calculating "Similar contract result extraction step" for selecting a plurality of past similar contract results having a predetermined similarity based on the similarity calculated in the process, and the similar contract result selected for the shipping pattern information already generated in the forecast target contract. Extracting the shipment timing in the similar contract result most similar to the predicted shipment timing from the shipment result pattern of the target, and extracting the shipment amount of the most similar shipment timing in the similar contract result in the “most similar timing shipment amount extraction step”. And a `` shipping amount calculation step '' of performing weighting on the most similar timing shipment amount extracted for each of the plurality of similar contract cases based on the similarity to calculate a shipment amount for the prediction target information. And a "prediction result output step" for displaying the predicted shipment quantity output from the shipment quantity calculation step. That.
[0010]
The computer-readable storage medium according to the present invention is characterized in that a program for causing a computer to function is recorded as each unit in the above-described device for predicting the amount of shipped products.
Another feature of the computer-readable storage medium of the present invention is that a program for causing a computer to execute the above-described method for predicting the shipment amount of a product is recorded.
[0011]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, with reference to the drawings, a description will be given of an apparatus, a method, and a computer-readable storage medium for predicting a product shipment amount according to the present invention.
FIG. 1 is a diagram showing a configuration of a device for predicting the shipment amount of a product according to the present embodiment. In the figure, reference numeral 101 denotes an actual data input unit. Past contract information and shipping pattern information corresponding to the contract are input to the shipment amount prediction device using, for example, a keyboard input device of a computer system. The contract information includes, for example, non-numerical information such as a product code, a customer code, a code of a mill for producing the product, a delivery location and a delivery condition code of the product, and a transportation means code relating to an order contract for a steel product, and a contract total. It is general that it is composed of numerical information such as shipping tonnage. Further, the shipping pattern divides a period after an order contract into unit periods such as one month, and indicates how many tons of shipping results have been generated in each unit period from the contract. FIG. 2 shows an example of contract information of a steel sheet product and shipping pattern information with a unit period of one month.
[0012]
Reference numeral 102 denotes a performance data storage unit that stores contract information and shipping pattern performance information input from the performance data input unit 101, and outputs the stored information as needed when a reference instruction is input. Is what you do. As a means for realizing the data storage unit, for example, there is a method using database software and a computer.
[0013]
Reference numeral 103 denotes a prediction target information input unit, which is contract information relating to an order contract for which shipment volume is to be predicted, such as a product code, a customer code, a mill code, a product delivery location and a delivery condition code in a steel product order contract, Enter the transportation code and the contracted total shipping tonnage. Furthermore, if shipment has already occurred from the order contract to be predicted, the already-generated shipping pattern information is input in addition to the contract information.
[0014]
Reference numeral 104 denotes a contract similarity calculation unit, which is realized by using, for example, a computer including a CPU, a memory, a hard disk storing a calculation algorithm, and the like. The contract information of the prediction target input by the prediction target information input unit 103 and the actual contract information stored in the actual data storage unit 102 are compared, and the similarity evaluation with the prediction target contract is performed for all contract results. As a method of evaluating the similarity using both non-numerical information such as a code and numerical information, for example, as shown in Expression (1), items of numerical information are geometrically evaluated such as Euclidean distance. Non-numerical information items are defined as norm 0 if the codes to be compared are the same code, norm 1 if the codes are different, and the norm of all items is added. There is a method of determining the similarity.
(Equation 1)
Figure 2004227301
Here, in order to correspond to taking the value of the norm is 0 or 1 non-numeric information item, for the numerical information, performs normalization processing of a k i and a k j, and the range of norm 0 1 I am trying to become. From equation (1), it can be seen that D ij has a small value when the similarity is high, and that D ij = 0 when contracts i and j have exactly the same conditions.
Further, in the equation (1), all items in the contract information are evaluated as having the same contribution rate to the similarity. However, when a specific item evaluates the similarity between contracts, another item is used. If it is known that the contribution ratio is higher than the item, the similarity process can be performed by setting the importance for each item as shown in Expression (2).
(Equation 2)
Figure 2004227301
Items with a large weighting factor are evaluated with a higher norm difference than other items, and consequently, contracts with similar items with a large weighting factor are evaluated as having high similarity. I do.
[0015]
Further, in the formulas (1) and (2), the numerical values of the numerical information items are distributed in a wide range, and it is found that the contribution ratio of the range information in which the numerical values exist is high in evaluating the similarity of the contract. If so, as shown in Expression (3), weighting may be performed according to the range of numerical information of each of the two contracts, and similarity processing may be performed.
[Equation 3]
Figure 2004227301
Here, w k i ′ j ′ is a symmetric matrix having a diagonal element value of 1 and n rows and n columns.
Figure 2004227301
Is the element. Here, the row and column indices of W k ij correspond to the range of values of a k as follows, for example.
(Equation 5)
Figure 2004227301
For each of a k i and a k j , a row and column index is determined according to the value, and the value of the element of W k ij corresponding to this index is used as a weight coefficient. Thus, when present in a k i and a k j same range, but is evaluated as a weight 1, if they are in different ranges, by weighting according to the equation (4), the similarity Processing is performed.
[0016]
Reference numeral 105 denotes a similar contract contract result extraction unit that extracts contract results similar to the forecast target contract using the similarity calculated by the contract similarity contract unit 104. As a method of extracting similar contracts, there is a method of sorting contract results in the order of similarity and selecting contracts with high similarity by the number of extracted contracts separately input to the apparatus. Alternatively, using a similarity limit index separately input to the apparatus, a method may be adopted in which all contract results having a higher similarity than the similarity limit index are made similar contracts.
[0017]
Reference numeral 106 denotes a most similar shipment timing shipment amount extraction unit. Based on the shipping pattern of the similar contract result obtained by the similar contract result extracting unit 105 and the shipping pattern information already generated input by the prediction target information input unit 103, the forecast target The timing most similar to the shipment timing of the contract is extracted, and further, the predicted shipment amount from the similar contract result is extracted. As a method of extracting the most similar timing, for example, for each unit period, the cumulative shipment amount before the forecast target period is subtracted from the contract total shipment amount, and the normalized contract remaining amount obtained by dividing by the contract total shipment amount is obtained. There is a method in which the normalized contract remaining amount of a contract and a similar contract are compared, and the timing closest to the normalized contract remaining amount of the forecast target contract in the normalized contract remaining amount pattern of the similar contract is set as the most similar shipping timing. This method will be described below with reference to an example.
[0018]
FIG. 3 shows an example of a shipping pattern already generated from a forecast target contract, and FIG. 4 shows an example of a shipping pattern of a similar contract example in a table format for each month. In this example, the purpose is to predict the amount of shipment in the N + February in the contract to be predicted, and three similar contract cases are extracted. In FIGS. 3 and 4, the normalized contract remaining amount is calculated for each month according to the above-described procedure, and is shown in a table format. FIGS. The normalized contract remaining amount in the (N + N) month in FIG. 5 is 50%. In the similar contract shipping pattern in FIG. 6, if the timing having the closest normalized contract remaining amount is extracted, the similarity to No1 is obtained. The most similar shipment timing is evaluated as N + February in the contract example, N + 1 month in the similar contract example of No2, and N + Feb in the similar contract example of No3. Further, using the shipping pattern from the similar contract in FIG. 4 and the most similar shipping timing information obtained in the above procedure, the estimated shipping amount corresponding to the N + February shipping amount of the contract to be predicted is obtained as shown in FIG. Can be
[0019]
Reference numeral 107 denotes a shipment amount calculation unit, which calculates a shipment amount for each similar contract obtained by the most similar shipment timing shipment amount extraction unit 106 and the similarity of the similar contract calculated by the contract similarity calculation unit 104. Calculate the estimated shipment volume. As a method of calculating the predicted shipment amount, for example, a weighting factor for each similar contract is calculated using the reciprocal of the similarity D ij calculated based on any of the formulas (1) to (3), and Is multiplied by a weighting factor to the estimated shipment amount of the above, and a linear sum is obtained.
[0020]
Reference numeral 108 denotes a shipment quantity prediction result display unit, which displays information effective in evaluating the prediction result, such as a shipment quantity prediction result of the target contract, a similar contract example and a shipping pattern, and a most similar shipment timing, which are the basis of the estimation. For example, it is output by a printing means such as a CRT in a computer system or a printer.
[0021]
According to the product shipment amount prediction apparatus according to the present embodiment described above, when searching for past contract results similar to contract information to be predicted, not only numerical information but also customer and transportation means information, etc. Since non-numerical information that has a high influence on the shipping volume estimation is also taken into account, it is possible to make predictions with high accuracy. In addition, even for contract cases that have been shipped in the past, the most similar contract can be extracted from existing contract results, and a prediction can be made based on the shipped result.
[0022]
【Example】
Hereinafter, an example in which the shipment amount of a steel sheet product is predicted by the present method will be described.
In the present embodiment, the shipment volume was estimated using information on about 150,000 contracts in five years and monthly shipping result information on each contract. Of the contract information, the first four years are stored in the actual data storage unit, and the remaining one year's shipping volume is estimated. For the predicted shipment volume, the prediction accuracy is evaluated by comparing the predicted shipment volume with the actual value.
[0023]
Items to evaluate the similarity of contracts include numerical information such as contract total shipping tonnage, non-numerical information such as product name code, user code, trading company code, sales department code, product delivery condition code, and delivery condition code. The location code, transportation mode code, and transportation code were used.
In calculating the similarity index, paying attention to the fact that the shipping pattern of sheet products is greatly affected by the total contracted shipping tonnage, the weighting factors for non-numerical information items are all set to 1 and the total The weighting factor Q for the contract shipment tonnage was variously changed, and the shipment patterns of the contract to be predicted and the extracted similar contract were compared. As a result, when Q = 3, it was evaluated that the most similar shipping patterns were extracted. Therefore, the shipping prediction is performed under this condition.
[0024]
Also, in the process of investigating the relationship between the total shipping tonnage and the shipping pattern, it became clear that shipping patterns can be generally grouped according to the total shipping tonnage range, so grouping according to the total contract tonnage range, The weighting factors between the groups were variously changed, and the shipping patterns of the contract to be predicted and the extracted similar contract were compared. As a result, the total contracted shipping tonnage is as follows: Group 1: Total contracted shipping tonnage 0 to 10 tons, Group 2: Total contracted shipping tonnage 10 to 100 tons, Group 3: Total contracted shipping tonnage 100 to 1000 tons, Group 4: When the Wij matrix in Equation (4) is classified as four with a total contract shipping tonnage of 1000 tons or more and is set as follows, it is evaluated that the shipping pattern with the highest similarity is extracted. Therefore, predictions are made under these conditions.
(Equation 6)
Figure 2004227301
Here, indices 1 to 4 of the matrix correspond to 1 to 4 of the above group.
[0025]
Next, regarding the extraction of similar contract cases, it is assumed that P cases are extracted from the contract results with the highest similarity, the value of P is variously changed, the estimated shipment volume is calculated, and compared with the shipment result. The condition that minimizes the error was searched for in the accuracy evaluation performed. As a result, when P = 3, the error was the smallest, and hence the prediction is performed under this condition.
[0026]
FIG. 8 shows information on a contract to be predicted, information on three similar contracts extracted by the present method, and respective shipping patterns. Each similar contract also describes the degree of similarity with the forecast target. From the contract to be predicted, shipments up to N + 1 month have already occurred, and the purpose is to predict the shipment amount for N + Feb. As a result of extracting the most similar shipment timing in the similar contract by this method, N + Feb was extracted in any case of the similar contract. FIG. 9 shows a reciprocal obtained from the similarity index of the similar contract and further normalized such that the sum of the linear sum coefficients becomes 1. This linear sum coefficient multiplied by the estimated shipment volume adds three similar contracts to 397.5 tons. Since the shipment result of the forecast target case in N + February was 412 tons, the prediction error regarding the contract case was relatively good at 3.5%.
[0027]
FIG. 10 is a graph in which the shipment amount for a specific steel type is predicted from all contracts every month and plotted together with the actual shipment amount. The average prediction error by this method is about 2.5%, and the shipment amount can be predicted with higher accuracy and stability than, for example, the conventional manual prediction.
By utilizing this result, it was possible to obtain an effect that a distribution plan of production amounts to each steelworks and a prediction of a change in product inventory can be performed with higher accuracy than before.
[0028]
In the present embodiment, the shipment amount prediction device is realized as a program on a computer, but may be configured by hardware combining an arithmetic device, a memory, and the like.
Further, the device for predicting the shipment amount of a product according to the present invention may be constituted by a plurality of devices or by one device.
Further, the above-described embodiment is configured by a computer CPU, MPU, RAM, ROM, or the like, and is implemented by operating a program recorded in the RAM or ROM. Therefore, means for supplying a program code of software for realizing the functions of the above-described embodiments to a computer, for example, a storage medium storing such a program code is included in the scope of the present invention.
[0029]
【The invention's effect】
As described above, according to the present invention, a similar contract example is extracted by utilizing both the numerical information and the non-numeric information in the contract information. By estimating the quantity, it is possible to predict the shipment quantity with higher accuracy.
In addition, even if it is difficult to obtain a shipping forecast model by learning because there are few similar cases in the past, since similar cases are searched using non-numerical information, appropriate Predictions are possible.
[Brief description of the drawings]
FIG. 1 is a diagram showing a configuration of a device for predicting the shipment amount of a product according to an embodiment of the present invention.
FIG. 2 is a diagram showing contract performance data and shipping pattern data of a product.
FIG. 3 is a diagram showing pattern data of a shipment already generated from a contract to be predicted.
FIG. 4 is an explanatory diagram showing shipping pattern data in a similar contract.
FIG. 5 is a diagram showing normalized contract remaining amount pattern data of a contract to be predicted.
FIG. 6 is a diagram showing normalized contract remaining amount pattern data derived from a similar contract shipping pattern.
FIG. 7 is an explanatory diagram showing the shipment timing of a similar contract most similar to the shipment timing of the prediction target case, and the shipment amount data for the following month.
FIG. 8 is a diagram showing prediction target information, contract information of a similar contract, and shipping pattern data.
FIG. 9 is a diagram showing a similarity index and a linear combination coefficient of similar contracts.
FIG. 10 is a diagram showing a predicted value and an actual value of the total shipment amount every month over one year.
[Explanation of symbols]
101: performance data input unit 102: performance data storage unit 103: prediction target information unit 104: contract similarity calculation unit 105: similar contract performance extraction unit 106: most similar shipment timing shipment amount extraction unit 107: shipment amount calculation unit 108: Estimated shipment volume display

Claims (10)

一つの受注契約に対して製品の出荷が複数回にわたる場合がある製品の過去の出荷実績情報、及び前記過去の出荷実績情報に対する契約情報を使用して、予測対象期間における予測対象契約からの出荷量を予測する製品の出荷量予測装置であって、
過去の契約実績情報とそれに対応する複数回にわたる製品出荷実績パターン情報を入力する「実績データ入力手段」と、
前記実績データ入力手段から入力された実績情報を蓄積する「実績データ蓄積手段」と、
予測対象となる契約情報と当該契約から既に発生した出荷パターン情報を入力する「予測対象情報入力手段」と、
前記予測対象となる契約情報と前記過去の契約実績情報の類似度を計算する「契約類似度計算手段」と、
前記契約類似度計算装置で算出した類似度に基づき所定の類似度を有する複数の過去の類似契約実績を選び出す「類似契約実績抽出手段」と、
前記予測対象契約における既に発生した出荷パターン情報と、選び出した前記類似契約実績の出荷実績パターンから、予測対象の出荷タイミングに最も類似した前記類似契約実績における出荷タイミングを抽出し、前記類似契約実績における最類似出荷タイミングの出荷量を抽出する「最類似タイミング出荷量抽出手段」と、
前記複数の類似契約事例それぞれに対して抽出された最類似タイミング出荷量に、前記類似度に基づいて重み付けを行って、前記予測対象情報に対する出荷量を演算する「出荷量演算手段」と、
前記出荷量演算手段から出力された予測出荷量を表示する「予測結果出力手段」を備えたことを特徴とする製品の出荷量予測装置。Using the past shipment result information of a product that may be shipped multiple times for one order contract, and the contract information for the past shipment result information, shipping from the forecast target contract in the forecast target period A device for predicting the quantity of products to be shipped,
"Result data input means" for inputting past contract result information and corresponding product shipment result pattern information for multiple times,
`` Result data storage means '' for storing the result information input from the result data input means,
"Prediction target information input means" for inputting contract information to be predicted and shipping pattern information already generated from the contract;
`` Contract similarity calculation means '' for calculating the similarity between the contract information to be predicted and the past contract performance information,
`` Similar contract result extraction means '' for selecting a plurality of past similar contract results having a predetermined similarity based on the similarity calculated by the contract similarity calculating device,
From the shipping pattern information that has already occurred in the forecast target contract, and the selected shipping result pattern of the similar contract result, the shipping timing in the similar contract result most similar to the shipping timing of the prediction target is extracted, and in the similar contract result, "Similar timing shipment quantity extraction means" for extracting the shipment quantity of the most similar shipment timing,
`` Shipping amount calculating means '' for performing weighting on the most similar timing shipment amount extracted for each of the plurality of similar contract cases based on the similarity and calculating the shipment amount for the prediction target information,
A device for predicting the shipment amount of a product, comprising: a "prediction result output unit" for displaying the estimated shipment amount output from the shipment amount calculation unit. 前記契約類似度計算手段が、契約情報における出荷量のような数値属性情報と、製品の種類や出荷先のような非数値属性情報の両者を用いて、前記予測対象となる契約情報の各項目と対応する前記過去の契約実績情報の各項目とのノルムを評価することにより、類似度計算処理するようにしたことを特徴とする請求項１に記載の製品の出荷量予測装置。The contract similarity calculation means uses both numerical attribute information such as the shipment amount in the contract information and non-numerical attribute information such as the product type and the shipping destination, and calculates each item of the contract information to be predicted. 2. The product shipping amount prediction apparatus according to claim 1, wherein the similarity calculation processing is performed by evaluating a norm of each item of the past contract result information corresponding to the item. 前記契約類似度計算手段が、契約情報における複数の項目を用いて類似度計算処理する場合に、前記複数の項目それぞれに対する重要度を個別に設定して類似度処理するようにしたことを特徴とする請求項１又は２に記載の製品の出荷量予測装置。When the contract similarity calculation means performs similarity calculation processing using a plurality of items in contract information, the similarity processing is performed by individually setting the importance for each of the plurality of items. The device for predicting the shipment amount of a product according to claim 1. 前記契約類似度計算手段が、契約情報における数値属性情報の項目を用いて類似度計算処理する場合に、前記数値属性情報項目の数値の範囲に基づいて決定される重み付けを行って類似度処理するようにしたことを特徴とする請求項１〜３のいずれか１項に記載の製品の出荷量予測装置。When the contract similarity calculation means performs similarity calculation processing using numerical attribute information items in contract information, the similarity calculation processing is performed by performing weighting determined based on the numerical value range of the numerical attribute information items. The apparatus for predicting the shipment amount of a product according to any one of claims 1 to 3, wherein the apparatus is configured as described above. 前記最類似タイミング出荷量抽出手段が、契約総出荷量から予測対象期間前の累積出荷量を差し引いたものを、更に契約総出荷量で除して得られる正規化契約残量を指標として最類似出荷タイミング量を抽出するようにしたことを特徴とする請求項１〜４のいずれか１項に記載の製品の出荷量予測装置。The most similar timing shipment amount extraction means extracts the contracted total shipment amount minus the cumulative shipment amount before the prediction target period, and further divides the result by the contracted total shipment amount to obtain the most similar contract remaining amount as an index. 5. The apparatus according to claim 1, wherein a shipment timing amount is extracted. 前記過去の契約実績情報に、製品の種類情報、総出荷量情報、需要家情報、生産工場情報、流通経路情報、輸送手段情報のいずれか一つ、或いは複数を含むことを特徴とする請求項１〜５のいずれか１項に記載の製品の出荷量予測装置。The past contract record information includes one or more of product type information, total shipment amount information, customer information, production factory information, distribution route information, and transportation means information. The device for predicting the shipment amount of a product according to any one of claims 1 to 5. 鉄鋼製品の出荷量予測に適用され、前記過去の出荷実績情報は、薄板、厚板、条鋼などの鉄鋼製品受注契約に対する月別の出荷量情報であることを特徴とする請求項１〜６のいずれか１項に記載の製品の出荷量予測装置。7. The method according to claim 1, wherein the past shipment result information is applied to a forecast of a shipment amount of a steel product, and the past shipment result information is a monthly shipment amount information for an order contract for a steel product such as a thin plate, a thick plate, and a bar. 2. The device for predicting the shipment amount of a product according to claim 1. 一つの受注契約に対して製品の出荷が複数回にわたる場合がある製品の過去の出荷実績情報、及び前記過去の出荷実績情報に対する契約情報を使用して、予測対象期間における予測対象契約からの出荷量を予測する製品の出荷量予測方法であって、
過去の契約実績情報とそれに対応する複数回にわたる製品出荷実績パターン情報を入力する「実績データ入力工程」と、
前記実績データ入力工程から入力された実績情報を蓄積する「実績データ蓄積工程」と、
予測対象となる契約情報と当該契約から既に発生した出荷パターン情報を入力する「予測対象情報入力工程」と、
前記予測対象となる契約情報と前記過去の契約実績情報の類似度を計算する「契約類似度計算工程」と、
前記契約類似度計算工程で算出した類似度に基づき所定の類似度を有する複数の過去の類似契約実績を選び出す「類似契約実績抽出工程」と、
前記予測対象契約における既に発生した出荷パターン情報と選び出した前記類似契約実績の出荷実績パターンから、予測対象の出荷タイミングに最も類似した前記類似契約実績における出荷タイミングを抽出し、前記類似契約実績における最類似出荷タイミングの出荷量を抽出する「最類似タイミング出荷量抽出工程」と、
前記複数の類似契約事例それぞれに対して抽出された最類似タイミング出荷量に、前記類似度に基づいて重み付けを行って、前記予測対象情報に対する出荷量を演算する「出荷量演算工程」と、
前記出荷量演算工程から出力された予測出荷量を表示する「予測結果出力工程」を備えたことを特徴とする製品の出荷量予測方法。Using the past shipment result information of a product that may be shipped multiple times for one order contract, and the contract information for the past shipment result information, shipping from the forecast target contract in the forecast target period A method for estimating the shipment amount of a product for estimating an amount,
"Result data input process" for inputting past contract result information and corresponding product shipment result pattern information for multiple times,
`` Result data accumulation step '' for accumulating the result information input from the result data input step,
A "prediction target information input step" for inputting contract information to be predicted and shipping pattern information already generated from the contract;
`` A contract similarity calculation step '' for calculating the similarity between the contract information to be predicted and the past contract performance information,
A `` similar contract result extraction step '' for selecting a plurality of past similar contract results having a predetermined similarity based on the similarity calculated in the contract similarity calculation step;
From the shipping pattern information that has already occurred in the contract to be predicted and the shipping result pattern of the selected similar contract result, the shipment timing in the similar contract result that is most similar to the shipping timing to be predicted is extracted, and the latest in the similar contract result is extracted. A “most similar timing shipment quantity extraction process” for extracting the shipment quantity of similar shipment timing,
A `` shipping amount calculation step '' for performing weighting on the most similar timing shipment amount extracted for each of the plurality of similar contract cases based on the similarity and calculating a shipment amount for the prediction target information;
A method for predicting the shipment amount of a product, comprising a "prediction result output step" for displaying the estimated shipment amount output from the shipment amount calculation step. 上記各手段として、コンピュータを機能させるプログラムを記録したことを特徴とするコンピュータ読み取り可能な記憶媒体。A computer-readable storage medium, wherein a program for causing a computer to function is recorded as each of the above means. 製品の出荷量予測方法をコンピュータに実行させるためのプログラムを記録したことを特徴とするコンピュータ読み取り可能な記憶媒体。A computer-readable storage medium storing a program for causing a computer to execute a method for predicting a shipment amount of a product.
JP2003014532A 2003-01-23 2003-01-23 Product shipment amount prediction apparatus, method, and computer-readable storage medium Expired - Fee Related JP4268408B2 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
JP2003014532A JP4268408B2 (en) 2003-01-23 2003-01-23 Product shipment amount prediction apparatus, method, and computer-readable storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
JP2003014532A JP4268408B2 (en) 2003-01-23 2003-01-23 Product shipment amount prediction apparatus, method, and computer-readable storage medium

Publications (2)

Publication Number Publication Date
JP2004227301A true JP2004227301A (en) 2004-08-12
JP4268408B2 JP4268408B2 (en) 2009-05-27

Family

ID=32902556

Family Applications (1)

Application Number Title Priority Date Filing Date
JP2003014532A Expired - Fee Related JP4268408B2 (en) 2003-01-23 2003-01-23 Product shipment amount prediction apparatus, method, and computer-readable storage medium

Country Status (1)

Country Link
JP (1) JP4268408B2 (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2006344186A (en) * 2004-06-14 2006-12-21 Univ Of Electro-Communications Flow control system, method, and program
JP2007026162A (en) * 2005-07-19 2007-02-01 Internatl Business Mach Corp <Ibm> Material allocating method, apparatus, and program
EP2053551A1 (en) 2007-10-26 2009-04-29 Hitachi Ltd. Execution decision support program and device of regulation measures
JP2009294952A (en) * 2008-06-05 2009-12-17 Sekisui Chem Co Ltd Production planning system and method
CN116680162A (en) * 2023-04-17 2023-09-01 中国软件评测中心(工业和信息化部软件与集成电路促进中心) Test case multiplexing method, device, medium, equipment and product

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH0668065A (en) * 1991-09-13 1994-03-11 Kanebo Ltd Demand forecasting device for commodity
JPH11203361A (en) * 1998-01-14 1999-07-30 Nec Corp System for easily calculating production schedule
JP2001106310A (en) * 1999-10-08 2001-04-17 Kirin Engineering Co Ltd Physical distribution system
JP2002073752A (en) * 2000-09-04 2002-03-12 Kao Corp Order receipt forecast system, order receipt control system, and production control system

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH0668065A (en) * 1991-09-13 1994-03-11 Kanebo Ltd Demand forecasting device for commodity
JPH11203361A (en) * 1998-01-14 1999-07-30 Nec Corp System for easily calculating production schedule
JP2001106310A (en) * 1999-10-08 2001-04-17 Kirin Engineering Co Ltd Physical distribution system
JP2002073752A (en) * 2000-09-04 2002-03-12 Kao Corp Order receipt forecast system, order receipt control system, and production control system

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2006344186A (en) * 2004-06-14 2006-12-21 Univ Of Electro-Communications Flow control system, method, and program
JP4706018B2 (en) * 2004-06-14 2011-06-22 国立大学法人電気通信大学 Flow number management system, method, and program
JP2007026162A (en) * 2005-07-19 2007-02-01 Internatl Business Mach Corp <Ibm> Material allocating method, apparatus, and program
JP4636609B2 (en) * 2005-07-19 2011-02-23 インターナショナル・ビジネス・マシーンズ・コーポレーション Material allocation method, apparatus and program
EP2053551A1 (en) 2007-10-26 2009-04-29 Hitachi Ltd. Execution decision support program and device of regulation measures
JP2009110044A (en) * 2007-10-26 2009-05-21 Hitachi Ltd Implementation determination support program and device for regulation countermeasure
US8121878B2 (en) 2007-10-26 2012-02-21 Hitachi, Ltd. Execution decision support program and device of regulation measures
JP2009294952A (en) * 2008-06-05 2009-12-17 Sekisui Chem Co Ltd Production planning system and method
CN116680162A (en) * 2023-04-17 2023-09-01 中国软件评测中心(工业和信息化部软件与集成电路促进中心) Test case multiplexing method, device, medium, equipment and product
CN116680162B (en) * 2023-04-17 2023-11-10 中国软件评测中心(工业和信息化部软件与集成电路促进中心) Test case multiplexing method, device, medium, equipment and product

Also Published As

Publication number Publication date
JP4268408B2 (en) 2009-05-27

Similar Documents

Publication Publication Date Title
AU2002353396B2 (en) Sales optimization
Yang et al. Integrated multi-period dynamic inventory classification and control
JP4678879B2 (en) Sales prediction system, method and computer program
JP6514342B2 (en) Online forecasting system and method
JP6370757B2 (en) Profit / loss prediction apparatus and profit / loss prediction program
JPH0668065A (en) Demand forecasting device for commodity
van Donselaar et al. Heuristics for setting reorder levels in periodic review inventory systems with an aggregate service constraint
JP2012247964A (en) Progress management apparatus and progress management program
Barron et al. Shortage decision policies for a fluid production model with MAP arrivals
JP5031715B2 (en) Product demand forecasting system, product sales volume adjustment system
CN115860800A (en) Festival and holiday commodity sales volume prediction method and device and computer storage medium
CN111768243A (en) Sales prediction method, prediction model construction method, device, equipment and medium
Ivanov et al. Demand forecasting
JP4296026B2 (en) Product demand forecasting system, product sales volume adjustment system
JP4268408B2 (en) Product shipment amount prediction apparatus, method, and computer-readable storage medium
JP5847137B2 (en) Demand prediction apparatus and program
US11238482B1 (en) Method and system for managing clearance items
JP2007115169A (en) Production planning device and method, and program
JP2007122264A (en) Prediction system for management or demand and prediction program to be used for the same
JPH11259564A (en) Sales prediction supporting system
JP2002007671A (en) Device and method for predicting demand, computer program and computer readable recording medium
JP2007026335A (en) Evaluation index forecast visualization method
JP6242362B2 (en) Profit / loss prediction apparatus and profit / loss prediction program
JP6704464B2 (en) A system for estimating accident factors in the operations of distribution warehouses
JPH0652191A (en) Device and method for analyzing commodity data

Legal Events

Date Code Title Description
A621 Written request for application examination

Free format text: JAPANESE INTERMEDIATE CODE: A621

Effective date: 20050914

A131 Notification of reasons for refusal

Free format text: JAPANESE INTERMEDIATE CODE: A131

Effective date: 20080729

A521 Written amendment

Free format text: JAPANESE INTERMEDIATE CODE: A523

Effective date: 20080924

A02 Decision of refusal

Free format text: JAPANESE INTERMEDIATE CODE: A02

Effective date: 20081111

A521 Written amendment

Free format text: JAPANESE INTERMEDIATE CODE: A523

Effective date: 20081208

A911 Transfer of reconsideration by examiner before appeal (zenchi)

Free format text: JAPANESE INTERMEDIATE CODE: A911

Effective date: 20090130

TRDD Decision of grant or rejection written
A01 Written decision to grant a patent or to grant a registration (utility model)

Free format text: JAPANESE INTERMEDIATE CODE: A01

Effective date: 20090217

A01 Written decision to grant a patent or to grant a registration (utility model)

Free format text: JAPANESE INTERMEDIATE CODE: A01

A61 First payment of annual fees (during grant procedure)

Free format text: JAPANESE INTERMEDIATE CODE: A61

Effective date: 20090220

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20120227

Year of fee payment: 3

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20120227

Year of fee payment: 3

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20130227

Year of fee payment: 4

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20130227

Year of fee payment: 4

S531 Written request for registration of change of domicile

Free format text: JAPANESE INTERMEDIATE CODE: R313531

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20130227

Year of fee payment: 4

R350 Written notification of registration of transfer

Free format text: JAPANESE INTERMEDIATE CODE: R350

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20130227

Year of fee payment: 4

S533 Written request for registration of change of name

Free format text: JAPANESE INTERMEDIATE CODE: R313533

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20130227

Year of fee payment: 4

R350 Written notification of registration of transfer

Free format text: JAPANESE INTERMEDIATE CODE: R350

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20140227

Year of fee payment: 5

LAPS Cancellation because of no payment of annual fees