TWI795809B - Business evaluation system and method therefore - Google Patents
Business evaluation system and method therefore Download PDFInfo
- Publication number
- TWI795809B TWI795809B TW110122201A TW110122201A TWI795809B TW I795809 B TWI795809 B TW I795809B TW 110122201 A TW110122201 A TW 110122201A TW 110122201 A TW110122201 A TW 110122201A TW I795809 B TWI795809 B TW I795809B
- Authority
- TW
- Taiwan
- Prior art keywords
- business
- change
- rate
- evaluation
- machine learning
- Prior art date
Links
Images
Abstract
Description
本案內容係關於一種業務評估系統,特別是關於一種具備機器學習之業務評估系統。The content of this case is about a business evaluation system, especially a business evaluation system with machine learning.
在金融機構推新型業務或服務時,若能準確的評估推出業務類別及時間點,將會促金融機構的營運管理並增加業務。如何提供最佳的業務類別及時間點係本領域重要的議題。When financial institutions launch new businesses or services, if they can accurately assess the type of business and the timing of the launch, it will promote the operation and management of financial institutions and increase business. How to provide the best service category and time point is an important issue in this field.
本揭示文件提供一種業務評估系統。業務評估系統包含伺服器、輸入介面以及業務評估主機。伺服器儲存業務評估項目的複數個評估因素,其中該些評估因素分別對應複數個權重。輸入介面通訊連接於該伺服器,輸入介面用以接收複數筆營運資料,其中該些筆營運資料分別對應該些評估因素。業務評估主機通訊連接於伺服器,業務評估主機儲存有機器學習模型,機器學習模型依據對應於該些評估因素的該些筆營運資料以及對應於該些評估因素的該些權重計算該業務評估項目的得分,並且該類機器學習模型更用以依據與該些評估因素其中一者對應之該些筆營運資料中之一筆營運資料以及一筆歷史營運資料計算第一變動率並且依據該得分以及歷史得分計算第二變動率,該機器學習模型依據該第一變動率與該第二變動率之差值調整該些權重。This disclosure provides a business evaluation system. The business evaluation system includes a server, an input interface and a business evaluation host. The server stores a plurality of evaluation factors of the service evaluation item, wherein the evaluation factors respectively correspond to a plurality of weights. The input interface is communicatively connected to the server, and the input interface is used to receive multiple pieces of operation data, wherein the pieces of operation data respectively correspond to the evaluation factors. The business evaluation host is connected to the server by communication, and the business evaluation host stores a machine learning model, and the machine learning model calculates the business evaluation item based on the operation data corresponding to the evaluation factors and the weights corresponding to the evaluation factors , and this type of machine learning model is further used to calculate the first rate of change based on one of the operating data corresponding to one of the evaluation factors and a piece of historical operating data, and based on the score and the historical score A second rate of change is calculated, and the machine learning model adjusts the weights according to a difference between the first rate of change and the second rate of change.
本揭示文件提供一種業務評估方法。業務評估方法包含下列步驟。藉由輸入介面接收複數筆營運資料,其中該些筆營運資料分別對應業務評估項目的複數個評估因素。藉由業務評估主機自伺服器取得與該些評估因素分別對應的複數個權重。藉由業務評估主機執行機器學習模型。藉由機器學習模型依據對應於該些評估因素的該些筆營運資料以及對應於該些評估因素的該些權重計算該業務評估項目的一得分。藉由機器學習模型依據與該些評估因素其中一者對應之該些筆營運資料中之一筆營運資料以及一筆歷史營運資料計算第一變動率並且依據該得分以及歷史得分計算第二變動率。機器學習模型依據第一變動率與第二變動率之差值調整該些權重。This disclosure provides a business valuation methodology. The business valuation methodology consists of the following steps. A plurality of pieces of operation data are received through the input interface, wherein the pieces of operation data respectively correspond to a plurality of evaluation factors of the business evaluation item. A plurality of weights respectively corresponding to the evaluation factors are obtained from the server by the business evaluation host. The machine learning model is executed by the business evaluation host. A score of the business evaluation item is calculated by a machine learning model according to the pieces of operation data corresponding to the evaluation factors and the weights corresponding to the evaluation factors. A machine learning model is used to calculate the first rate of change based on one piece of operating data and a piece of historical operating data corresponding to one of the evaluation factors, and calculate the second rate of change based on the score and the historical score. The machine learning model adjusts the weights according to the difference between the first rate of change and the second rate of change.
綜上所述,本揭示文件的業務評估系統可算出可信的業務評估項目的得分,藉以作為業務管理作為銀行分支機構未來獲利策略及佈局之方向。To sum up, the business evaluation system in this disclosure document can calculate the scores of credible business evaluation items, which can be used as business management as the direction of future profit strategy and layout of bank branches.
下文係舉實施例配合所附圖式作詳細說明,以更好地理解本案的態樣,但所提供之實施例並非用以限制本案所涵蓋的範圍,而結構操作之描述非用以限制其執行之順序,任何由元件重新組合之結構,所產生具有均等功效的裝置,皆為本案所涵蓋的範圍。The following is a detailed description of the embodiment in conjunction with the accompanying drawings to better understand the aspect of the case, but the provided embodiment is not used to limit the scope of the case, and the description of the structure and operation is not used to limit it. The sequence of execution, any structure recombined by components to produce devices with equivalent functions are all within the scope of this case.
此外,在本文中所使用的用詞『包含』、『包括』、『具有』、『含有』等等,均為開放性的用語,即意指『包含但不限於』。此外,本文中所使用之『及/或』,包含相關列舉項目中一或多個項目的任意一個以及其所有組合。In addition, the words "comprising", "including", "having", "containing", etc. used in this article are all open terms, meaning "including but not limited to". In addition, "and/or" used herein includes any one and all combinations of one or more items in the relevant listed items.
請參閱第1圖,第1圖為依據本揭露一實施例之業務評估系統100。如第1圖所示,業務評估系統100包含業務評估主機110輸入介面120、伺服器130。伺服器130儲存多個業務項目。業務項目可以依據欲評估之金融機構的類型而有所調整。舉例而言,業務項目以企今授信為例,企金授信的分析評估可包含多個分析類別,例如,企金授信的分析類別有整體業務執行情形、業務別客戶融資情況以及營運管理情形。每一個分析類別皆可包含多個評估因素,例如,整體業務執行情形包含五項評估因素,分別是平均總融資餘額、平均總額度、平均總戶數、平均總收益、總收益於全行佔比,但不以此為限。不同評估因素具有相應的權重。同一分析類別所有評估因素的權重總和為所述分析類別的權重。相異分析類別可以具有相異的權重。所有分析類別的權重總和可以為100%。上述的權重可以儲存於伺服器。Please refer to FIG. 1 . FIG. 1 is a
輸入介面120透過有線或無線網路與業務評估主機110通訊連接,電子裝置S、輸入介面120可以與業務評估主機110通訊連接。電子裝置S可以由個人電腦、行動裝通訊裝置或平板電腦實現。用戶可透過電子裝置S、輸入介面120將關聯於業務項目的營運資料輸入至業務評估主機110。The
業務評估主機110包含處理器112、內部記憶體114及外部記憶體116。內部記憶體114用以儲存機器學習模型114a,外部記憶體116用以儲存營運資料。機器學習模型114a包含類神經網路、身心語言程式學、模糊邏輯模型、隱馬爾可夫模型、決策樹、貝氏演算法、條件隨機域或支持向量機。The
處理器112電性耦接內部記憶體114以及外部記憶體116。處理器112透過有線或無線網路與輸入介面120以及伺服器130通訊連接。The
請一併參閱第2圖,第2圖為依據本揭示一實施利之業務評估方法S100的流程圖。業務評估方法S100包含步驟S110~S130。在步驟S110中,藉由輸入介面接收複數筆營運資料,該些營運資料分別對應複數個評估因素。在步驟S120中,藉由業務評估主機取得與該些評估因素分別對應的複數個權重。在步驟S130中,藉由業務評估主機執行機器學習模型以調整該些權重,並依據調整後的該些權重計算業務評估項目的調整得分。Please also refer to FIG. 2 . FIG. 2 is a flowchart of a business evaluation method S100 implemented according to the present disclosure. The business evaluation method S100 includes steps S110-S130. In step S110, a plurality of pieces of operation data are received through the input interface, and the operation data respectively correspond to a plurality of evaluation factors. In step S120, a plurality of weights respectively corresponding to the evaluation factors are obtained by the service evaluation host. In step S130, the business evaluation host executes the machine learning model to adjust the weights, and calculates the adjusted score of the business evaluation item according to the adjusted weights.
為了更佳的理解本揭示文件,請一併參閱下列的表一。在步驟S110中,藉由輸入介面120接收複數筆營運資料D1~D5,該些營運資料D1~D5分別對應複數個評估因素F11~F15。在一些實施例中,輸入介面120亦接收複數筆營運資料D6~D10,該些營運資料D6~D10也分別對應複數個評估因素F11~F15。在另一些實施例中,營運資料D6~D10已儲存於外部記憶體116,處理器112可以自外部記憶體116提取營運資料D6~D10。並且,在步驟S120中,藉由業務評估主機110取得與該些評估因素F11~F15分別對應的複數個權重。
如表一所示, 整體業務執行情形所涵蓋的評估因素F11~ F15分別是平均總融資餘額、平均總額度、平均總戶數、平均總收益以及總收益於全行比重。As shown in Table 1, the evaluation factors F11~F15 covered by the overall business execution situation are the average total financing balance, the average total amount, the average total number of accounts, the average total income, and the proportion of total income to the whole bank.
在一些實施例中,平均總融資餘額(評估因素F11)代表的是欲評估月份(例如,當月)總融資餘額佔過去五年平均總融資餘額之比例。平均總額度(評估因素F12)代表的是欲評估月份(例如,當月) 總額度佔過去五年平均總額度之比例。平均總戶數(評估因素F13)代表的是欲評估月份(例如,當月)總戶數佔過去五年平均總戶數之比例。平均總收益(評估因素F14)代表的是欲評估月份(例如,當月) 總收益佔過去五年平均總收益之比例。總收益於全行比重(評估因素F15)代表的是欲評估月份(例如,當月) 總收益於全行比重佔過去五年平均總收益之比例。In some embodiments, the average total financing balance (assessment factor F11 ) represents the ratio of the total financing balance of the month to be evaluated (for example, the current month) to the average total financing balance of the past five years. The average total amount (assessment factor F12) represents the ratio of the total amount of the month to be evaluated (for example, the current month) to the average total amount of the past five years. The average total number of households (assessment factor F13) represents the ratio of the total number of households in the month to be evaluated (for example, the current month) to the average total number of households in the past five years. The average total income (assessment factor F14) represents the ratio of the total income of the month to be evaluated (for example, the current month) to the average total income of the past five years. The proportion of total income in the whole bank (assessment factor F15) represents the proportion of the total income in the month to be evaluated (for example, the current month) to the average total income of the past five years.
需注意的是,營運資料D1~D5是欲評估月份(例如,當月)的資料,營運資料D6~D10可以是與營運資料D1~D5在相同月份的多年度資料的數值平均。在一些實施例中,營運資料D6~D10可以是與營運資料D1~D5在相同月份的近五年內資料的數值平均。舉例而言,營運資料D1~D5是108年1月的營運資料,營運資料D6~D10是104年1月、105年1月、106年1月、107年1月、108年1月的營運資料的數值平均。It should be noted that the operation data D1~D5 are the data of the month to be evaluated (for example, the current month), and the operation data D6~D10 can be the numerical average of multi-year data in the same month as the operation data D1~D5. In some embodiments, the operating data D6-D10 may be the numerical average of data in the same month as the operating data D1-D5 within the past five years. For example, the operation data D1~D5 are the operation data in January 2010, and the operation data D6~D10 are the operation data in January 104, January 105, January 106, January 107, and January 108 The numerical average of the data.
在一些實施例中,評估因素F11~ F15各自具有4%的權重,業務評估項目三大類別之一的整體業務執行情形合計了評估因素F11~ F15各自的權重,整體業務執行情形會具有20%的權重。In some embodiments, the evaluation factors F11-F15 each have a weight of 4%, and the overall business execution situation of one of the three major categories of business evaluation items adds up the respective weights of the evaluation factors F11-F15, and the overall business execution situation will have a weight of 20%. the weight of.
在步驟S120中,藉由業務評估主機110取得與評估因素F11~ F15分別對應的多個權重。舉例而言,業務評估主機110自伺服器130取得與評估因素F11~ F15分別對應的多個權重,例如,評估因素F11~ F15的多個權重皆為4%,如表一所示。In step S120 , a plurality of weights respectively corresponding to the evaluation factors F11 - F15 are obtained by the
在步驟S130中,藉由業務評估主機110執行機器學習模型114a以調整所述權重,並依據調整後的權重計算業務評估項目的調整得分。In step S130, the
為了更佳清楚易懂步驟S130,請一併參閱第3圖。第3圖為依據本揭示一實施例之第2圖之業務評估方法中之步驟S130的流程圖。步驟S130包含步驟S132~S138。在步驟S132中,藉由機器學習模型依據該些筆營運資料以及該些權重,計算業務評估項目的得分。在步驟S134中,藉由機器學習模型依據該些筆營運資料中之一筆營運資料以及一筆歷史營運資料計算第一變動率,並且依據業務評估項目的得分以及一歷史得分計算第二變動率。在步驟S134中,判斷第一變動率與第二變動率之差值是否大於預期值。在步驟S137中,藉由機器學習模型優先調整該些權重中之數值最高者以及數值最低者,使第一變動率與第二變動率之差值小於預期值。在步驟S138中,不調整該些權重。For better understanding of step S130 , please also refer to FIG. 3 . FIG. 3 is a flow chart of step S130 in the business evaluation method in FIG. 2 according to an embodiment of the present disclosure. Step S130 includes steps S132-S138. In step S132, a machine learning model is used to calculate the scores of the business evaluation items according to the pieces of operating data and the weights. In step S134, the machine learning model is used to calculate the first change rate according to one piece of operation data and one piece of historical operation data among the pieces of operation data, and to calculate the second change rate according to the score of the business evaluation item and a historical score. In step S134, it is determined whether the difference between the first rate of change and the second rate of change is greater than an expected value. In step S137 , the machine learning model is used to preferentially adjust the weight with the highest value and the weight with the lowest value, so that the difference between the first rate of change and the second rate of change is smaller than the expected value. In step S138, these weights are not adjusted.
在步驟S132中,藉由機器學習模型依據評估因素F11~F15各自對應的營運資料D1~D10及權重,計算業務評估項目的得分。以評估因素F11(平均總融資餘額)為例,108年1月總融資餘額的營運資料D1是75萬,104年至108年各年度1月總融資餘額平均的營運資料D6是156萬。108年1月總融資餘額對104年至108年各年度1月總融資餘額平均的占比是營運資料D1除以營運資料D6=75/156=0.4807,而評估因素F11(平均總融資餘額)具有4%的權重,因此108年1月評估因素F11(平均總融資餘額)的得分係0.4807*4=1.92分。In step S132 , the scores of the business evaluation items are calculated by using the machine learning model according to the corresponding operation data D1 - D10 and weights of the evaluation factors F11 - F15 . Taking the evaluation factor F11 (average total financing balance) as an example, the operating data D1 of the total financing balance in January 2018 is 750,000, and the operating data D6 of the average total financing balance in January of each year from 2014 to 2010 is 1.56 million. The proportion of the total financing balance in January 2018 to the average total financing balance in January of each year from 2014 to 2010 is the operating data D1 divided by the operating data D6=75/156=0.4807, and the evaluation factor F11 (average total financing balance) It has a weight of 4%, so the score of the evaluation factor F11 (average total financing balance) in January 2018 is 0.4807*4=1.92 points.
再舉一個例子,以評估因素F12(平均總額度)為例,108年1月總額度的營運資料D2是83萬,104年至108年各年度1月總額度平均的營運資料D7是114萬。108年1月總額度對104年至108年各年度1月總額度平均的占比是營運資料D2除以營運資料D7=83/114=0.7280,而評估因素F12(平均總額度)具有4%的權重,因此108年1月評估因素F12(平均總額度)的得分係0.7280*4=2.91分。依此類推,於108年1月,評估因素F13(平均總戶數)、評估因素F14(平均總收益)、評估因素F15(總收益於全行比重)的得分依序是3.91分、2.56分及3.56分。To give another example, take the evaluation factor F12 (average total quota) as an example, the operating data D2 of the total quota in January 2018 is 830,000, and the average operating data D7 of the total quota in January of each year from 2014 to 2010 is 1.14 million . The proportion of the total quota in January 2018 to the average total quota in January of each year from 2010 to 2018 is the operating data D2 divided by the operating data D7=83/114=0.7280, and the evaluation factor F12 (average total quota) has 4% Therefore, the score of the evaluation factor F12 (average total amount) in January 2018 is 0.7280*4=2.91 points. By analogy, in January 2018, the scores of assessment factor F13 (average total number of households), assessment factor F14 (average total income), and assessment factor F15 (proportion of total income to the entire bank) were 3.91 points and 2.56 points and 3.56 points.
需要注意的是,在當月對平均占比超過設定值(例如,1)時,會以設定值(例如,1)進行計算。It should be noted that when the average proportion of the current month exceeds the set value (eg, 1), the calculation will be performed with the set value (eg, 1).
如此,在整體業務執行情形最高得分為20分(亦即,具有20%權重)時,依據評估因素F11~F15各自的得分,機器學習模型114a 可計算得出108年1月的整體業務執行情形的得分為1.92+2.91+3.91+2.56+3.56=14.86分。In this way, when the highest score of the overall business execution situation is 20 points (that is, with 20% weight), according to the respective scores of evaluation factors F11~F15, the
在一些實施例中,業務評估項目的另外兩大類別,業務別客戶融資情況以及營運管理情形的權重可以分別被設定在36%及44%,請參閱表二及表三。
如表二所示,在第二大類別,業務別客戶融資情況之中,評估因素F21~F26各自具有6%的權重,業務評估項目三大類別之一的業務別客戶融資情況合計了評估因素F21~F26各自的權重,業務別客戶融資情況會具有36%的權重。並且,營運資料D11~D22亦分別對應評估因素F21~F26。藉由步驟S110~S132計算業務別客戶融資情況(評估因素F21~F26)的得分的分是類似於由步驟S110~S132計算整體業務執行情形(F11~F15)的方式,在此不再贅述。As shown in Table 2, in the second largest category, customer financing by business, the evaluation factors F21~F26 each have a weight of 6%, and the financing of customers by business, one of the three major categories of business evaluation items, adds up to the evaluation factors For the respective weights of F21~F26, the financing situation of customers by business will have a weight of 36%. Moreover, the operating data D11~D22 also correspond to the assessment factors F21~F26 respectively. Calculating the scores of customer financing conditions (assessment factors F21-F26) by business in steps S110-S132 is similar to the way of calculating the overall business execution conditions (F11-F15) in steps S110-S132, and will not be repeated here.
在一些實施例中,各業務平均時間成本(評估因素F21)代表的是欲評估月份(例如,當月) 新客戶開發或額度申請起案至額度核准所需時間。在一些實施例中,各業務平均時間成本(評估因素F21)可以由過去五年時間成本數值中最高減欲評估月份(例如,當月)時間成本所得之數值,該數值佔過去五年平均時間成本之比例計算。In some embodiments, the average time cost of each business (assessment factor F21) represents the time required from new customer development or quota application initiation to quota approval in the month to be evaluated (eg, the current month). In some embodiments, the average time cost of each business (assessment factor F21) can be the value obtained by subtracting the time cost of the month to be evaluated (for example, the current month) from the highest value of time cost in the past five years, and this value accounts for the average time cost of the past five years ratio calculation.
在一些實施例中,各業務平均收益(評估因素F22)代表的是欲評估月份(例如,當月)收益佔過去五年平均收益之比例。各業務平均戶數變動率(評估因素F23)代表的是欲評估月份(例如,當月)戶數變動率佔過去五年平均戶數變動率之比例。各業務平均額度變動率(評估因素F24)代表的是欲評估月份(例如,當月)額度變動率佔過去五年平均額度變動率之比例。各業務平均動撥金額(評估因素F25)代表的是欲評估月份(例如,當月)動撥金額佔過去五年平均動撥金額之比例。各業務平均融資餘額(評估因素F26)代表的是欲評估月份(例如,當月)融資餘額佔過去五年平均融資餘額之比例。In some embodiments, the average revenue of each business (assessment factor F22) represents the ratio of the revenue of the month to be evaluated (for example, the current month) to the average revenue of the past five years. The change rate of the average number of households of each business (assessment factor F23) represents the ratio of the change rate of the number of households in the month to be evaluated (for example, the current month) to the average change rate of the number of households in the past five years. The average quota change rate of each business (assessment factor F24) represents the ratio of the quota change rate of the month to be evaluated (for example, the current month) to the average quota change rate of the past five years. The average transfer amount of each business (evaluation factor F25) represents the ratio of the transfer amount in the month to be evaluated (for example, the current month) to the average transfer amount in the past five years. The average financing balance of each business (assessment factor F26) represents the ratio of the financing balance of the month to be evaluated (for example, the current month) to the average financing balance of the past five years.
請參閱表二,評估因素F21~F26的得分依序是4.35分、5.35分、6分、4.47分、6分、6分。Please refer to Table 2, the scores of evaluation factors F21~F26 are 4.35 points, 5.35 points, 6 points, 4.47 points, 6 points, 6 points.
如此,在業務別客戶融資情況最高得分為36分(亦即,具有36%權重)時,依據評估因素F21~F26各自的得分,機器學習模型114a可計算得出108年1月的業務別客戶融資情況的得分為4.35+5.35+6+4.47+6+6=32.17分。
如表三所示,在第三大類別,營運管理情形之中,評估因素F31~F34各自具有11%的權重,業務評估項目三大類別之一的業務別客戶融資情況合計了評估因素F31~F34各自的權重會具有44%的權重。並且,營運資料D23~D30亦分別對應評估因素F31~F34。藉由步驟S110~S132計算營運管理情形(評估因素F31~F34)的得分的分是類似於由步驟S110~S132計算整體業務執行情形(F11~F15)的方式,在此不再贅述。As shown in Table 3, in the third category, operation management situation, the evaluation factors F31~F34 each have a weight of 11%, and one of the three major categories of business evaluation items, business-specific customer financing, adds up to the evaluation factors F31~F34 The respective weights of F34 would have a weight of 44%. Moreover, the operating data D23~D30 also correspond to the assessment factors F31~F34 respectively. Calculating the score of the operation management situation (assessment factors F31~F34) through steps S110~S132 is similar to the method of calculating the overall business execution situation (F11~F15) through steps S110~S132, and will not be repeated here.
在一些實施例中,各資本額平均風險主評等(評估因素F31)代表的是評估月份(例如,當月)風險主評等佔過去五年平均風險主評等之比例。各資本額平均戶數(評估因素F32)代表的是評估月份(例如,當月)戶數佔過去五年平均戶數之比例。各資本額平均既有客戶維護情形/新客戶開發成效量化自我評分(評估因素F33) 代表的是評估月份(例如,當月)既有客戶(新客戶)維護情形(開發成效)量化自我評分佔過去五年平均既有客戶(新客戶)維護情形(開發成效)量化自我評分。各資本額平均未來營運預測自我評分(評估因素F34) 代表的是評估月份(例如,當月)未來營運預測自我評分佔過去五年平均未來營運預測自我評分之比例。In some embodiments, the average main risk rating of each capital amount (assessment factor F31) represents the ratio of the main risk rating of the assessment month (for example, the current month) to the average main risk rating of the past five years. The average number of households for each capital amount (assessment factor F32) represents the ratio of the number of households in the assessment month (for example, the current month) to the average number of households in the past five years. Quantified self-score (assessment factor F33) of the average existing customer maintenance status/new customer development effectiveness of each capital amount represents the quantitative self-score score of the existing customer (new customer) maintenance status (development effectiveness) of the evaluation month (for example, the current month) accounted for in the past Five-year average existing customer (new customer) maintenance status (development effectiveness) quantified self-rating. The average future operating forecast self-score (assessment factor F34) for each capital amount represents the ratio of the future operating forecast self-score for the assessment month (eg, the current month) to the past five-year average future operating forecast self-score.
請參閱表三,評估因素F31~F34的得分依序是9.27分、8.45分、6.82分及7.30分。Please refer to Table 3, the scores of evaluation factors F31~F34 are 9.27 points, 8.45 points, 6.82 points and 7.30 points in sequence.
如此,在營運管理情形最高得分為44分(亦即,具有44%權重)時,機器學習模型114a 可計算得出108年1月的營運管理情形的得分為9.27+8.45+6.82+7.30=31.84分。In this way, when the maximum score of the operation management situation is 44 points (that is, with 44% weight), the
因此,機器學習模型114a 可以計算出在欲評估月份的業務評估項目中之整體業務執行情形、業務別客戶融資情況及營運管理情形所涵蓋的全部評估因素F11~F15、F21~F26及F31~F34在欲評估月份的總和等於14.86+32.17+31.84=78.87分。Therefore, the
在一些實施例中,機器學習模型114a 可以將欲評估月份的業務項目的得分(例如,78.87分)與近五年該月的業務項目的得分進行比較,以判斷/評估欲評估的金融機構在當年與近年同月份的營運狀況。In some embodiments, the
請再參閱第3圖,在步驟S134中,藉由機器學習模型114a 依據該些筆營運資料D1~30中之對應於評估因素F14(平均總收益)的一筆營運資料D4以及一筆歷史營運資料計算第一變動率,並且依據業務評估項目的得分以及一歷史得分計算第二變動率。Please refer to FIG. 3 again. In step S134, the
其中第一變動率的公式:(當月評估因素F14的得分-上月評估因素F14的得分)/ 上月評估因素F14的得分。在一些實施例中,上月評估因素F14的得分可儲存於伺服器130或外部記憶體116,處理器112可以自伺服器130或外部記憶體116提取上月評估因素F14的得分,並執行機器學習模型114a 。並且,由於在步驟S132之中,機器學習模型114a 可以由對應於評估項目F14的營運資料D4及D9計算出當月評估因素F14的得分,即可計算得出第一變動率。在一些實施例中,上月評估因素F14的得分可以由歷史營運資料計算,歷史營運資料可以是在欲評估月份的前一個月份且對應於評估因素F14的營運資料。例如,欲評估月份是108年1月,所述歷史營運資料可以是107年12月且對應於評估因素F14的營運資料。依據歷史營運資料計算上月評估因素F14的方式類似於前述步驟S110~S132,在此不再贅述。由於當月評估因素F14的得分是依據對應於評估因素F14的營運資料D4計算而得。如此,依據對應於評估因素F14的營運資料D4以及所述歷史營運資料可以計算第一變動率。The formula for the first rate of change: (the score of the assessment factor F14 in the current month - the score of the assessment factor F14 in the previous month)/the score of the assessment factor F14 in the previous month. In some embodiments, the score of last month’s assessment factor F14 can be stored in
在一些實施例中,第二變動率的公式:(當月業務評估項目的得分平均-上月業務評估項目的得分平均)/ 上月業務評估項目的得分平均。當月業務評估項目的得分平均可以是多年內欲評估月份的業務評估項目的得分平均,例如104至108年各年度1月的業務評估項目得分的平均。並且,上月業務評估項目的得分平均可以是多年內欲評估月份前一月的業務評估項目的得分平均,例如103至107年各年度12月的業務評估項目得分的平均。換言之,上月業務評估項目的得分平均可以是該業務評估項目在歷年的欲評估月份的歷史得分平均。In some embodiments, the formula of the second rate of change: (the average score of the business evaluation item in the current month−the average score of the business evaluation item in the previous month)/the average score of the business evaluation item in the previous month. The average score of the business evaluation items in the current month may be the average score of the business evaluation items in the desired month of evaluation for many years, for example, the average score of the business evaluation items in January of each year from 104 to 108. In addition, the average score of the business evaluation items in the previous month may be the average score of the business evaluation items in the month before the desired evaluation month for many years, for example, the average score of the business evaluation items in December of each year from 103 to 107. In other words, the average score of the business evaluation item in the previous month may be the historical average score of the business evaluation item in the month to be evaluated in the calendar year.
在另一些實施例中,第二變動率的公式:(當月業務評估項目的得分-上月業務評估項目的得分)/ 上月業務評估項目的得分。上月業務評估項目的得分可以是欲評估月份前一月的業務評估項目的得分,例如,欲評估月份是108年1月,當月業務評估項目的得分是108年1月的得分,上月業務評估項目的得分是107年12月的業務評估項目的得分。換言之,上月業務評估項目的得分可以是該業務評估項目在前一年度的欲評估月份的歷史得分。由於在欲評估月份(例如,當月)的業務評估項目的得分在前述步驟S110~S132中已詳細說明其計算方式。如此,依據當月業務評估項目的得分以及所述上月業務評估項目的得分可以計算第二變動率。In some other embodiments, the formula of the second rate of change: (score of the business evaluation item of the current month−score of the business evaluation item of the previous month)/score of the business evaluation item of the previous month. The score of the business evaluation item in the previous month can be the score of the business evaluation item in the month before the month to be evaluated. For example, the month to be evaluated is January 108, and the score of the business evaluation item in the current month is the score in January 108. The score for the assessment item is the score for the business assessment item in December 107. In other words, the score of the business evaluation item in the previous month may be the historical score of the business evaluation item in the month to be evaluated in the previous year. Since the scores of the business evaluation items in the month to be evaluated (for example, the current month) have been calculated in detail in the aforementioned steps S110-S132. In this way, the second rate of change can be calculated according to the score of the business evaluation item of the current month and the score of the business evaluation item of the previous month.
在一些實施例中,上月業務評估項目的得分平均以及非當年度欲評估月份的業務評估項目的得分亦可儲存於伺服器130或外部記憶體116,處理器112可以自伺服器130或外部記憶體116提取上月業務評估項目的得分平均以及非當年度欲評估月份的業務評估項目的得分,並執行機器學習模型114a 。並且,由於在步驟S132之中,機器學習模型114a 可以由營運資料D1~D30計算出業務評估項目的得分,在配合上月業務評估項目的得分平均以及非當年度欲評估月份的業務評估項目的得分,即可計算得出第二變動率。In some embodiments, the average scores of the business evaluation items in the previous month and the scores of the business evaluation items in the month not to be evaluated in the current year can also be stored in the
在步驟S136中,判斷第一變動率與第二變動率之差值是否大於預期值。在一些實施例中,可將預期值設定在5%。In step S136, it is determined whether the difference between the first rate of change and the second rate of change is greater than an expected value. In some embodiments, the expected value may be set at 5%.
在一些實施例中,若第一變動率與第二變動率之差值大於預期值,進行步驟S137,藉由機器學習模型114a 優先調整前述權重中之數值最高者以及數值最低者,使第一變動率與第二變動率之差值小於預期值。接著,再進行步驟S132,藉由機器學習模型114a 依據該些筆營運資料D1~D30以及調整後的權重,計算業務評估項目的調整得分。並接續步驟S134~S136,即可驗證第一變動率與第二變動率之差值是否小於預期值,藉此取得業務評估項目在欲評估月份的可信的調整得分。如此,業務評估系統100中的機器學習模型114a依據第一變動率以及第二變動率調整權重,可以更準確地計算及評估業務評估項目在待評估月份的得分以正確的判斷該分支機構最佳業務推展時機。具體而言,藉由步驟S110~S130可以計算業務評估項目在各個月份的多個得分,業務評估項目在各個月份的多個得分中之最高者可作為該分支機構最佳業務推展時機。In some embodiments, if the difference between the first rate of change and the second rate of change is greater than the expected value, proceed to step S137, and use the
在另一些實施例中,若第一變動率與第二變動率之差值大於預期值,機器學習模型114a 依據各個業務評估項目各個評估因素F11~F15、F21~F26及F31~F34的得分中之最高者以及最低者調整該兩個評估因素的權重。In some other embodiments, if the difference between the first rate of change and the second rate of change is greater than the expected value, the
若第一變動率與第二變動率之差值小於或等於預期值,接續步驟S138,不調整該些權重。If the difference between the first rate of change and the second rate of change is less than or equal to the expected value, proceed to step S138 and not adjust the weights.
請參閱第4圖,第4圖為依據本揭示一實施例之業務評估方法S100的流程圖。如第4圖所示,業務評估方法S100更包含步驟S140~S160。在步驟S140中,藉由輸入介面接收複數個業務類型各自的複數筆營運子資料,該些營運子資料分別對應複數個評估因素中之一部分。在步驟S150中,藉由機器學習模型依據該些業務類型各自的該些筆營運子資料以及對應的權重,計算該些業務類型的得分。在步驟S160中,藉由機器學習模型依據該些業務型的得分,判斷該些業務類型中之最佳者。Please refer to FIG. 4 . FIG. 4 is a flowchart of a business evaluation method S100 according to an embodiment of the present disclosure. As shown in FIG. 4 , the business evaluation method S100 further includes steps S140 - S160 . In step S140, a plurality of operation sub-data of each of the plurality of business types are received through the input interface, and the operation sub-data respectively correspond to one part of the plurality of evaluation factors. In step S150, the machine learning model is used to calculate the scores of the business types according to the respective operation sub-data and corresponding weights of the business types. In step S160, a machine learning model is used to determine the best of the business types according to the scores of the business types.
在步驟S140中,藉由輸入介面120接收接收複數個業務類型各自的複數筆營運子資料,該些營運資料分別對應複數個評估因素中之一部分。請一併參下列閱表四。
如表四所示,業務類型A可以為資本額100萬以下的融資。業務類型B可以為101至500萬元的融資。業務類型C可以為501至3,000萬元的融資。業務類型D可以為3,001萬元至1億元的融資。業務類型E可以為1億元以上的融資。As shown in Table 4, business type A can be used for financing with a capital amount of less than 1 million. Business Type B can provide financing for $101 to $5 million. Business Type C can be used for financing from 501 to 30 million yuan. Business Type D can provide financing ranging from RMB 30.01 million to RMB 100 million. Business type E can provide financing of more than 100 million yuan.
在步驟S150中,藉由機器學習模型114a 依據業務類型A~E各自的筆營運子資料以及對應的權重,計算業務類型A~E的得分。以業務類型A(100萬以下的融資)為例,於108年1月,對應於評估因素F21(平均時間成本)的業務類型A的營運子資料是83小時,對應於評估因素F21(平均時間成本)的業務類型A~E的營運子資料合計是508小時。對應於評估因素F21(平均時間成本)的業務類型A於業務類型A~E中的占比為83小時除以508小時=0.16。對應於評估因素F21的業務類型A於業務類型A~E中的占比乘以評估因素F21的得分即為業務類型A於評估因素F21的得分(亦即,0.16*4.35=0.71分)。在業務別客戶融資情況之中的所有評估因素F21~F26之中,業務類型A的得分是0.71+0.95+6.32+0.84+0.69+0.93=10.44。業務類型B~E的計算方式類似於業務類型A,在此不再贅述。In step S150 , the
如此,在業務別客戶融資情況之中的所有評估因素F21~F26,業務類型A~E各自的得分依序是10.44分、0.28分、15.14分、2.59分、3.72分。 In this way, among all evaluation factors F21~F26 in the customer financing situation by business category, the respective scores of business types A~E are 10.44 points, 0.28 points, 15.14 points, 2.59 points, and 3.72 points respectively.
在步驟S160中,藉由機器學習模型114a依據該些業務型的得分,判斷該些業務類型A~E中之最佳者。舉例而言,業務類型A~E各自的得分依序是10.44分、0.28分、15.14分、2.59分、3.72分,機器學習模型114a便可判斷在業務類型A~E中得分最高的業務類型C為108年1月最佳者。
In step S160, the
在另一些實施例中,機器學習模型114a針對業務別客戶融資情況的所有評估因素F21~F26,依各項業務類型A~E分別比對整體業務執行情形之評估因素F14(平均總收益)及評估因素F15(總收益於全行比重),以及營運管理情形之評估因素F34(平均未來營運預測自我評分),取得總分最高者,為業務評估項目的該大類別(業務別客戶融資情況)中之預測最佳獲利策略及佈局之分析結果。
In some other embodiments, the
第5圖為依據本揭示一實施例之業務評估方法S100的流程圖。業務評估方法S100更包含步驟S170、S180、S182及S184。在步驟S170之中,藉由機器學習模型114a依據該些業務類型A~E中之最佳者的該些筆營運子資料其中一者以及一筆歷史營運子資料計算第三變動率。
FIG. 5 is a flowchart of a business evaluation method S100 according to an embodiment of the present disclosure. The business evaluation method S100 further includes steps S170, S180, S182 and S184. In step S170, the third rate of change is calculated by the
其中第三變動率的公式:(當月業務類型N的得分-上月業務類型N的得分)/上月業務類型N的得分,其中業務類型N可以是前述的業務類型A~E。在一些實施例中,上月業務類型N的得分可儲存於伺服器130或外部記憶體116,處理器112可以自伺服器130或外部記憶體116提取上月業務類型N的得分,並執行機器學習模型114a 。並且,由於在步驟S132之中,機器學習模型114a 可以由對應於評估項目F21~F26的營運子資料計算出當月評估因素F14的得分,即可計算得出第三變動率。在一些實施例中,上月業務類型N的得分可以由歷史營運子資料計算。在一些實施例中,依據評估項目F21~F26的數量,歷史營運子資料可以是一筆或多筆。歷史營運子資料可以是在欲評估月份前一個月的業務類型N的營運子資料。例如,欲評估月份是108年1月,所述歷史營運子資料可以是107年12月的營運子資料。計算歷史營運子資料的方式類似於前述步驟S110~S132,在此不再贅述。The formula for the third rate of change: (the score of the business type N in the current month-the score of the business type N in the previous month)/the score of the business type N in the previous month, wherein the business type N can be the aforementioned business types A~E. In some embodiments, the score of last month’s business type N can be stored in the
在步驟S180中,機器學習模型114a 判斷第一變動率與第三變動率之差值是否大於預期值。在一些實施例中,可將預期值設定在5%。In step S180, the
在一些實施例中,若第一變動率與第三變動率之差值大於預期值,進行步驟S182藉由機器學習模型114a 優先調整該些權重中之數值最高者以及數值最低者,使第一變動率與第三變動率之差值小於預期值。接著,再進行步驟S132,藉由機器學習模型114a 依據該些筆營運資料D1~D30以及調整後的權重,計算業務評估項目的調整得分。In some embodiments, if the difference between the first rate of change and the third rate of change is greater than the expected value, proceed to step S182 by using the
在另一些實施例中,若第一變動率與第三變動率之差值大於預期值,機器學習模型114a 依據各個業務評估項目各個評估因素F11~F15、F21~F26及F31~F34的得分中之最高者以及最低者調整該兩個評估因素的權重。In some other embodiments, if the difference between the first rate of change and the third rate of change is greater than the expected value, the
並接續步驟S134~S136,即可驗證第一變動率與第二變動率之差值是否小於預期值,藉此取得業務評估項目更加精確且可信的調整得分。And continuing with steps S134-S136, it can be verified whether the difference between the first rate of change and the second rate of change is smaller than the expected value, thereby obtaining a more accurate and credible adjusted score for the business evaluation item.
若第一變動率與第三變動率之差值小於或等於預期值,接續步驟S184,不調整該些權重。If the difference between the first rate of change and the third rate of change is less than or equal to the expected value, proceed to step S184 and not adjust the weights.
綜上所述,藉由業務評估系統100中的機器學習模型114a依據第一變動率、第二變動率調整權重,可以更準確地計算及評估業務評估項目在待評估月份的得分以判斷該分支機構最佳業務推展時機。並且業務評估系統100中的機器學習模型114a可以計算業務類型A~E中得分最高者,為銀行分支機構最佳獲利策略及佈局之分析結果。再者,可以依據業務評估項目在各月的歷年平均得分中之最高者作為該分支機構最佳業務推展時機,依據各業務類型的得分最高者作為該分支機構最佳業務推展類型。如此,由業務評估系統100中的機器學習模型114a 算出可信的業務評估項目的得分以及業務類型A~E各自的得分,可以被結合為業務管理作為銀行分支機構未來獲利策略及佈局之方向。To sum up, by adjusting the weights of the
雖然本案已以實施方式揭露如上,然其並非限定本案,任何熟習此技藝者,在不脫離本案之精神和範圍內,當可作各種之更動與潤飾,因此本案之保護範圍當視後附之申請專利範圍所界定者為準。Although this case has disclosed the above in the way of implementation, it does not limit this case. Anyone who is familiar with this technology can make various changes and modifications without departing from the spirit and scope of this case. Therefore, the scope of protection of this case should be regarded as attached The one defined in the scope of the patent application shall prevail.
為使本揭露之上述和其他目的、特徵、優點與實施例能更明顯易懂,所附符號之說明如下:
100:業務評估系統
110:業務評估主機
112:處理器
114:內部記憶體
114a:機器學習模型
116:外部記憶體
120:輸入介面
S:電子裝置
S100:業務評估方法
S110,S120,S130,S132,S134,S136,S137,S138,S140,S150,S160,S170,S180,S182,S184:步驟
In order to make the above and other purposes, features, advantages and embodiments of the present disclosure more obvious and easy to understand, the descriptions of the attached symbols are as follows:
100: Business Evaluation System
110: Business Evaluation Host
112: Processor
114:
為使本揭露之上述和其他目的、特徵、優點與實施例能更明顯易懂,所附圖式之說明如下: 第1圖為依據本揭露一實施例之業務評估系統。 第2圖為依據本揭示一實施例之業務評估方法的流程圖。 第3圖為依據本揭示一實施例之第2圖之業務評估方法中之步驟S130的流程圖。 第4圖為依據本揭示一實施例之業務評估方法的流程圖。 第5圖為依據本揭示一實施例之業務評估方法的流程圖。 In order to make the above and other purposes, features, advantages and embodiments of the present disclosure more comprehensible, the accompanying drawings are described as follows: Fig. 1 is a business evaluation system according to an embodiment of the present disclosure. FIG. 2 is a flowchart of a business evaluation method according to an embodiment of the present disclosure. FIG. 3 is a flow chart of step S130 in the business evaluation method in FIG. 2 according to an embodiment of the present disclosure. FIG. 4 is a flowchart of a business evaluation method according to an embodiment of the present disclosure. FIG. 5 is a flowchart of a business evaluation method according to an embodiment of the present disclosure.
國內寄存資訊(請依寄存機構、日期、號碼順序註記) 無 國外寄存資訊(請依寄存國家、機構、日期、號碼順序註記) 無 Domestic deposit information (please note in order of depositor, date, and number) none Overseas storage information (please note in order of storage country, institution, date, and number) none
S100:業務評估方法 S100: Business evaluation methods
S110,S120,S130:步驟 S110, S120, S130: steps
Claims (10)
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
TW110122201A TWI795809B (en) | 2021-06-17 | 2021-06-17 | Business evaluation system and method therefore |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
TW110122201A TWI795809B (en) | 2021-06-17 | 2021-06-17 | Business evaluation system and method therefore |
Publications (2)
Publication Number | Publication Date |
---|---|
TW202301238A TW202301238A (en) | 2023-01-01 |
TWI795809B true TWI795809B (en) | 2023-03-11 |
Family
ID=86658150
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
TW110122201A TWI795809B (en) | 2021-06-17 | 2021-06-17 | Business evaluation system and method therefore |
Country Status (1)
Country | Link |
---|---|
TW (1) | TWI795809B (en) |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
TW200844894A (en) * | 2007-03-23 | 2008-11-16 | Res Affiliates Llc | Using accounting data based indexing to create a portfolio of financial objects |
CN109102140A (en) * | 2018-06-11 | 2018-12-28 | 北交金科金融信息服务有限公司 | A kind of growing quality evaluation method of base artificial intelligence and big data technology |
US20190066020A1 (en) * | 2014-03-31 | 2019-02-28 | Kountable, Inc. | Multi-Variable Assessment Systems and Methods that Evaluate and Predict Entrepreneurial Behavior |
CN110969002A (en) * | 2018-09-28 | 2020-04-07 | 北京国双科技有限公司 | Financial index analysis report generation method and device |
CN111612549A (en) * | 2020-05-28 | 2020-09-01 | 中认英泰检测技术有限公司 | Construction method of platform operation service system |
CN112862223A (en) * | 2021-03-29 | 2021-05-28 | 中信银行股份有限公司 | Method and device for evaluating business activity value, decision method, equipment and medium |
TWM619067U (en) * | 2021-06-17 | 2021-11-01 | 華南商業銀行股份有限公司 | Business evaluation system |
-
2021
- 2021-06-17 TW TW110122201A patent/TWI795809B/en active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
TW200844894A (en) * | 2007-03-23 | 2008-11-16 | Res Affiliates Llc | Using accounting data based indexing to create a portfolio of financial objects |
US20190066020A1 (en) * | 2014-03-31 | 2019-02-28 | Kountable, Inc. | Multi-Variable Assessment Systems and Methods that Evaluate and Predict Entrepreneurial Behavior |
CN109102140A (en) * | 2018-06-11 | 2018-12-28 | 北交金科金融信息服务有限公司 | A kind of growing quality evaluation method of base artificial intelligence and big data technology |
CN110969002A (en) * | 2018-09-28 | 2020-04-07 | 北京国双科技有限公司 | Financial index analysis report generation method and device |
CN111612549A (en) * | 2020-05-28 | 2020-09-01 | 中认英泰检测技术有限公司 | Construction method of platform operation service system |
CN112862223A (en) * | 2021-03-29 | 2021-05-28 | 中信银行股份有限公司 | Method and device for evaluating business activity value, decision method, equipment and medium |
TWM619067U (en) * | 2021-06-17 | 2021-11-01 | 華南商業銀行股份有限公司 | Business evaluation system |
Also Published As
Publication number | Publication date |
---|---|
TW202301238A (en) | 2023-01-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109977151B (en) | Data analysis method and system | |
Huang | A new perspective for optimal portfolio selection with random fuzzy returns | |
CN110796539A (en) | Credit investigation evaluation method and device | |
Akçiğit et al. | Firm-level impact of credit guarantees: Evidence from Turkish credit guarantee fund | |
Peresetsky et al. | Probability of default models of Russian banks | |
TWM619067U (en) | Business evaluation system | |
TWI795809B (en) | Business evaluation system and method therefore | |
Tikhonov et al. | The relationship between the financial performance of banks and the quality of credit scoring models | |
KR20200111064A (en) | System for Debt Repayment Capability Evaluation Of Corporation | |
García et al. | Evolution of sovereign rating models in the current crisis | |
Omoshoro‐Jones et al. | Intra‐regional spillovers from Nigeria and South Africa to the rest of Africa: New evidence from a FAVAR model | |
So et al. | Multivariate GARCH models with correlation clustering | |
Kurmanova et al. | Simulation of Banking Sector’s Investment Potential | |
Park et al. | Determinants of domestic public debt crisis | |
Pensieroso et al. | The Gold Standard and the Great Depression: a Dynamic General Equilibrium Model | |
CN110232593A (en) | The total retail sales of consumer goods add up the prediction technique and system of amplification | |
TWI749938B (en) | Scoring system and method for financial operation | |
Cicowiez et al. | A Simple Stylized Long-run Growth Model for Haiti | |
Bilgin | Endogenous Production Networks and Supply Chain Disruptions | |
Genemo | Velocity of Money in Ethiopia | |
Zhang et al. | The effect of new electronic money on money supply—A study about money based on information technology | |
İlhan et al. | Analyzing exchange market pressure dynamics with Markov regime switching: The case of Turkey | |
Jadwani et al. | Operational risk management: analysis of challenges faced by banks in India | |
Döpp et al. | Modeling Non-Maturing Deposits: A Procedure for the Determination of the Minimal Time Series Length Required for Model Calibration and Back-Testing | |
TWM609222U (en) | Financial operation grading system |