WO2000054186A1 - Systeme de prevision financiere et procede d'appreciation et de gestion des risques - Google Patents

Systeme de prevision financiere et procede d'appreciation et de gestion des risques Download PDF

Info

Publication number
WO2000054186A1
WO2000054186A1 PCT/US2000/006186 US0006186W WO0054186A1 WO 2000054186 A1 WO2000054186 A1 WO 2000054186A1 US 0006186 W US0006186 W US 0006186W WO 0054186 A1 WO0054186 A1 WO 0054186A1
Authority
WO
WIPO (PCT)
Prior art keywords
applicant
population
forecast
portfolio
generating
Prior art date
Application number
PCT/US2000/006186
Other languages
English (en)
Other versions
WO2000054186A8 (fr
Inventor
Anand V. Deo
Original Assignee
Mathematical Modellers Inc.
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 Mathematical Modellers Inc. filed Critical Mathematical Modellers Inc.
Publication of WO2000054186A1 publication Critical patent/WO2000054186A1/fr
Publication of WO2000054186A8 publication Critical patent/WO2000054186A8/fr

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
    • G06Q40/08—Insurance

Definitions

  • the present invention geneially relates to risk analysis and management in the financial industry Specifically, the present invention relates to systems and methods for performing risk assessment of financial service applications and risk management of financial portfolios
  • the ability ol a financial institution to become and remain a profitable entity is highly dependent on assessing and managing ⁇ sks associated with its investments
  • a financial service e g , a loan or a line ot credit
  • the threshold value may be used to control the level of risk associated with a financial institution's investments If the credit worthiness is accurately measured and the threshold value is properly set, a financial institution is well equipped to prosper
  • the conventional method for determining credit worthiness is called credit scoring Credit sco ⁇ ng is premised on the idea that the recent past is an indicator of the near future Credit sco ⁇ ng provides a method by which risk may be measured and quantified, and is p ⁇ ma ⁇ ly used for determining ⁇ sk associated with credit applications, but is also used for determining collections strategies, determining credit limits, evaluating account renewal, autho ⁇ zing transactions and developing marketing strategies
  • the credit sconng method is discussed in the context of credit and loan applications
  • the credit sco ⁇ ng process involves collecting histo ⁇ cal data from a population of individuals relating to certain att ⁇ butes of the individuals, for example, credit history, debts, assets, employment, and residence These att ⁇ butes are then correlated to "good” and "bad” loans.
  • the next step m the credit scoring process involves correlating the attributes of an individual applicant to attributes from the population to generate a set of matched attributes. The corresponding weight for each matched att ⁇ bute is added to obtain a score for the individual applicant.
  • the score is a quantification of ⁇ sk associated with the application of the individual, and represents the completion of the ⁇ sk assessment process using the credit sco ⁇ ng method
  • the application If the score is less than the threshold value as set by the financial institution, the application is rejected If, on the other hand, the score is greater than the threshold value, the application is approved An approved application paves the way for an agreement between the financial institution and the individual.
  • the agreement provides a loan or line of credit to the individual from the financial institution in exchange for a promise to repay according to specified terms (pe ⁇ od, interest rate, etc.)
  • the agreement also represents a discrete investment to be added to the financial institution's portfolio.
  • the credit scoring method is premised on the idea that the recent past is a predictor of the near future, the near future must be a mirror image of the recent past Va ⁇ ables from a past time period are associated with good and bad loans and are assumed to have the same association in a future time period of equal duration
  • the accuracy of the credit sco ⁇ ng method is limited to the accuracy of this assumption.
  • the accuracy of this assumption is limited to the extent that the future holds the exact same va ⁇ able association as the past
  • the accuracy of the credit sconng method is limited bv the assumption that the future is a static replication of the past.
  • the credit sco ⁇ ng method also treats individual applicants as static objects that are assigned a score corresponding to their credit worthiness at the time of the application. In reality, however, individual applicants are dynamic in that their att ⁇ butes change with time and thus their credit worthiness changes with time. Thus, the accuracy of the credit scoring method is also limited because it fails to treat individual applicants as dynamic entities
  • the accuracy of the credit sco ⁇ ng method is limited because it fails to treat individual applicants as dynamic entities and it fails to tieat time as a dynamic function
  • the present invention recognizes that individuals are dynamic entities wherein their att ⁇ butes change with time and that the future will inevitably present different events occur ⁇ ng at different times than in the past
  • the present invention provides a system and method lor generating data such as a forecast indicative of risk associated with an application tor financial services (e.g , loans, lines of credit, etc ) for purposes of application decision making
  • an application tor financial services e.g , loans, lines of credit, etc
  • the present invention provides a system and method for generating data such as a forecast indicative of ⁇ sk associated with the portfolio for purposes of portfolio ⁇ sk management
  • the present invention is a significant improvement over the p ⁇ or art m that it does not rely on the future to mirror the past, but rather uses histo ⁇ cal expe ⁇ ence to provide strignos and associated probabilities to generate a forecast of performance.
  • the present invention provides a method for generating data indicative of risk associated with an application for a financial service by an applicant
  • the method includes the basic step of generating a forecast of performance of the applicant wherein the forecast comp ⁇ ses a se ⁇ es of discreterialnos applied over a finite time line, using a template called an aging stnp Such a forecast is indicative of nsk associated with the application
  • the forecast may be compared to a ⁇ sk threshold of the financial institution and the application may be accepted or rejected if the forecast is greater than or less than the ⁇ sk threshold
  • a plurality of forecasts are generated including an optimistic forecast, a pessimistic forecast and a neutral forecast
  • the series of discrete plausible are selected from a set of applicant strignos
  • the process of generating a forecast involves modifying an applicant performance curve
  • the applicant performance curve compnses applicant energy as a function of time having an initial energy, and may also be expressed in terms of an algo ⁇ thm or a senes of algonthms
  • the curve is modified by applying the senes of discrete scenarios points in time based on the probability of occunence of each discrete strig ⁇ o at each point in time
  • the energy at each point in time changes in an amount corresponding to the magnitude and direction of the discrete scenario applied at the point m time to obtain a modified performance curve
  • the modified performance curve essentially compnses the forecast of performance, which is indicative of ⁇ sk
  • the process of generating applicant scenarios involves sampling the applicant data for changes in the att ⁇ butes as a function of time to generate applicant sequences Each sequence is characte ⁇ zed as a positive, negative or neutral influence A probability of occurrence and an intensity is calculated for each sequence The magnitude, direction and probability of occurrence of each applicant scenario co ⁇ elates to the intensity, influence and probabilitv of occunence of each applicant sequence, respectively
  • the process of generating population toysnos involves sampling the population data for changes of the attnbutes as a function of time to generate population sequences
  • the population sequences are conelated to successful loans and failed loans to generate population sequences having a positive influence and a negative influence, respectively
  • a probability of occurrence and an intensity is calculated for each population sequence
  • the population sequences are then sampled for common patterns to generate stable population sequences
  • the stable population sequences are then classified based on association with class attnbutes of the population to generate classes of stable population sequences having class attnbutes
  • the class attributes are matched to applicant att ⁇ butes to generate matched population sequences applicable to the applicant
  • the magnitude, direction and probability of occurrence of each population scenario co ⁇ elates to the intensity, influence and probability of occunence of each matched population sequence, respectively.
  • the present invention provides a method for generating data indicative of ⁇ sk associated with a portfolio of financial service agieements
  • the method includes the basic step of generating a forecast of performance of the portfolio wherein the forecast comprises a senes of discrete scenarios applied over a finite time line, using a template called an agmg st ⁇ p.
  • a forecast is indicative of risk associated with the portfolio
  • the series of discrete scenarios is applied to a set of individual forecasts of performance for each agreement or a class of agreements in the portfolio to generate the portfolio forecast.
  • the senes ot discrete scenarios aie selected from a set of global scenarios generated from global attributes affecting all agreements within the portfolio
  • the global data may include macio economic information and or financial institution information.
  • nsk management process Important to the nsk management process is the step of identifying steady state conditions within the portfolio. More significant to the process is the step of identifying a steady state condition of the entire portfolio, refe ⁇ ed to as the transient equilibrium point of the portfolio This may be accomplished by synchronizing the agreements within the portfolio as a function of time The portfolio charactenstics may then be identified at the steady state conditions and the transient equihbnum point
  • FIG. 1 is a flow chart illustrating the credit sco ⁇ ng method of the prior art
  • FIG. 2 is a flow chart illustrating a computer implemented method for generating data indicative of nsk associated with an application for financial services in accordance with an exemplary embodiment of the present invention
  • FIG 3 is a flow chart illustrating a computer implemented method for generating scenarios for use in the method shown in FIG 2,
  • FIG 4 is a flow chart illustrating a computer implemented method for generating an applicant forecast for use in the method shown m FIG. 2
  • FIG 5 is a flow chart illustrating a computer implemented method for generating applicant strignos for use in the method shown m FIG. 3,
  • FIG 6 is a flow chart illustrating a computer implemented method for generating population scenarios for use m the method shown in FIG 3,
  • FIGS. 7A-7C illustrate the method of applying a series of discrete scenarios to a finite time line using a template called an aging strip to provide an optimistic forecast
  • FIGS. 7D-7F illustrate the method of applying a senes of discrete strignos to a finite time line using a template called an aging st ⁇ p to provide a pessimistic forecast
  • FIG 8 is a flow chart illustrating a computer implemented method for generating data indicative of ⁇ sk associated with a portfolio of financial service agreements in accordance with an exemplary embodiment of the present invention
  • FIG 9 is a flow chart illustrating a computer implemented method for generating a portfolio forecast for use m the method shown in FIG. 8; and
  • FIG. 10 is a schematic diagram illustrating a computer system for generating data indicative of ⁇ sk associated with an application for financial services and/or a portfolio of financial service agreements in accordance with an exemplary embodiment of the present invention
  • a method or algorithm is herein, generally, conceived to be a self-consistent sequence of steps leading to a desire result. These steps are those requiring physical manipulations of physical quantities. Usually, though not necessarily, these quantities take the form of electrical or magnetic signals capable of being stored, transfened, combined, compared, and otherwise manipulated. It is often convenient, principally for reasons of common usage, to refer to these signals as bits, values, elements, symbols, characters, terms, numbers, data, or the like. It should be kept in mind, however, that all of these and similar terms are to be associated with the appropriate physical quantities and are merely convenient labels applied to these quantities.
  • the manipulations performed are often referred to in terms such as adding, compa ⁇ ng, generating, modifying, applying, conelating, calculating, sampling, and the like, which are commonly associated with mental operations performed by human operators. No such compatibility of a human operator is necessary, or desirable in most cases, in any of the operations described herein.
  • the methods and operations contemplated herein are machine or computer operations. Useful machines for perfo ⁇ ning the operations and methods of the present invention include general-purpose digital computers or other similar devices.
  • the present invention relates to method steps for operating a computer in processing electrical or other (e.g., mechanical, chemical, magnetic) physical signals to generate other desired physical signals.
  • the present invention also relates to an apparatus for performing these methods and operations.
  • This apparatus may be specially constructed for the required purposes or it may comprise a general-purpose computer selectively activated or reconfigured by a computer program stored in the computer.
  • the algorithms and methods presented herein are not inherently related to a particular computer system or other apparatus.
  • Various general-purpose computer systems may be used with computer programs written in accordance with the teachings of the present invention, or it may prove to be more convenient to construct a specialized apparatus to perform the required method steps.
  • the required structure for such machines or computers will be apparent to those skilled in the art in light of the description given below.
  • the present invention is preferably implemented for practice by a computer, e.g., a source code expression of the present invention is input to the computer to control operations therein. It is contemplated that a number of source code expressions, in one or many computer languages, may be utilized to implement the present invention.
  • a variety of computer systems can be used to practice the present invention, including, for example, a personal computer, an engineering workstation, an enterprise server, etc. The present invention, however, is not limited to the practice on any one particular computer system, and the selection of a particular computer system can be made for many reasons.
  • the credit scoring method 10 is the conventional method for determining credit worthiness of an applicant who is applying for a financial service such as a loan or line of credit.
  • a loan is used as the financial service.
  • the background processes 12 Prior to beginning the credit scoring method 10, a number of background processes 12 are performed.
  • the background processes 12 begin with the step 14 of obtaining credit history data of a population.
  • the credit history data be obtained from a credit bureau or a financial institution.
  • the historical data relates to certain attributes of individuals within the population. For example, the attributes may include information related to debts, assets, employment, and residence.
  • the next step 16 is to correlate the attributes to "good” and "bad” loans.
  • the definition of a "good” or "bad” loan varies depending on the financial institution's use of the loan, but generally conelates to a profitable loan or a non-profitable loan, respectively.
  • the next step 18 is to assign weights the to population attributes.
  • Weights are assigned to the attributes of the population as a function of how much of an impact the particular attribute is perceived to have on the outcome of the loan as either a "good” or "bad” loan
  • a collection of population attnbutes and corresponding weights are provided by performing the background processes 12 Once the background processes 12 are complete, the credit sconng method 10 may be applied to a loan application
  • the initial step 22 in the credit sco ⁇ ng process is to collect att ⁇ butes 22 of the applicant
  • the attributes of the applicant are obtained from applicant data 24 which, in turn, is obtained from a credit report agency and/or the loan application
  • the next step 26 is to compare and match attributes Specifically, the attributes of the applicant are compared to the attnbutes of the population to obtain a set of matched attnbutes with corresponding weights
  • the next step 28 is to add the co ⁇ esponding weight for each matched attribute to obtain a score for the individual applicant
  • the score is a qualification of risk associated with the individual and completes the credit sco ⁇ ng process 10
  • the next step 30 m the loan application process is to compare the score of the applicant to a risk threshold of the financial institution considering the application.
  • the risk threshold of the financial institution is obtained from financial institution nsk data 32 If the score of the applicant is less than the ⁇ sk threshold of the financial institution, the loan application is rejected 34 If the applicant score is greater than the nsk threshold of the financial institution, the loan application is approved 36 If the loan application is approved, a loan agreement 38 may be executed between the applicant and the financial institution After a decision has been reached with regard to the application, the loan application process is complete 40
  • the credit scoring method 10 is premised on the theory that the recent past is a predictor of the near future
  • the accuracy of the credit scoring method 10 is limited to the extent that the future holds the exact same va ⁇ able association (1 e , co ⁇ elation of attributes to "good” and “bad” loans) as
  • the present invention does not rely on the future to mirror the past, but rather utilizes histoncal expe ⁇ ence to provide narrativenos and associated probabilities to generate a forecast of performance
  • the forecast of performance is indicative of risk
  • a financial institution may utilize such a forecast to render a decision and take action with regard to the application or portfolio, with a higher degree of accuracy and confidence than with prior art methods
  • FIG. 2 shows a flow chart illustrating a computer implemented method 200 for generating data indicative of risk associated with an application for financial services in accordance with an exemplary embodiment of the present invention
  • the method 200 involves the basic steps 60, 70 of generating a forecast of performance of the applicant utilizing a series of discrete scenarios
  • the step 70 of generating the applicant forecast requires the pnor step 60 of generating scenarios from data 100 Data 100 mav comp ⁇ se data specific to the applicant, data specific to applicants within a population, and/or data applicable to an entire population
  • a decision 80 may be rendered with regard to the application as to whether to accept or reject the application.
  • the decision to accept or reject the application involves the steps of companng the forecast to a risk threshold of the financial institution and accepting the application if the forecast is greater than or otherwise better than the ⁇ sk threshold Conversely, the application may be rejected if the forecast is less than or otherwise worse than the ⁇ sk threshold Risk threshold, as used herein, is more likely to be stated in terms of a plurality of vanable limitations, as opposed to a single discrete value
  • the nsk threshold of the financial institution may be set bv the financial institution based on financial institution nsk experience If the application is accepted, a financial service agreement 82 may be entered into between the applicant and the financial institution
  • FIG 3 shows a flow chart illustrating m detail the step 60 of generating narrativenos for use in the method 200 shown in Figure 2
  • the step 60 of generating cutenos may involve three discreet subprocesses, namely the step 1 10 of generating applicant sentencesnos, the step 130 ot generating of population toysnos, and the step 150 of generating global narrativenos, m order of importance
  • the generation of applicant catsnos utilizes applicant data 102, which may be obtained, for example, from the financial service application or from a credit bureau.
  • the applicant data 102 is indicative of attnbutes of the applicant These attnbutes include, for example, age, mantal status, income, educational expenence, professional expenence, assets, liabilities, etc , applicable to the applicant.
  • the generation of applicant scenarios is discussed in more detail with reference to Figure 5
  • the generation of population toys requires the use of both applicant data 102 and population data 104, because a conelation must be established between applicant att ⁇ butes and population attnbutes m order to obtain matched jewenos
  • Population data 104 may be obtained, for example, from a financial institution's historical data
  • the population data 104 is indicative of attributes of a plurality of individuals withm the population
  • For the generation of population strignos the atomicity of the individuals withm the population is violated
  • the atomicity or independence ol the individuals withm the population is violated for purposes of sampling attributes and generating sequences discussed hereinafter
  • the population data 104 is essentially turned on its side and treated as a pool of continuous va ⁇ ables du ⁇ ng this process
  • the population att ⁇ butes are the same as the applicant attributes described above, except the population attributes are applicable to a plurality of applicants withm the population, rather than a single individual.
  • global data 106 which may be obtained, for example, from the financial institution and from general economic data commonly available to the public
  • the global data 106 is indicative of global attnbutes affecting all applicants withm the population.
  • the global attnbutes are common to each and every applicant within the population.
  • global data or attnbutes include two distinct types- global to an individual application but internal to the financial institution and global to the financial institution Applicant strignos pnma ⁇ ly use the first while nsk management process uses the later The ⁇ sk management process is explained later in this document.
  • Examples of global attnbutes include, for example, increases in the pnme rate, general macro-economic tiends, general political trends having economic impact, financial institution policies, etc
  • the next step is to combine the plausiblenos to obtain a set of plausiblenos 62
  • the set of scenarios 62 preferably includes all three types (applicant, population and global) of strignos
  • the set may merely compnse a subset of the three
  • the set of strignos 62 may only compnse applicant sentencesnos, or a combination of applicant narrativenos and population narrativenos
  • Each scenario includes a probability of occurrence and an effect on future performance of the applicant Specifically, each strigno generated has a probability of occunence at any given point in time This probability of occurrence should not be confused with the probability of a scenario being added to the set of possible narrativenos 62 This later probability is indicative of the accuracy of the strigno generation process and indirectly the accuracy of the forecast
  • each scenario has an influence or an effect on future performance that may be charactenzed as a positive influence, a negative influence, or a neutral influence
  • FIG 4 shows a flow chart illustrating in detail the step 70 of generating a forecast for use in the method 200 shown in Figure 2
  • the process of generating an applicant forecast begins with the step 72 of generating an initial performance curve
  • the initial applicant performance curve provides applicant energy as a function of time Energy, in this context, is used figuratively, not literally
  • the initial and modified applicant performance curves may be linear or non-lmear, and further, may be continuous or discontinuous
  • the initial applicant performance curve is assumed to be a linear continuous curve having an initial energy and an initial decay rate or slope
  • the initial energy and the initial decay rate may be selected based on applicant attributes at the time the application is filed
  • An applicant with a large number of positive attnbutes may have a high initial energy and a slow decay rate
  • an applicant with a large number of negative attnbutes may have a low initial energy and a fast decay rate
  • the next step 74 is to apply a senes of discreet toys to the performance curve
  • the senes of discreet scenarios are applied to the performance curve at points in time based on the probability of occurrence of each discreet scenario at each respective point in time.
  • the application of a scenario at a given point m time may determined using a nearest neighbor model which is based m part on the probability of occurrence of the strigno
  • the nearest neighbor model may be descnbed as the process of creating a multi-dimensional matrix or hyper-cube or non-directional graphs to find the most probable anangement and timing of dinosaurnos
  • the edges of the graphs contain characteristics such as delay from previous neighbor, confidence level as well as any other optimization goals that the financial institution desires
  • the energy value or the slope of the curve changes at each point in an amount corresponding to the magnitude and direction of the strigno applied at each point.
  • the result is a modified performance curve comprising a forecast of performance 76
  • the generation of the performance curve and the application of the strignos thereto are discussed in more detail with reference to Figures 7A - 7F It is preferable to obtain more than one forecast of performance m order to demonstrate circumstances under which the applicant will succeed and fail Accordingly, if the decision 78 is rendered to generate an additional applicant forecast, the scenario selection is modified 73 and the new selection of plausiblenos is reapphed 74 to generate another forecast of performance 76
  • Modification of the strigno selection may be performed to generate an optimistic forecast and a pessimistic forecast Alternatively, the modification of the scenario selection may be performed to obtain an optimistic forecast, a pessimistic forecast, and a neutral forecast In some circumstances, however, it may not be possible to obtain an optimistic forecast or a neutral forecast if the applicant has a large number of negative attributes (I
  • the probability of occunence of discreet scenarios having a positive influence or direction may be increased Similarly, the magnitude ot each scenario having a positive influence or direction may be increased Alternativ ely, the timing mav be changed such that the positive scenarios occur earlier in time and the negative scenarios occur later in time
  • FIG. 5 shows a flow chart illustrating in detail the step 1 10 of generating applicant scenarios foi use m the method 60 illustrated in Figure 3
  • the process of generating applicant strignos begins with the step 1 12 of sampling applicant data 1 12
  • the applicant data 102 is sampled for changes in applicant attributes as a function of time to generate applicant sequences 1 14
  • the applicant sequences 1 14 represent changes in the applicant att ⁇ butes as a function of time
  • Each sequence withm the set of applicant sequences 1 14 is then characte ⁇ zed 1 16 as a positive, negative, or neutral influence
  • the sequence characterization is based on the effect that each sequence is perceived to have on the economic welfare of the particular applicant For example, a change in employment status wherein the applicant is laid-off will have a negative influence.
  • a change m employment wherein the applicant is promoted with a pay raise will have a positive influence
  • each change in an applicant attribute is charactenzed according to its economic impact on the applicant
  • the next step 1 18 is to calculate a probability of occurrence for each sequence
  • the probability of occurrence may be calculated by taking the ratio of the number of times the sequence was part of the group of loans that showed the desired outcome and the total number of applicants in this group Flence in a group of 10,000 loans, if a specific scenario existed 9000 times the probability is 0 9
  • the next step 120 is to calculate an intensity for each sequence
  • the intensity corresponds to the amount of influence the particular sequence will have on the economic welfare of the applicant
  • a sequence such as a change in mantal status may have a significant effect on the economic welfare of the applicant, and therefore have a relatively e a large mtensitv
  • a sequence such as a small change in net fraction revolving burden (NFRB) may have little effect on the economic condition of the applicant, and therefore have a relatively low intensity
  • the intensity for each sequence may be calculated by establishing a gradation among outcomes and then representing the change on a fixed scale throughout the forecast process Intensity is a path dependent quantity Within the scope of an applicant's past history, intensity is calculated by grading the vanous positive and negative changes internal to the applicant The sequences that are related to these changes are then assigned a relative intensity depending on the seventy of the change
  • Each financial institution's risk management requirements determine the gradation of the outcomes Usually the profitability function of an institution is used to grade the vanous
  • the intensity, influence, and probability of occunence of each applicant sequence corresponds to the magnitude, direction, and probability of occunence of each applicant scenario such that a set of applicant narrativenos 122 may be generated from the set of applicant sequences 1 14 This completes the step of generating applicant scenarios 1 10
  • FIG. 6 shows a flow chart illustrating in detail the step 130 of generating population narrativenos for use in the method 60 shown in Figure 3
  • the process for generating population strignos begins with the step 132 of sampling the population data 104 Specifically, the population data 104 is sampled for changes in the population attributes as a function of time to generate population sequences 134.
  • the population sequences 134 represent changes in the population attributes as a function of time.
  • the population sequences are then conelated 136 to successful loans and failed loans.
  • the step 136 of co ⁇ elating the population sequences generates population sequences having a positive influence or a negative influence conesponding to successful loans or failed loans, respectively.
  • the next step 138 is to calculate the probability of each population sequence.
  • the probability of occunence may be calculated as discussed previously with regard to applicant sequences but utilizing different data sets as discussed herein.
  • the next step 140 is to calculate an intensity for each population sequence.
  • the intensity may be calculated as discussed previously with regard to applicant sequences but utilizing different data sets as discussed herein.
  • the probability of occunence for each population sequence represents the likelihood of the particular sequence occurring at discreet points in time.
  • the intensity for each population sequence correlates to the amount of influence the population sequence has on the economic welfare of the individuals within the population.
  • the next step 141 is to sample the population sequences for common patterns to generate stable population sequences 142.
  • the stable population sequences represent sequences which occur on a regular basis within the population.
  • the sampling process 141 eliminates sequences which are unique to particular individuals within the population and are thus relatively uncommon sequences not applicable to other individuals within the population.
  • the stable population sequences are then classified 143 to generate classes of stable population sequences 144 having class attributes.
  • the stable population sequences are classified based on association with class attributes of the population.
  • Class attributes comprise a collection of population attributes common to a group of individuals within the population. For example, a class attribute may comprise single males professionally employed for five years or less. This class attribute conesponds to a class of individuals within the population.
  • the class attributes are then matched 145 or otherwise conelated to the applicant attributes to generate matched population sequences 146 applicable to the particular applicant under consideration Specifically, the applicant under consideration may fall into one or more classes as defined by the class attnbutes that match the applicant's attnbutes.
  • the matched classes have conespondmg stable population sequences which may then be matched to the applicant to generate matched population sequences 146
  • the intensity, influence, and probability of occurrence of each matched population sequence conelates to the magnitude, direction, and probability of occunence of each population strigno 147 This completes the step of generating population strignos 130
  • Eq 1 Scenano ⁇ I d , ⁇ D S , P s , Eff s , ⁇ [(D S
  • I u is the user defined identification of the strigno (name or number),
  • P s is the probability of occunence of scenario
  • Eff s is the overall effect of strigno as a vector term, Ds, j is the delay when the predecessor strigno S M is used,
  • E, j is the effect when placed after S
  • P, j is the probability of placement after scenario (cumulative effect of decay accuracy)
  • G j is the goal represented by this combination
  • Equation 2 Equation 2
  • an applicant forecast 76 may be generated 70 as discussed with reference to Figures 4 and 7A-7F using the aging strip method described
  • An anchor scenario may be used as a starting scenario An anchor plausibleno is selected from individual or class strignos If the applicant's individual data produced a very high probability plausibleno ( I e , greater than a user defined threshold), then the highest probability scenario is used as the anchor strigno If not, the anchor scenario of the highest-class match is used as the anchor strigno Note that it is possible to have several anchors This simply means that several initial paths may be plotted on the agmg st ⁇ p
  • the entire set of strignos is searched to find all possible strignos that follow the anchor towards the user defined goal cntena (optimistic, pessimistic etc)
  • the best possible scenario is then selected to follow the anchor It is preferable to minimize the reduction in accuracy between alterations and have the highest probability of accuracy as the two mam cntena for selecting the best strigno
  • a user may define other criteria for best follower association
  • Figures 7A through 7F illustrate the process 70 of generating an applicant forecast by the applying a senes of discreet strignos to a finite timeline using a template called an agmg stnp (energy vs time graph)
  • Figures 7 A through 7F graphically illustrate the process 70 for generating an applicant forecast as discussed with reference to Figures 2 and 4
  • Figures 7A through 7C illustrate an optimistic forecast
  • Figures 7D through 7F illustrate a pessimistic forecast
  • This graphic illustration is for demonstrative purposes only and those skilled in the art will readily recognize that the curves, vectors, and manipulations thereof may be expressed m terms of one or more of algorithms
  • an initial applicant performance curve 210 is provided compnsmg applicant energy (E) as a function of time (t)
  • the initial applicant performance curve 210 may be expressed as a continuous linear or non- linear algonthm, or a discontinuous senes of linear oi non-lmear algonthms
  • the initial applicant performance curve 210 is shown as a continuous linear curve having an initial energy and an initial decay rate
  • the initial applicant performance curve 210 typically decays from its initial energy to zero energy in a finite time period, but may also increase or remain steady, depending on the att ⁇ butes of the particular applicant
  • the initial applicant performance curve 210 illustrated in Figure 7 A represents the performance of the applicant assuming no occunence of strignos, positive or negative, occur du ⁇ ng the time penod illustrated
  • Figure 7B illustrates the application of a series of discreet scenarios 212, 214,
  • the initial applicant performance curve 210 is shown in phantom Figure 7B illustrates the discreet dinosaurnos 212, 214, 216, and 218 as vectors placed at specific points on the time line based on the probability of occurrence of each plausibleno at each point in time
  • the magnitude of each scenario is reflected by the height of the vector, and the influence of each protagonist is reflected by the direction of the vector
  • Scenano 212 illustrates a strigno having a negative influence with a moderate magnitude
  • narrativeno 214 illustrates a Russianno having a positive influence with a large magnitude
  • narrativenos 216 and 218 illustrate dinosaurnos having a negative influence with a relatively small magnitude
  • An example of scenario 212 may be a change in residence wherein the applicant encounters an increase in monthly mortgage payments
  • An example of strigno 214 may be a change in employment status wherein the applicant receives a promotion and a conespondmg substantial increase in income
  • An example of strignos 216 and 218 include a sequential occunence of two increases in NFRB
  • the probability of an occunence, magnitude, and direction of each scenario 212, 214, 216, and 218 illustrated in Figure 7B are merely illustrative, as a plethora ot scenarios with various probabilities, magnitudes and directions are possible
  • the strignos may be applicant strignos, population scenarios, global scenarios, a combination thereof, or a
  • Figure 7C shows the discreet strignos 212, 214, 216, and 218 applied to the initial performance curve 210 (shown in phantom) resulting in a modified performance curve 230
  • the senes of discreet strignos 212 - 218 are applied to the initial performance curve 210 at points in time based on the probability of occunence of each discreet strigno at each point in time
  • the slope at each point in time changes in an amount conespondmg to the magnitude and direction of the discreet scenario applied thereto
  • the energy at each point in time may be changed in an amount conespondmg to the magnitude and direction of each discreet narrativeno applied thereto
  • a modified performance curve 230 is obtained.
  • the modified performance curve 230 compnses the forecast 76 of applicant performance refened to in Figure 4
  • the modified performance curve 230 remains positive for the duration of the financial service agreement, and thus is an optimistic forecast
  • Applying the discreet toys 212 - 218 to the initial performance curve 210 may be accomplished by simple vector addition or by other mathematical means.
  • a senes of partial differential equations representing the change at each discrete point may be solved to produce the net change in the energy of the applicant over time
  • the equations are either solved simultaneously to produce a single equation or the vectors of the vanous toysnos are added to produce a single net effect vector at the a point in time
  • the net effect vector then becomes a single partial differential equation that is applied sequentially It is possible to simply provide vector additions of all the strignos
  • the differential equation approach is more ⁇ ersatile and allows for better flexibility by allowing us to vary the rate of change in terms of the variables or events individually
  • FIGS 7D through 7F illustrate a pessimistic forecast of applicant performance
  • an initial applicant performance curve 210 is provided having an initial energy and an initial decav rate oi slope
  • the p ⁇ mary difference between the optimistic forecast illustrated m Figures 7A through 7C and the pessimistic forecast illustrated in Figures 7D through 7F is the order of occunence of the strignos 212 - 218 Specifically, as seen in Figure 7E, each of the scenarios having a negative influence, namely strignos 212, 216, and 218, occur early along the finite time line, and the only strigno having a positive influence, namely narrativeno 214, occurs late along the finite timeline
  • the result is illustrated in Figure 7F by the modified performance curve 220
  • the early occunence of the strignos having a negative influence 212, 216, and 218 results in a modified performance curve 220 reaching a zero energy level at a relatively early stage
  • Figure 8 shows a flow chart illustrating a computer implemented method 400 for generating data indicative of risk associated with a portfolio of financial service agreements in accordance with an exemplary embodiment of the present invention
  • the method 400 illustrated in Figure 8 includes the basic steps 320, 330 of generating of a series ot discrete narrativenos and generating a portfolio forecast based on the scenarios
  • the basic steps 320, 330 of method 400 generally correspond to the basic steps 60. 70 of method 200
  • the step 320 of generating may be accomplished by sampling global data 106 for changes in global attributes as a function of time to thereby generate global sequences Each global sequence has an intensity, an influence on the portfolio, and a probability of occunence, which correlate to a magnitude, a direction and a probability of occunence of each global strigno
  • These global sequences embody the overall framework or conditions under which the portfolio performance is to be evaluated
  • a portfolio forecast may be generated
  • the step of generating a portfolio forecast 330 is discussed m detail with reference to Figure 9
  • the generation of a portfolio forecast 330 begins by synchronizing 332 individual forecasts from the set of individual forecasts 360
  • the set of individual forecasts 360 comprises a forecast of performance 76 for each individual withm the portfolio
  • the set of individual forecasts 360 may therefore be generated by collecting a forecast of performance 76 for each individual in the portfolio utilizing the methods discussed with reference to Figures 2 through 7
  • the synchronization step 332 involves placing each forecast on the same time line in real time Once synchron
  • the global portfolio plausiblenos as generated by step 320 may then be selectively applied 336 to the initial portfolio forecast at points in time based on the probability of occunence of each discreet scenario at each point m time and based on the effect (if anv) of the global scenarios to strignos used m the individual forecasts
  • the global scenarios are selectively applied only to those individual forecasts that contain a strigno that will be influenced (e g , dampened, amplified) by the particular global strigno
  • the energy or slope at each point m time on the initial portfolio forecast changes in an amount conespondmg to the magnitude and direction of the discreet scenario applied thereto
  • the result is a modified portfolio performance curve
  • the modified portfolio performance curve is then checked 338 for stability Stability may be checked by assessing the convergence of the steady state conditions of vanous groups of loans That is, if the variance of the number of applicants in a group is minimal or below an acceptable enor rate, the group is considered stable In terms of counting statistics, the stability check is equivalent to the need to take several measurements to assess the reliability of the underlying process
  • the portfolio may be charactenzed or modified 350
  • Portfolio charactenzation 350 is preferably based on portfolio charactenstics du ⁇ ng steady state conditions within the portfolio, and more preferably, at a steady state condition of the entire portfolio, I e , the transient equilibrium point of the portfolio.
  • the identification of a steady state condition of the entire portfolio or a portion thereof is an important aspect of nsk management using the forecasting method described herein When several individual forecasts are synchronized and aged simultaneously, it is possible to get many steady states
  • the definition of a steady state is unique to each financial institution, but in general terms, it is when a high percentage ot individual forecasts withm the portfolio fall within some specified parameters of behavior Transient equihbnum points are large steady state events where overall portfolio information or charactenstics thereof may be reliably gathered
  • the method 400 tor generating data indicative of ⁇ sk associated with a portfolio of financial service agreements as in the present invention is an improvement over pnor art methods because factors influencing the performance of the portfolio, both internal and external, are applied over a finite timeline
  • the portfolio forecast allows the portfolio to be charactenzed and/or modified pnor to gaining substantial expenence with the portfolio
  • FIG. 10 illustrates a computer system 500 for generating data indicative of nsk associated with an application for a financial service by an applicant and or associated with a portfolio of financial service agreements
  • the computer system 500 mav be anv suitable data processing system including a processor 502, an input device 504, an output device 506, and a data storage means 508
  • the input device 504 genencally refers to any means for providing data to the processor such as a keyboaid or a telecommunications receiver
  • the output device 506 may comprise any suitable means to present the forecast 510 to the end user, and in particular the financial service institution
  • the output device 506 may compnse a display, a p ⁇ nter, or a telecommunications transmitter
  • the data storage means 508 may compnse any means for temporanly or indefinitely sto ⁇ ng data for use by the processor 502
  • the data storage means 508 may compnse RAM or a disk dnve Those skilled in the art will recognize that many suitable alternatives for
  • the processor 502 of the computer system 500 performs the majonty of the operations and steps discussed previously with regard to Figures 2 through 9 Specifically, the processor 502 provides means for generating a forecast, means for comparing the forecast to a risk threshold, means for generating a set of applicant, population, and/or global scenarios, means for generating an applicant or portfolio performance curve, means for modifying such a performance curve, means for sampling data, means for characterizing such data or subsets thereof, means for calculating probabilities associated with such data, means for calculating intensity associated with such data, means for correlating data, means for classifying data, means for matching data, etc
  • the output device 506 or an extension thereof provides means for accepting or rejecting an application and means for providing a forecast indicative of risk associated with a portfolio or a financial service application
  • the data storage means 508 provides means for stonng applicant, population, and global data indicative of attnbutes thereof, and means for storing derivative data thereof such as sequences, strignos, att ⁇ butes, matched va ⁇ able
  • the present invention provides a system 500 and method 200 for generating data indicative of risk associated with an application for financial service by an applicant in addition to a system 500 and method 400 for generating data indicative of risk associated with a portfolio of financial service agreements
  • the methods 200, 400 include and the system 500 implements the basic step of generating a forecast of performance of the applicant/portfolio wherein the forecast compnses a series of discreet narrativenos applied over a finite timeline, using a template referred to as an agmg st ⁇ p.
  • a forecast is indicative of ⁇ sk and is tremendously useful data.

Abstract

L'invention concerne un système et un procédé pour générer des données (Fig.1), telles qu'une prévision représentative du risque associé à l'application de services financiers et/ou d'un portefeuille d'accords financiers, à des fins d'appréciation et de gestion (30) des risques. Ledit procédé consiste à générer une prévision de résultat comprenant une série de scénarios différents appliqués sur une ligne temporelle finie, au moyen d'un modèle appelé bande chronologique. Ladite prévision indique le risque. Les système et procédé de l'invention constituent une amélioration sensible par rapport à ceux de la technique antérieure, en ce que le futur n'est pas le reflet du passé mais en ce que l'expérience historique est utilisée pour la production de scénarios et de probabilités associées pour la génération d'une prévision de résultat.
PCT/US2000/006186 1999-03-09 2000-03-09 Systeme de prevision financiere et procede d'appreciation et de gestion des risques WO2000054186A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US26571699A 1999-03-09 1999-03-09
US09/265,716 1999-03-09

Publications (2)

Publication Number Publication Date
WO2000054186A1 true WO2000054186A1 (fr) 2000-09-14
WO2000054186A8 WO2000054186A8 (fr) 2001-04-19

Family

ID=23011606

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2000/006186 WO2000054186A1 (fr) 1999-03-09 2000-03-09 Systeme de prevision financiere et procede d'appreciation et de gestion des risques

Country Status (1)

Country Link
WO (1) WO2000054186A1 (fr)

Cited By (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6546545B1 (en) 1998-03-05 2003-04-08 American Management Systems, Inc. Versioning in a rules based decision management system
US6601034B1 (en) 1998-03-05 2003-07-29 American Management Systems, Inc. Decision management system which is cross-function, cross-industry and cross-platform
US6609120B1 (en) 1998-03-05 2003-08-19 American Management Systems, Inc. Decision management system which automatically searches for strategy components in a strategy
US6708155B1 (en) 1999-07-07 2004-03-16 American Management Systems, Inc. Decision management system with automated strategy optimization
US7006994B1 (en) 1999-07-16 2006-02-28 American Management Systems, Inc. Automated receivables management system
US7318046B1 (en) 1998-03-05 2008-01-08 American Management Systems, Inc. Collector's account payment promise option advisory apparatus and method
US7319971B2 (en) * 2001-01-31 2008-01-15 Corprofit Systems Pty Ltd System for managing risk
US20100138354A1 (en) * 2001-03-20 2010-06-03 Goldman Sachs & Co Construction Industry Risk Management Clearinghouse
US7881995B2 (en) 2004-07-02 2011-02-01 Capital Tool Company Systems and methods for objective financing of assets
US8364578B1 (en) 1998-03-05 2013-01-29 Cgi Technologies And Solutions Inc. Simultaneous customer/account strategy execution in a decision management system
US8996481B2 (en) 2004-07-02 2015-03-31 Goldman, Sach & Co. Method, system, apparatus, program code and means for identifying and extracting information
US9058581B2 (en) 2004-07-02 2015-06-16 Goldman, Sachs & Co. Systems and methods for managing information associated with legal, compliance and regulatory risk
US9063985B2 (en) 2004-07-02 2015-06-23 Goldman, Sachs & Co. Method, system, apparatus, program code and means for determining a redundancy of information
US11055731B2 (en) 2013-04-08 2021-07-06 Oracle International Corporation Parallel processing historical data

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8121937B2 (en) 2001-03-20 2012-02-21 Goldman Sachs & Co. Gaming industry risk management clearinghouse
CN112700280A (zh) * 2020-12-31 2021-04-23 上海竞动科技有限公司 一种短期非连续用户行为评价方法及装置

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4975840A (en) * 1988-06-17 1990-12-04 Lincoln National Risk Management, Inc. Method and apparatus for evaluating a potentially insurable risk
US5611052A (en) * 1993-11-01 1997-03-11 The Golden 1 Credit Union Lender direct credit evaluation and loan processing system
US5797133A (en) * 1994-08-31 1998-08-18 Strategic Solutions Group, Inc Method for automatically determining the approval status of a potential borrower
US5809478A (en) * 1995-12-08 1998-09-15 Allstate Insurance Company Method for accessing and evaluating information for processing an application for insurance
US5812988A (en) * 1993-12-06 1998-09-22 Investments Analytic, Inc. Method and system for jointly estimating cash flows, simulated returns, risk measures and present values for a plurality of assets
US5970479A (en) * 1992-05-29 1999-10-19 Swychco Infrastructure Services Pty. Ltd. Methods and apparatus relating to the formulation and trading of risk management contracts
US6078901A (en) * 1997-04-03 2000-06-20 Ching; Hugh Quantitative supply and demand model based on infinite spreadsheet

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4975840A (en) * 1988-06-17 1990-12-04 Lincoln National Risk Management, Inc. Method and apparatus for evaluating a potentially insurable risk
US5970479A (en) * 1992-05-29 1999-10-19 Swychco Infrastructure Services Pty. Ltd. Methods and apparatus relating to the formulation and trading of risk management contracts
US5611052A (en) * 1993-11-01 1997-03-11 The Golden 1 Credit Union Lender direct credit evaluation and loan processing system
US5812988A (en) * 1993-12-06 1998-09-22 Investments Analytic, Inc. Method and system for jointly estimating cash flows, simulated returns, risk measures and present values for a plurality of assets
US5797133A (en) * 1994-08-31 1998-08-18 Strategic Solutions Group, Inc Method for automatically determining the approval status of a potential borrower
US5809478A (en) * 1995-12-08 1998-09-15 Allstate Insurance Company Method for accessing and evaluating information for processing an application for insurance
US6078901A (en) * 1997-04-03 2000-06-20 Ching; Hugh Quantitative supply and demand model based on infinite spreadsheet

Cited By (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8364578B1 (en) 1998-03-05 2013-01-29 Cgi Technologies And Solutions Inc. Simultaneous customer/account strategy execution in a decision management system
US6601034B1 (en) 1998-03-05 2003-07-29 American Management Systems, Inc. Decision management system which is cross-function, cross-industry and cross-platform
US6609120B1 (en) 1998-03-05 2003-08-19 American Management Systems, Inc. Decision management system which automatically searches for strategy components in a strategy
US6546545B1 (en) 1998-03-05 2003-04-08 American Management Systems, Inc. Versioning in a rules based decision management system
US7062757B2 (en) 1998-03-05 2006-06-13 American Management Systems, Inc. Decision management system which is cross-function, cross-industry and cross-platform
US7318046B1 (en) 1998-03-05 2008-01-08 American Management Systems, Inc. Collector's account payment promise option advisory apparatus and method
US7318224B2 (en) 1998-03-05 2008-01-08 American Management Systems, Inc. Versioning in a rules based decision management system
US6708155B1 (en) 1999-07-07 2004-03-16 American Management Systems, Inc. Decision management system with automated strategy optimization
US7006994B1 (en) 1999-07-16 2006-02-28 American Management Systems, Inc. Automated receivables management system
US7319971B2 (en) * 2001-01-31 2008-01-15 Corprofit Systems Pty Ltd System for managing risk
US20100138354A1 (en) * 2001-03-20 2010-06-03 Goldman Sachs & Co Construction Industry Risk Management Clearinghouse
US8285615B2 (en) * 2001-03-20 2012-10-09 Goldman, Sachs & Co. Construction industry risk management clearinghouse
US7881995B2 (en) 2004-07-02 2011-02-01 Capital Tool Company Systems and methods for objective financing of assets
US8996481B2 (en) 2004-07-02 2015-03-31 Goldman, Sach & Co. Method, system, apparatus, program code and means for identifying and extracting information
US9058581B2 (en) 2004-07-02 2015-06-16 Goldman, Sachs & Co. Systems and methods for managing information associated with legal, compliance and regulatory risk
US9063985B2 (en) 2004-07-02 2015-06-23 Goldman, Sachs & Co. Method, system, apparatus, program code and means for determining a redundancy of information
US11055731B2 (en) 2013-04-08 2021-07-06 Oracle International Corporation Parallel processing historical data

Also Published As

Publication number Publication date
WO2000054186A8 (fr) 2001-04-19

Similar Documents

Publication Publication Date Title
US7813981B2 (en) Apparatus and method for simulating an analytic value chain
US8204813B2 (en) System, method and framework for generating scenarios
US8200557B2 (en) Systems and methods for providing migration and performance matrices
US7783551B1 (en) Method and system for simulating risk factors in parametric models using risk neutral historical bootstrapping
US7890407B2 (en) System and method for estimating conduit liquidity requirements in asset backed commercial paper
WO2000054186A1 (fr) Systeme de prevision financiere et procede d'appreciation et de gestion des risques
US6968326B2 (en) System and method for representing and incorporating available information into uncertainty-based forecasts
CN110009417B (zh) 目标客户筛选方法、装置、设备及计算机可读存储介质
US20030225659A1 (en) Retail lending risk related scenario generation
Leung Estimating maintenance effort by analogy
CN115203496A (zh) 基于大数据的项目智能预测及评估方法、系统及可读存储介质
US7966254B2 (en) Method and system for credit decisioning using activity based costing and compartmental modeling
Langarudi et al. Resurrecting a forgotten model: Updating Mashayekhi’s model of Iranian Economic Development
Saha A real options perspective to enterprise architecture as an investment activity
CN110245985A (zh) 一种信息处理方法及装置
CN114297052A (zh) 测试数据生成方法及装置
JP7298286B2 (ja) モデル提供プログラム、モデル提供方法及びモデル提供装置
JP4723758B2 (ja) 不動産担保貸出引当額計算システム、不動産担保貸出引当額計算プログラム及びそのプログラムを記録した記録媒体
JP2004094662A (ja) 信用リスク管理の最適化モデル適用方法及び装置
Pagan et al. Econometric analysis and prediction of recurrent events
KR20240066782A (ko) 빅데이터와 기계학습을 이용한 대출 연체 정보 예측 시스템 및 그 방법
CN115271934A (zh) 信贷客户的多层次评价方法、装置、设备及存储介质
CN114742126A (zh) 一种样本扩充方法、装置、设备及可读存储介质
CN116128185A (zh) 资产配置方法、系统及存储介质
CN113240513A (zh) 一种用户授信额度的确定方法和相关装置

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A1

Designated state(s): CA JP US

AL Designated countries for regional patents

Kind code of ref document: A1

Designated state(s): AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE

121 Ep: the epo has been informed by wipo that ep was designated in this application
DFPE Request for preliminary examination filed prior to expiration of 19th month from priority date (pct application filed before 20040101)
AK Designated states

Kind code of ref document: C1

Designated state(s): CA JP US

AL Designated countries for regional patents

Kind code of ref document: C1

Designated state(s): AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE

CFP Corrected version of a pamphlet front page
CR1 Correction of entry in section i
122 Ep: pct application non-entry in european phase