`(12) Patent Application Publication
`Keyes
`
`(10) Pub. No. : US 2001/0044766 A1
`Nov. 22, 2001
`(43) Pub. Date:
`
`IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII
`US 20010044766A1
`
`(54) METHODS AND SYSTEMS FOR MODELING
`USING CLASSIFICATION AND
`REGRESSION TREES
`
`(76)
`
`Inventor:
`
`Tim K. Keyes, West Redding, CT (US)
`
`Correspondence Address:
`John S. Beulick
`Armstrong Teasdale LLP
`One Metropolitan Sq. , Suite 2600
`St. Louis, MO 63102 (US)
`
`(21) Appl. No. :
`
`09/746, 411
`
`(22) Filed:
`
`Dec. 21, 2000
`
`Related U. S. Application Data
`
`of provisional
`application No.
`(63) Non-provisional
`60/174, 057, filed on Dec. 30, 1999.
`
`Publication Classification
`
`(51) Int. Cl.
`(52) U. S. Cl.
`
`G06F 17/60
`705/36; 705/7
`
`ABSTRACT
`
`(57)
`A method of valuation of large groups of assets using
`trees is described. The method
`classification and regression
`assess-
`relevant portfolio segmentations,
`includes defining
`ing performance of the classification
`and regression
`tree
`based model against a simple model and ranking all port-
`folio segments based upon performance of the models.
`Iterative and adaptive statistical evaluation of all assets and
`statistical inferences are used to generate
`the segmentations.
`The assets are collected into a database, grouped by credit
`variable, subdivided by ratings as to those variables and then
`individually. The assets are
`rated
`then
`regrouped
`and a
`is established by cumulating
`collective valuation
`individual
`valuations.
`
`2i2
`
`DEFINE
`
`~ 216
`
`UW
`SAMPLE
`
`218
`
`CHECK
`FOR
`ATTRIB.
`COMB,
`
`220
`
`222
`
`SFT
`ATTRIB,
`
`CLASSIFY
`
`EXPERT
`OPINION
`
`214
`
`224 g
`
`VALUE
`CLUSTER
`
`226
`
`DESEG,
`INTO
`LOANS
`BY
`RULE
`
`228
`
`TABLE
`
`TRULIA - EXHIBIT 1014
`
`
`
`Patent Application Publication Nov. 22, 2001 Sheet 1 of 14
`
`US 2001/0044766 A1
`
`12
`
`p10
`
`PORTFOLIO
`
`UNDERWRITE UNTIL BID
`
`} — 18
`
`UNDER-
`WRITTEN
`PORTION
`
`20
`
`26
`
`BID
`
`24
`
`22
`
`PRIOR ART
`
`
`
`Patent Application Publication Nov. 22, 2001 Sheet 2 of 14
`
`US 2001/0044766 A1
`
`28
`
`DD TIMELINE
`
`BEGIN
`
`12
`
`|4
`40
`
`38
`
`44
`178
`
`82
`
`PORTFOLIO
`
`170
`172
`
`76
`
`34 I
`30
`174 /-78
`80
`EXTRAPOLATION
`IMPROVES
`
`BID
`
`~ 658 60 62
`50 46
`4 8
`
`64
`30
`
`84
`
`72
`
`SAMPLED
`FULLY
`UNDER-
`WRITTEN
`
`32
`
`"BEST" VALUE
`CONTINUED
`IMPROVE
`TO
`
`48
`
`52
`
`66
`
`46
`50
`
`54
`
`58 60
`
`62
`
`64
`
`
`
`Patent Application Publication Nov. 22, 2001 Sheet 3 of 14
`
`US 2001/0044766 A1
`
`206
`
`208
`p
`
`134
`
`136
`
`SUPERVISED
`LEARNING
`PROCESS
`
`UNSUPERVISED
`LEARNING
`PROCESS
`
`UPLOAD
`SOFTWARE
`
`UNDERWRITING
`CLUSTERS
`TABLE
`
`40
`
`138
`
`120
`
`122
`
`BETA ADJ
`FOR CREDIT
`SCORE
`
`ADJ CREDIT
`ANALYST
`
`TABLE
`
`126
`
`SAMPLE
`1QQ( A
`RANDOM B&C
`
`LOAN LEVEL
`RE-UNDERWRITE
`PROCESS
`
`MANUAL
`DATA
`ENTRY
`
`CREDIT
`ANALYST
`TABLE
`
`UNTOUCHED
`
`LOAN
`TABLE
`
`144
`
`142
`
`INFERRED?
`
`132
`
`CREDIT?
`
`SUPER A?
`
`PARTIAL
`CASH?
`
`98
`
`100%%u
`CASH?
`
`PORTFOLIO
`
`108
`
`ASSET CLASS
`BETA ADJUST
`
`'4 ~ «0
`
`128
`
`100%%uo
`
`M SAMPLE
`
`GROUP LEVEL
`UNDERWRITE
`
`130
`
`B ADJUST
`CREDIT ANALYST
`TABLE
`
`«4
`
`RULE SET TO
`DESEGREGATE
`TO LOAN LEVEL
`
`ELECTRONIC
`
`UPLOAD TO
`SUPER A TABLE
`
`116
`
`«8
`
`92
`
`PARTIAL
`VALUE
`ASSETS
`
`100
`
`102
`
`TEAM
`AUTHENTICATION
`
`PARTIAL
`CASH TABLE
`
`14
`
`90
`
`94
`
`96
`
`86
`
`FULL
`
`VALUE
`LOANS
`
`TEAM
`AUTHENTICATION
`
`100%%u CASH
`
`TABLE
`
`85
`
`168 g
`
`TO FIG. 4
`
`
`
`Patent Application Publication Nov. 22, 2001 Sheet 4 of 14
`
`US 2001/0044766 A1
`
`FROM FIG, 3
`
`148
`
`i52 ~ l68
`
`DETERMINISTIC
`CASH FLOW
`BRIDGE
`
`STOCHASTIC
`CASH FLOW
`BRIDGE
`
`CASH FLOW ~ i50
`
`LOAN LEVEL
`VALUATION
`
`146
`
`TIMING TABLE
`
`i54
`
`TRANCHE
`BID PRICE
`
`156
`
`TRANCHE
`IRR, TTP
`
`MODEL
`
`iSO
`
`MEAN
`
`IRR
`
`NPV&0
`
`GE PURSE
`PREFERENCES
`TRANCHE PRIORTY
`
`BID OPENING
`SIMULATION
`
`OTHER BIDDER
`PREFERNCES
`
`OTHER BIDDER
`PURSES
`
`BID PROCESS
`RULE SET
`
`i58
`
`PARTNERSHIP
`FINANCIAL
`PRO FORMA
`
`MAX EXPECTED
`IRR SUBJECT
`TO NPV&0
`
`IRR
`MEAN
`&30?
`
`SENIOR MGT,
`SETS BID
`
`BID FORMS
`& BID
`
`PARTNER
`ROUND TABLE
`BID PRICE
`
`
`
`Patent Application Publication Nov. 22, 2001 Sheet 5 of 14
`
`US 2001/0044766 A1
`
`100%%uo
`
`194
`
`182
`
`196
`
`204
`
`200
`
`198
`
`~ 180
`
`202
`
`188
`
`186
`
`192
`
`1, 0
`
`190
`
`184
`
`
`
`Patent Application Publication Nov. 22, 2001 Sheet 6 of 14
`
`US 2001/0044766 A1
`
`212
`
`DEFINE
`
`~ 216
`
`UW
`SAMPLE
`
`218
`
`CHECK
`FOR
`ATTRIB.
`COMB,
`
`220
`
`222
`
`SET
`ATTRIB,
`
`CLASSIFY
`
`EXPERT
`OPINION
`
`214
`
`224 g
`
`VALUE
`CLUSTER
`
`226
`
`DESEG.
`INTO
`LOANS
`BY
`RULE
`
`228
`
`TABLE
`
`
`
`Patent Application Publication Nov. 22, 2001 Sheet 7 of 14
`
`230
`
`232
`
`US 2001/0044766 A1 ~ 208
`
`LOAN
`DATA
`
`DATA
`ACQUISITION
`
`VARIABLE
`SELECTION
`
`HIERARCHICAL
`SEGMENTATION
`
`78
`
`236
`
`FCM
`
`238
`
`138
`
`UNDER-WRITING
`REVIEW
`
`CASH FLOW/
`RISK SCORES
`
`
`
`Patent Application Publication Nov. 22, 2001 Sheet 8 of 14
`
`US 2001/0044766 A1
`
`240
`
`SAMPLE
`ASSETS
`
`242
`
`MANUAL
`
`UW
`
`FORM
`CLUSTERS
`
`246
`
`BUILD
`MODELS
`
`248
`
`SEI ECT
`BEST
`MODELS
`
`CALCULATE
`COUNTS
`
`250
`
`252
`
`APPLY
`MODELS
`
`254
`
`
`
`Patent Application Publication Nov. 22, 2001 Sheet 9 of 14
`
`US 2001/0044766 A1
`
`z
`
`I—
`I B
`~O
`
`W
`
`zg &nO m
`
`OL
`
`I—
`
`LIJ O O
`
`O
`
`LIJ
`
`I— I
`W O z D
`
`O
`Q W ~O
`W OI
`CL W ~)-I—
`y ~W
`CL O~~
`ov«
`~ZCLD
`
`H
`
`V)
`O
`O O
`W z
`OI ——
`O&- DO
`W ~ O — O ~
`~m O~ W ~
`W — D g
`0 ~ O
`
`W
`
`~&- z
`OzO MI-
`O—
`cC WI — O&
`O CL
`OOW ~
`WZ&~
`~ ~O
`
`z R OI—
`~ C/) ze
`
`O CL
`z: X
`~cn
`
`W
`
`LIJ R
`I—
`Vl cCWZ
`I — Ve
`z Ioo
`5 CL W &
`W OOI — V
`OZOw
`I-OZZ
`
`
`
`Patent Application Publication Nov. 22, 2001 Sheet 10 of 14
`
`US 2001/0044766 A1
`
`0 (n w
`
`CD
`
`CL o~ Q
`
`0
`
`Q 0 I—
`
`CA
`
`V)
`LIJ
`
`z
`
`0
`
`W
`
`0-
`O
`w
`I R
`
`zo 0&
`
`ZI—
`
`LLI N=
`
`0 W 0
`ww
`CL O wo
`
`M
`LL
`
`go
`CL I—
`«C Z ow
`O w ww
`z o&
`CL OC-I
`I — W CL
`go
`
`X 0 cn
`I — 0
`U) CD
`W ~
`w
`
`w
`U)
`
`0 PL
`R ~0 ~
`I—
`I—
`O Z
`
`CL
`
`W I—
`I—
`
`I—
`
`I—
`«C Q
`
`V) Z w 00 — C5
`LLI 0 0
`CD CL W K Z 0 CQ
`Q w
`II 0 Z
`CL 0
`Q OI-
`CD «C O
`OQ CL-
`O A
`
`II
`
`0
`0 0
`
`W
`w w
`0 «C
`0
`
`0 CL
`
`«C
`W CL
`
`Q o D Z w
`Z
`CL og R o~ «C
`0 CD
`
`Z 0
`I—
`O
`
`C5
`
`0cn
`O CL ~
`
`LLI 6 LIJ
`~I — O
`«C — Z
`CL Z w
`I DCL
`EJJ
`CD(D LL
`Z:ZZ
`g&o
`CL CL LL
`«C «C RAW
`CL
`III N &-
`
`0
`0
`CL 0 CL
`
`O
`
`w
`
`0
`
`C
`
`w+
`
`l/) 0
`CD I— ~ U)
`N W
`M
`
`w «C Z
`0-
`O w
`
`LJJ
`
`CL w
`Cf)
`
`o 0O ) Z
`
`LIJ
`CA
`
`
`
`Patent Application Publication Nov. 22, 2001 Sheet 11 of 14
`
`US 2001/0044766 A1
`
`-J W
`CI O
`
`X X
`
`X X X
`
`X X
`
`X X X X
`
`X X
`
`CI 0
`
`w
`CI 0
`
`LIJ
`
`CI 0 X X XXX
`
`X X
`
`XX X
`
`X X
`
`IJJ
`
`CI 0
`
`X X X XX
`
`X X
`
`4J
`
`LIJ
`CI
`
`N
`
`I/I
`«C O
`Q Q
`CL
`
`CL'
`
`LLJ
`
`I
`
`I
`
`W z N
`~ w OQ
`O
`w w
`I — IJJ
`«C O O CL
`— O&(cid:30)-
`0-
`IWQQ) — Z
`W
`Q Q P Q I — 0
`«CQ
`Q
`«C» r WI
`Q «C
`OOZ Z OO
`+~ww CLO
`mocDcD Q — I
`
`0
`
`4J
`
`4J
`m
`
`CL )
`I/) WK wo mQ
`«C CL 0 — WW
`
`Q
`
`OZ~
`
`Q
`
`CQ I — I
`I — «C
`4J W~
`CD 0
`WI
`I/I I/I O
`
`0 (~
`
`m
`(/)
`
`LLJ
`
`CI 0
`
`CD
`
`IJJ
`
`Z J
`LIJ O oo
`I/I Q z ~ o o Q
`O
`~ ~& I/I
`w W4J
`0 I/I CL C0 o
`CJ Cf) CL
`w gZ
`Z
`4- W ) — Z «C
`(/I OQ
`I — 0 — CL W
`0 OI 0
`I — CI I/I
`Z WI—
`~ KZ
`P& CL CD
`O I — LLJ
`I/I If) +
`4J
`m (/I
`W
`~ CIZ
`Lf) I—
`O O& 2 CL
`~ CIW
`CD
`O OI-
`4J
`«CQW
`W CLCI
`
`Q
`
`U)
`
`I/I ~
`2 0 O Q O CD
`
`CL CL
`
`LLJ
`
`C5 o
`
`«C
`
`I—
`
`«C ~ O
`I — ~ 0 O
`
`LJJ
`
`I — Z OW
`I — CI
`I/I 4J
`
`CL
`
`I—
`
`& W 0 O
`I — Cn 4J
`
`CL
`
`0
`I—
`
`«C
`I—
`
`CL o o
`
`Q 4J
`0 O
`0
`I— O
`
`CL 0
`
`
`
`Patent Application Publication Nov. 22, 2001 Sheet 12 of 14
`
`US 2001/0044766 A1
`
`LJ
`V)
`
`LJ
`V)
`LIJ
`
`VI
`
`tD
`
`(3
`LJ
`
`C3
`LJ
`
`O O
`
`O I—
`
`II
`LJ
`VI
`
`LJJ
`
`II
`LJ
`C/l
`
`LIJ
`
`II
`(/l
`LIJ 0-
`
`ll
`
`LIJ 0-
`
`) «C
`
`O
`
`O
`
`0
`
`LJ
`V)
`LJ
`
`CD )
`C) )
`
`CD )
`O ) LJ
`
`I
`
`O
`
`LJ
`V)
`LJ
`
`11
`
`IJJ
`0-
`CL
`
`O
`
`LIJ
`LJ
`
`CD
`V)
`
`CD
`
`C5 O
`
`LIJ
`
`LJ
`CD
`
`IJJ
`
`)
`
`O
`LJ
`
`O
`
`LJ
`CA
`LJ
`
`LJ 0-
`
`O
`
`V)
`LJ
`
`IJJ
`V)
`LJJ
`
`O
`I—
`U) O
`
`CL
`
`LJ
`
`O
`O I—
`
`CO
`LIJ
`
`O
`
`O
`
`N
`
`H
`
`O
`4J
`CL
`
`O
`
`VI
`LJ 0-
`
`(/)
`LJ 0- O
`
`4J
`CD
`
`) C3
`
`LJ
`CL
`
`LJ
`V)
`
`I— z:
`
`LIJ
`V)
`I—
`
`LJ
`
`I— z: LJ
`
`CL
`
`LJ
`
`LIJ
`CL 0-
`I—
`
`LIJ
`
`Cl
`0-
`I—
`
`Ck'
`
`«C
`
`O
`
`C)
`I—
`
`O
`
`O
`
`I—
`O
`
`CD
`
`CL O
`
`V)
`
`LJ
`Q O
`
`O
`
`LJ
`
`O
`
`CL
`«C
`
`CD
`
`O
`
`O
`I—
`il) O
`
`CL
`
`LJ
`
`O O
`
`LIJ
`
`I—
`
`CD
`
`CD
`
`CL O
`LJ
`
`CD
`
`C5
`
`CL
`
`CL
`I—
`
`LJ
`CL 4J
`
`O I—
`I— z: 4J
`
`I—
`
`LIJ
`
`«C
`CL
`
`O
`
`
`
`Patent Application Publication Nov. 22, 2001 Sheet 13 of 14
`
`US 2001/0044766 A1
`
`260
`
`LOAN
`SECURED?
`
`NO
`
`YES
`
`LOAN
`TYPE?
`
`NON-
`REVOLVING
`
`REVOLVING
`
`SHAKER?
`TREE ii
`
`SHAKER?
`TREE i6
`
`272
`
`262
`
`LOAN
`TYPE?
`
`NON-
`REVOLVING
`
`LAST
`PAYMENT?
`
`REVOLVING
`
`LAST
`PAYMENT?
`
`NON-
`ZERO
`
`ZERO
`
`NON-
`ZERO
`
`ZERO
`
`SHAKER 2
`TREE 9
`
`SHAKER7
`TREE io
`
`264
`
`266
`
`SHAKER'P
`TREE i3
`268 i
`
`SHAKER'?
`TREE 14
`
`270
`
`
`
`Patent Application Publication Nov. 22, 2001 Sheet 14 of 14
`
`US 2001/0044766 A1
`
`p76
`
`DATABASE
`
`304
`
`COMPUTER
`
`306
`
`DATABASE
`SERVER
`
`304
`
`COMPUTER
`
`300
`
`
`
`US 2001/0044766 A1
`
`Nov. 22, 2001
`
`METHODS AND SYSTEMS FOR MODELING
`USING CLASSIFICATION AND REGRESSION
`TREES
`
`CROSS REFERENCE TO RELATED
`APPLICATIONS
`[0001] This application claims the benefit of U. S. Provi-
`sional Application No. 60/174, 057, filed Dec. 30, 1999,
`incorporated by reference
`which is hereby
`in its entirety.
`
`BACKGROUND OF THE INVENTION
`
`[0002] This invention
`to valuation meth-
`relates generally
`ods for financial
`to rapid
`and more particularly
`instruments
`valuation of large numbers of financial
`instruments.
`[0003] A large number of assets such as loans, e. g. , ten
`sometimes
`loans or other financial
`instruments,
`thousand
`become available
`for sale due to economic conditions,
`the
`planned or unplanned divestiture of assets or as the result of
`legal remedies. The sale of thousands of commercial
`loans
`or other
`sometimes
`financial
`involving
`instruments
`the
`equivalent of billions of dollars
`in assets must sometimes
`occur within a few months. Of course, the seller of assets
`the value of the portfolio,
`and will
`wants
`to optimize
`in "tranches. " The
`assets
`sometimes
`term
`the
`group
`"tranche" as used herein is not limited
`to foreign notes but
`assets and
`also
`includes
`financial
`instrument
`groupings
`regardless of country or jurisdiction.
`[0004] Bidders may submit bids on all tranches, or on only
`some tranches. In order to win a tranche, a bidder typically
`the highest bid for that tranche. In connection
`must submit
`to submit on a particular
`a bid amount
`with determining
`to evaluate
`tranche, a bidder often will engage underwriters
`loans as possible within a tranche and within
`as many
`the
`limited time. Up until the time for submitting a bid
`available
`the loans under-
`is about to expire, the bidder will evaluate
`to extrapolate a value
`written at that time, and then attempt
`then been analyzed by
`loans
`that have not
`to
`the
`the
`underwriters.
`[0005] As a result of this process, a bidder may signifi-
`a tranche and submit a bid that is not
`cantly undervalue
`competitive or bid higher
`value and
`the underwritten
`than
`risk. Of course, since the objective is to
`assume unquantified
`win each tranche at a price that enables a bidder to earn a
`losing a tranche due to significant undervaluation of
`return,
`It would be desir-
`the tranche represents a lost opportunity.
`able to provide a system that facilitates accurate valuation of
`a large number of financial
`in a short period of
`instruments
`the associated probabilities of return for
`time and understand
`a given bid.
`
`BRIEF SUMMARY OF THE INVENTION
`
`[0006]
`In an exemplary
`embodiment,
`iterative
`and
`an
`is provided wherein a portfolio is divided
`adaptive approach
`into three major valuations. Full underwriting of a first type
`of valuation of an asset portfolio is performed based upon an
`sample. A second valuation
`is elficiently
`adverse
`type
`from categories of common descriptive attributes,
`sampled
`in the selective
`the assets
`random
`sample are fully
`and
`to sta-
`type is subjected
`underwritten. The third valuation
`tistically
`values and
`inferred valuation using underwriting
`variances of the first and second portions
`and applying
`
`value each asset in the
`statistical
`to individually
`inference
`and data reduction
`are used
`third portion. Clustering
`in
`the third portion.
`valuing
`[0007] As the process proceeds and more assets are under-
`the number of assets in the first and second portions
`written,
`the number of assets
`increase
`in the
`third portion
`and
`decreases and the variance of the valuation of the assets in
`the third portion becomes more and more defined. More
`the assets in the third portion are evaluated by
`specifically,
`into clusters based on similarity
`the assets
`to
`grouping
`valuations of assets in the first and second portions. Hypo-
`thetical bids are generated using the valuations
`to determine
`an optimum bid within parameters determined by the bidder.
`bid is identified
`The optimum
`iterative bid
`an
`through
`generation process.
`[0008] One method for grouping assets based on similarity
`tree analysis of asset
`uses a classification
`and regression
`the steps of defining
`portfolios, where
`includes
`the method
`assessing performance of
`relevant portfolio segmentations,
`the classification and regression
`tree based model against a
`simple model and ranking all portfolio segments based upon
`performance of the models.
`
`for
`
`BRIEF DESCRIPTION OF THE DRAWINGS
`[0009] FIG. 1 is a flow diagram
`a known
`illustrating
`process for valuing a portfolio of assets;
`[0010] FIG. 2 is a flow diagram
`illustrating
`a
`valuing
`portfolio of assets in accordance with one embodiment of
`invention;
`the present
`[0011] FIG. 3 is a flow diagram
`in more detail,
`illustrating,
`one embodiment of a first portion of a rapid valuation
`process for large asset portfolios
`that breaks assets
`into
`categories of variance;
`[0012] FIG. 4 is a flow diagram
`a second
`illustrating
`portion of a rapid valuation process for a large asset port-
`folios that aggregates
`from a basis to a tranche or portfolio
`basis;
`[0013] FIG. 5 illustrates
`distribution
`a probability
`exemplary assets whose recovery value is inferred;
`[0014] FIG. 6 is a flow diagram of a supervised
`step of the process of FIG. 3;
`[0015] FIG. 7 is a flow diagram of an unsupervised
`learning step of the process of FIG. 3;
`[0016] FIG. 8 is an embodiment of the process for unsu-
`pervised
`learning;
`[0017] FIG. 9 is an embodiment of the generation 1 (first
`pass) rapid asset valuation process;
`[0018] FIG. 10 is a flow diagram of a fuzzy clustering
`learning of FIG. 8;
`method used in the unsupervised
`[0019] FIG. 11 is a pair of tables showing an example of
`for a rapid asset
`model selection
`and model weighting
`evaluation process;
`[0020] FIG. 12 is a table showing exemplary attributes for
`a rapid asset valuation process; and
`[0021] FIG. 13 is a cluster diagram of an exemplary
`clustering method for a rapid asset valuation process; and
`[0022] FIG. 14 is a computer network schematic.
`
`learning
`
`
`
`US 2001/0044766 A1
`
`Nov. 22, 2001
`
`DETAILED DESCRIPTION OF THE
`INVENTION
`[0023] FIG. 1 is a diagram 10 illustrating a known process
`for valuing a large portfolio of assets 12 through an under-
`to making a bid for purchasing
`writing cycle and through
`in an auction. FIG. 1 is a
`asset portfolio 12, for example,
`high level overview of a typical underwriting
`and extrapo-
`lation process 10 which is not iterative and not automated.
`14 a number of
`In diagram 10, underwriters
`underwrite
`assets from portfolio 12 to generate
`an under-
`individual
`written first portion 16 and an untouched
`remainder portion
`18. Before any of the assets are underwritten,
`first portion 16
`is zero percent and remainder portion 18 is one hundred
`percent of portfolio 12. As
`process
`the underwriting
`first portion 16 increases and remainder portion
`progresses,
`18 decreases. The objective is to underwrite
`as many assets
`as possible before a bid is submitted for the purchase of asset
`portfolio. The team of underwriters
`continues
`individually
`underwriting 14 until just before a bid must be submitted. A
`gross extrapolation 20 is made to evaluate remainder portion
`18. The extrapolated value 20 becomes the non-underwritten
`inferred value 24. The gross extrapolation generates a valu-
`ation 24 for remainder portion 18. Valuation 22 is simply the
`in first portion 16.
`total of the
`asset values
`individual
`However, valuation 24 is a group valuation generated by
`and may be discounted
`accordingly. Valua-
`extrapolation
`tions 22 and 24 are then totaled to produce the portfolio asset
`value 26. Valuation processes are performed on each tranche
`of the portfolio.
`[0024] FIG. 2 is a diagram
`illustrating one embodiment of
`in FIG. 2 are
`a system 28 for rapid asset valuation. Included
`of process steps
`taken by system 28 in
`representations
`valuating asset portfolio 12. System 28 individually
`evalu-
`ates (" touches" ) every asset, except for a very small quantity
`30 of untouched assets considered statistically
`insignificant
`or financially
`immaterial. Specifically, all assets in portfolio
`12 other than quantity 30 undergo an iterative and adaptive
`valuation 32 in which
`the assets in portfolio 12 are indi-
`tables
`listed
`valued,
`individually
`then
`vidually
`in
`and
`into any desired or
`selected from
`the tables and grouped
`groups or
`for bidding
`tranches
`(as
`required
`purposes
`described below. ) As in diagram 10, underwriters
`begin a
`full underwrite 14 of individual
`in portfolio 12 to
`assets
`first portion 16 of assets.
`produce
`a fully underwritten
`also underwrite 34 a sample of assets
`Underwriters
`in a
`second portion 36 of portfolio 12, and a computer 38
`infers 40 value for a third portion 42 of portfolio
`statistically
`12. Computer 38 also
`generates 44 tables
`repetitively
`(described below) showing values assigned
`to the assets in
`portions 16, 36 and 42 as described below. In one embodi-
`ment, computer 38 is configured as a stand alone computer.
`computer 38 is configured
`In another embodiment,
`as a
`server connected
`to at least one client system
`a
`through
`in FIG. 14), such as a
`and described
`network
`(shown
`wide-area network (WAN) or a local-area network (LAN).
`to FIG. 2, an
`[0025] For example,
`and still referring
`portion 46 of a third por-
`and non-underwritten
`unsampled
`tion 42 of portfolio 12 is subjected
`to a statistical
`inference
`procedure 40 using fuzzy-C means clustering ("FCM") and
`("HELTR")
`a composite High/Expected/Low/Timing/Risk
`two categories 48 and 50. HELTR is
`defined as H — High cash flow, E — Expected cash flow,
`score
`to generate
`L — Low cash flow, T — Timing of cash flow (for example
`
`in
`
`months: 0-6, 7-18, 19-36, 37-60), and R — Risk assessment
`of borrower (9 — boxer used by credit analysts). Category 48
`
`for evaluation as
`to have sufflcient commonality
`is deemed
`a whole. Category 50 is further divided
`into clusters 52 and
`54 that are, in turn, further subdivided. Cluster 52 is divided
`into subclusters 56 and 58, while cluster 54 is subdivided
`into subclusters 60, 62 and 64. Cluster and subclusters are
`shown both in a "tree" chart 66 and as boxes in valuation
`block 68. These individual asset values are then regrouped
`into tranches 70, 72 and 74 for bid purposes. Any number of
`tranches could be assembled
`set by the
`in any arrangement
`seller.
`[0026]
`Individual asset data (not shown) for each asset in
`portfolio 12 is entered
`into a database 76 from which
`selected data 78 is retrieved based on a given criteria 80 for
`the iterative and adaptive process 32. When criteria 80 is
`for valuation of any asset, that established cri-
`established
`teria 80 is stored in database 76 for use in valuating other
`asset data in database 76 which shares such an established
`and adaptive valuation process 32 thus
`criteria. Iterative
`develops 82 valuations
`(described below) and groups 84
`them for use in bidding.
`[0027] FIGS. 3 and 4 together
`form a flowchart 85
`a functional overview of one embodiment of
`illustrating
`system 28 (shown in FIG. 2) for evaluation of a large asset
`portfolio 12. Valuation procedures 14, 34 and 40 (see also
`FIG. 2) are simultaneously
`and sequentially used in system
`28 in a manner described below. As described above, full
`14 is a first
`type of valuation procedure.
`underwriting
`Grouping and sampling underwriting 34 with fall underwrit-
`ing of the samples
`is a second type of valuation procedure.
`Statistical inference 40 is a third type of valuation procedure,
`which is an automated grouping
`and automated valuation.
`Procedures 14, 34 and 40 are based on objective criteria
`established as described below.
`[0028] "Underwriting"
`as used herein means a process in
`which a person (" underwriter" ) reviews an asset in accor-
`dance with established principles and determines
`a current
`the asset. During underwriting,
`purchase price for buying
`the
`underwriter uses pre-existing or established criteria 80 for
`the valuations. "Criteria" means rules relevant
`to asset value
`and a rating based on such categories. For example, as a
`three years of cash
`criteria, an underwriter might determine
`flow history of the borrower
`to be a category of information
`to asset valuation and might give a certain rating to
`relevant
`levels of cash flow.
`various
`14 is done in two ways, a full
`[0029] Full underwriting
`cash basis manner 86 and a partial cash basis manner 88.
`Both full cash basis manner 86 and partial cash basis manner
`88 start with sets 90 and 92 of assets that are fully individu-
`ally reviewed 14 (see FIG. 2). Such full review 14 is usually
`large dollar, or other appropriate
`to
`currency,
`the
`due
`amounts of the assets being reviewed relative to other assets
`in the portfolio or due to the borrower being so well known
`or so reliable that the assets can be quickly and reliably fully
`or the assets are marked
`to market such that
`underwritten
`there is very little variance associated with the value of said
`assets. Asset set 90 is evaluated by underwriters 94 and each
`asset in set 90 receives a valuation with very little variation
`such as an asset backed with cash or a tradable commodity
`with full cash value and is placed in a full value table 96.
`Selected individual values for assets in table 96 are stored as
`group value 98.
`a fully underwritten
`
`
`
`US 2001/0044766 A1
`
`Nov. 22, 2001
`
`[0030] Set 92 is evaluated by a team of underwriters 100,
`which could be the same as team 94, but each asset receives
`a discounted or partial value and is placed in a partial value
`table 102. Selected individual values for assets in a tranche
`in table 102 are stored as a partial value fully underwritten
`group value 104. Criteria 80 (shown in FIG. 2) for full cash
`basis manner 86 and partial cash basis manner 88 are stored
`in database 76 (shown in FIG. 2) in a digital storage memory
`in FIG. 2) for use in
`(not shown) of computer 38 (shown
`learning 208 of
`learning 206 and unsupervised
`supervised
`automated valuation 40.
`[0031] Sampling underwriting 34 is accomplished using
`two procedures, a full sampling 106 procedure and a partial
`sampling 108 procedure. Full sampling 106 is utilized
`for
`categories of large assets and includes a one hundred percent
`sampling 110 of the sample groups in the categories of assets
`in full sampling 106 are not
`being sampled. The assets
`but rather are underwritten
`in full
`individually underwritten
`sampling groups 112 based on a determined commonality. A
`is
`group valuation
`full sampling
`resulting
`(not shown)
`based on a rule 114 to
`created and
`then desegregated
`table 116.
`full sample asset value
`an individual
`generate
`in table 116 are then
`full sample asset values
`Individual
`into any full sampling group valu-
`uploaded electronically
`ation 118 required for bidding as suggested by the grouping
`of assets in a tranche. The number of assets in an under-
`can be as little as one
`to any
`sample grouping
`writing
`number of assets. Partial
`sampling 108 is for medium
`categories of assets and includes forming a cluster sample
`group 120 by one hundred percent sampling of a represen-
`a cluster of the groups being
`tative group
`from within
`sampling of the other groups
`sampled and random
`in the
`cluster. In partial sampling 108, all groups are sampled, but
`from cluster sample
`some are partly valued by extrapolation
`group 120. Partial sampling 108 includes
`an asset level
`re-underwrite 122 with manual data entry 125 to produce an
`table 126 which is given an asset class
`alpha credit analyst
`adjustment 128 to produce an adjusted credit analyst
`table
`130. As described above, individual assets are selected from
`adjusted credit analyst table 130 according to tranche group-
`ing to produce a partial sampling credit value 132 for use in
`in FIG. 2).
`bidding on tranche 70 (shown
`[0032] Automatic valuation procedure 40 utilizes super-
`vised learning process 206, an unsupervised
`learning pro-
`cess 208 and an upload from a statistical
`inferencing algo-
`table 136
`rithm 134 to generate
`clusters
`an underwriting
`is stored
`in a digital storage device. In supervised
`which
`process 206, an experienced
`underwriter who
`learning
`to ask to establish value, assists the
`knows what questions
`in determining whether or not an asset is a good
`computer
`the asset. In unsupervised
`to value
`and how
`investment
`learning process 208, the computer segments and classifies
`self-evaluates
`the assets based on
`assets and objectively
`feedback from the data. An underwriter periodically
`reviews
`learning process 208 to determine whether
`the unsupervised
`conclusions.
`is making sensible underwriting
`the computer
`The computer uses statistical algorithms 134 to make
`its
`inferences. For example, but not by way of limitation, one
`the Design For Six Sigma ("DFSS")
`uses
`embodiment
`and used by General Electric
`quality paradigm developed
`("DD") asset
`in a Due Diligence
`and applied
`Company
`product devel-
`valuation process using a multi-generational
`("MGPD") mode
`the asset data with
`to value
`opment
`increasing accuracy. Learning processes 206 and 208 incor-
`
`knowledge
`accumulated
`the valuation
`porate
`as
`the
`into cash flow recovery and probability of recov-
`progresses
`ery calculations on an ongoing, real time basis. Supervised
`learning process 206 uses business rules to identify clusters
`of assets having common aspects for valuation purposes.
`learning process 208 uses feedback from prior
`Unsupervised
`if
`data valuations performed by procedure 40 to determine
`is being made with respect to increasing valuation
`progress
`Identification of all available
`confidence.
`raw data and
`discovery of interrelationships
`of clusters of these available
`raw data is possible due to the use of high-speed computers,
`as is described below.
`
`[0033] In one exemplary embodiment,
`a fuzzy clustering
`means ("FCM") process of unsupervised
`organization of
`raw data using a HELTR scoring technique
`is employed
`to
`infer valuations of credit scores onto assets in portfolios, as
`described below. Such clustering
`have been
`techniques
`classification
`to more sophisticated
`developed
`in response
`to describe assets and high asset counts in port-
`segments
`folios that must be assessed in time periods that do not allow
`manual processing.
`
`[0034] One exemplary method
`first organizes valuation
`scores (static and/or probabilistic
`recoveries)
`in a comput-
`erized system. Adjustments
`are then made to the valuation
`scores for special factors and business decisions. Then a
`reconciliation of multiple valuation
`scores describing
`the
`to interview/override
`same asset and an overall adjustment
`is performed.
`the inferred valuation
`
`scores is performed by col-
`[0035] Organizing valuation
`in electronic form, a cluster number, a cluster name,
`lating,
`descriptive attributes of the cluster(s), probabilistic recovery
`is a HELTR score) and the
`values (an illustrative
`example
`in each cluster's valuation based
`underwriter's
`confidence
`the strengths of each cluster's descriptive
`attributes.
`upon
`identifier of a specific set of
`The cluster number
`is a unique
`that are facts about an asset which a
`descriptive attributes
`person skilled in evaluations uses to assess value of an asset.
`Examples of descriptive
`include, but are not
`attributes
`type, borrower's
`credit
`status, asset
`limited
`to, payment
`worthiness expressed as a score, location and seniority of a
`claim. The cluster name is, in one embodiment,
`an alpha-
`the cluster's
`that describes
`descriptive
`numeric
`name
`attributes or sources. One example of descriptive attributes
`in FIG. 12, described below.
`is found
`
`[0036] Descriptive attributes are the facts or dimensions or
`the asset's value. Com-
`vectors that were used to develop
`puter logic is used to check for replicated clusters, if any, and
`alert the analysts or underwriters.
`
`[0037] Because each asset can be described by many
`combinations of descriptive
`levels of
`attributes, various
`value for the same asset may occur. Probabilistic
`recovery
`indication of the
`values or credit score or any numerical
`asset's worth
`indicators of worth designated
`at the
`are
`discrete asset level. All of the information
`from the various
`such that a purchase or
`is synthesized
`descriptive attributes
`sale price can be ascertained as a fixed value or a probabi-
`is the
`listic one. An illustrative
`used herein
`embodiment
`HELTR score. Each cluster has a unique set of descriptive
`attributes and designated HELTR score.
`
`[0038] Every cluster's unique
`attributes
`of cluster value. Dilferent
`valuation
`
`contribute
`combinations
`
`to a
`of
`
`
`
`US 2001/0044766 A1
`
`Nov. 22, 2001
`
`attributes provide a higher confidence or confidence
`interval
`of a particular cluster's score. For example, if any asset was
`and width equal to 5" — one might ascribe a value of 0 to
`described as a green piece of paper with height equal to 2. 5n
`1000 dollars and place very little confidence
`in this assess-
`ment. If this same asset was described with one more fact or
`attribute or vector as being a real $20 US bill, one would
`place a very high confidence factor on this cluster value of
`$20 US dollars.
`[0039] A cluster's valuation and confidence
`is determined
`at a point in time and recorded. Sometimes new information
`becomes available and the analyst would
`like to alter the
`value(s). The value is altered manually or automatically with
`a data field and decision rules, in the automated
`fashion via
`computer code. The prior values are manipulated
`to reflect
`information. As an illustrative
`example, assume
`new
`the
`prior cluster confidence was recorded at 0. 1 and it is learned
`that a dilferent asset with exact descriptive attributes as in
`this cluster just sold for over the predicted "most probable"
`value. Rules were in elfect such that if this event occurred,
`cluster confidence is multiplied by 10. 0. 1x10=1 which is the
`revised cluster confidence.
`[0040] The purpose of such a process
`is to reconcile
`scores for
`for
`the same asset, controlling
`the
`multiple
`confidence associated with each source of valuation of each
`dimension of valuation. Using the HELTR as an illustrative
`example with sample data points on a particular asset:
`
`are of two
`response variable. Attribute
`variables
`types,
`continuous and categorical. The cross correlations are com-
`puted by the correlation tool between all variables of interest
`and their bin or level and presented,
`in one embodiment,
`in
`a two dimensional matrix for easy identification of trends
`the assets in the portfolios.
`amongst
`
`[0044] First, the cross-correlation
`tool identifies attribute
`in the portfolio of assets as one of continuous or
`variables
`levels are com-
`categorical. For each variable aggregation
`and by value for
`puted by bins for continuous variables
`categorical variables.
`
`[0045] A user looking to identify correlations with the tool
`will select a response variable, Y„ for example, an expected
`recovery or count. For all combinations of pairs of attribute
`variables (x1 and x2) and their levels (a and b), compute the
`average value of the response variable, Y„according to:
`
`Y, =sum(Y(xi=a and x2=b))/count(xi=a
`
`and x2=b).
`
`[0046] An expected value, Y, , „of the response variable
`is calculated according to:
`
`Y, „t=(sum(Y(xi=a))*count(xi=a)+sum(Y(x2=
`b)Pcount(x2=b)))/(count(xi =a) e count(x2=b)).
`
`[0047] A deviation, Y„, „of the chosen response variable,
`Y„ from the expected value, Y, , „using weighted values
`of o