# US 2021/0174207 A1 · Analytic system for interactive direct functional principal component analysis

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2021/0174207 A1
- Document ID: US20210174207A1
- Status: Published application
- Kind: A1
- Application number: 17/158,062
- Filed: 2021-01-26
- Published: 2021-06-10
- Earliest claimed: 2019-12-05
- Assignees: SAS Institute Inc.
- Applicants: SAS Institute Inc.
- Inventors: Ryan Jeremy Parker; Clayton Adam Barker; Christopher Michael Gotwalt
- Web page: https://patentgrep.com/patent/US20210174207A1

## CPC

- G06N 3/08: Learning methods
- G06N 20/10: using kernel methods, e.g. support vector machines [SVM]
- G06F 16/2228: {Indexing structures}
- G06F 17/16: Matrix or vector computation {, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization (matrix transposition G06F7/78)}
- G06F 17/17: Function evaluation by approximation methods, e.g. inter- or extrapolation, smoothing, least mean square method ({G06F17/18 takes precedence } ; interpolation for numerical control G05B19/18)
- G06N 5/01: Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
- G06N 20/20: Ensemble learning
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

An analytic system provides direct functional principal component analysis. (A) A next group variable value is selected from values of a group variable. (B) Explanatory variable values of observations having the selected next group variable value are sorted in ascending order. (C) The response variable value associated with each sorted explanatory variable value is stored in a next row of a data matrix. (D) (A) through (C) are repeated. (E) An eigenfunction index is incremented. (F) An FPCA is performed using the data matrix to define an eigenfunction for the eigenfunction index. (G) (E) and (F) are repeated. (H) FPCA results from the performed FPCA are presented within a window of a display. The FPCA results include an eigenvalue and an eigenfunction associated with the eigenvalue for each functional principal component identified from the performed FPCA in (F).

## Claims

### Claim 1

A non-transitory computer-readable medium having stored thereon computer-readable instructions that when executed by a computing device cause the computing device to:
read a dataset that includes a plurality of observation vectors defined for each value of a plurality of values of a group variable, wherein each observation vector of each plurality of observation vectors includes an explanatory variable value of an explanatory variable and a response variable value of a response variable;
(A) select a next group variable value from the plurality of values of the group variable;
(B) sort in ascending order the explanatory variable values of the plurality of observation vectors having the selected next group variable value;
(C) store the response variable value associated with each sorted explanatory variable value from (B) in a next row of a data matrix;
(D) repeat (A) through (C) until each unique group variable value is selected from the plurality of values of the group variable in (A);
(E) increment an eigenfunction index;
(F) perform a functional principal component analysis (FPCA) using the data matrix to define an eigenfunction for the incremented eigenfunction index;
(G) repeat (E) and (F) until a predefined number of eigenfunctions is defined in (E) or an eigenfunctions complete value indicates that defining eigenfunctions is done; and
(H) present FPCA results from the performed FPCA within a window of a display, wherein the FPCA results include an eigenvalue and an eigenfunction associated with the eigenvalue for each functional principal component identified from the performed FPCA in (F).

### Claim 2 (depends on 1)

The non-transitory computer-readable medium of claim 1, wherein, after (A) and before (B), the computer-readable instructions further cause the computing device to replace a missing explanatory variable value with a linearly interpolated value computed by linearly interpolating between a previous explanatory variable value and a next explanatory variable value of the of the plurality of observation vectors having the selected next group variable value.

### Claim 3 (depends on 1)

The non-transitory computer-readable medium of claim 1, wherein, after (D) and before (E), the computer-readable instructions further cause the computing device to:
compute a mean value for each column of the data matrix from the data stored in a respective column of the data matrix; and
center the data matrix using the computed mean value for each column of the data matrix.

### Claim 4 (depends on 3)

The non-transitory computer-readable medium of claim 3, wherein the data matrix is centered using Mi,j=Mi,j−μj, i=1, . . . , Ng, j=1, . . . , N, where Mi,j is an i, jth element of the data matrix, μj is the mean value computed for a jth column of the data matrix, Ng is a number of unique group variable values of the plurality of values of the group variable, and N is a number of observation vectors of the plurality of observation vectors having the selected next group variable value.

### Claim 5 (depends on 4)

The non-transitory computer-readable medium of claim 4, wherein the mean value is computed for the jth column of the data matrix using
μ
j
=
1
N
g

∑
i
=
1
N
g


M
ij
,
j
=
1
,
…

,
N
.

### Claim 6 (depends on 1)

The non-transitory computer-readable medium of claim 1, wherein, after (F) and before (G), the computer-readable instructions further cause the computing device to normalize the eigenfunction using a normalization factor.

### Claim 7 (depends on 6)

The non-transitory computer-readable medium of claim 6, wherein the eigenfunction is normalized using ek=ek/F, where ek is a kth eigenfunction, k is the eigenfunction index incremented in (E), and F is the normalization factor.

### Claim 8 (depends on 7)

The non-transitory computer-readable medium of claim 7, wherein the normalization factor is computed using F=√{square root over ((N−1)/N)}, where N is a number of observation vectors of each plurality of observation vectors.

### Claim 9 (depends on 1)

The non-transitory computer-readable medium of claim 1, wherein, after (F) and before (G), the computer-readable instructions further cause the computing device to smooth the eigenfunction using a smoothing function.

### Claim 10 (depends on 9)

The non-transitory computer-readable medium of claim 9, wherein the smoothing function is selected from the group consisting of a b-splines basis function, a p-splines basis function, a power basis function, a Fourier basis function, a wavelets function, and an exponential basis function.

### Claim 11 (depends on 1)

The non-transitory computer-readable medium of claim 1, wherein, after (F) and before (G), the computer-readable instructions further cause the computing device to normalize the eigenfunction to have a unit norm.

### Claim 12 (depends on 11)

The non-transitory computer-readable medium of claim 11, wherein the eigenfunction is normalized to have unit norm using
e
k

[
i
]
=
e
k

[
i
]
F
,
i
=
1
,
…

,
N
,
where ek is a kth eigenfunction, k is the eigenfunction index incremented in (E), F is a normalization factor, and N is a number of observation vectors of each plurality of observation vectors.

### Claim 13 (depends on 12)

The non-transitory computer-readable medium of claim 12, wherein the normalization factor is computed using F=√{square root over (ek1ek)}, where T indicates a transpose.

### Claim 14 (depends on 1)

The non-transitory computer-readable medium of claim 1, wherein, after (F) and before (G), the computer-readable instructions further cause the computing device to orthogonalize the eigenfunction to the previous k−1 eigenfunctions.

### Claim 15 (depends on 14)

The non-transitory computer-readable medium of claim 14, wherein the eigenfunction is orthogonalized to the previous k−1 eigenfunctions using
e
k

[
i
]
=
e
k

[
i
]
*
(
1
-
e
k
-
1
T

e
k
e
k
-
1
T

e
k
-
1
)
,
i
=
1
,
…

,
N
,
where ek is a kth eigenfunction, k is the eigenfunction index incremented in (E), N is a number of observation vectors of each plurality of observation vectors, and T indicates a transpose.

### Claim 16 (depends on 14)

The non-transitory computer-readable medium of claim 14, wherein, after (F) and before (G), the computer-readable instructions further cause the computing device to compute a functional principal component score from the orthogonalized eigenfunction.

### Claim 17 (depends on 16)

The non-transitory computer-readable medium of claim 16, wherein the functional principal component score is computed using FPCi=Mi1ek, i=1, . . . , Ng, where FPCi is an ith functional principal component score value, Mi is an ith row of the data matrix, ek is a kth orthogonalized eigenfunction, k is the eigenfunction index incremented in (E), Ng is a number of unique group variable values of the plurality of values of the group variable, and T indicates a transpose.

### Claim 18 (depends on 16)

The non-transitory computer-readable medium of claim 16, wherein, after (F) and before (G), the computer-readable instructions further cause the computing device to remove elements from the data matrix based on the computed functional principal component score.

### Claim 19 (depends on 18)

The non-transitory computer-readable medium of claim 18, wherein elements are removed from the data matrix using Mi,j=Mi,j−FPCixek[i], i=1, . . . , Ng, j=1, . . . , N, where Mi,j is an i,jth element of the data matrix, FPCi is an ith functional principal component score value, ek is a kth orthogonalized eigenfunction, k is the eigenfunction index incremented in (E), Ng is a number of unique group variable values of the plurality of values of the group variable, N is a number of observation vectors of the plurality of observation vectors having the selected next group variable value, and where x indicates a cross product.

### Claim 20 (depends on 1)

The non-transitory computer-readable medium of claim 1, wherein, after (H), the computer-readable instructions further cause the computing device to:
receive an indicator of a request to perform functional analysis using the FPCA results and the dataset based on a predefined factor variable, wherein the indicator is received in association with the window of the display, wherein each observation vector of each plurality of observation vectors further includes a factor variable value of the predefined factor variable;
(I) train a model using the eigenvalue and the eigenfunction computed as a result of the performed FPCA for each plurality of observation vectors using the factor variable value associated with each observation vector of each plurality of observation vectors as a model effect; and
(J) present trained model results from the trained model within a sub-window of the window of the display.

### Claim 21 (depends on 20)

The non-transitory computer-readable medium of claim 20, wherein a selector associated with the first window of the display provides user selection of a type of model to train.

### Claim 22 (depends on 21)

The non-transitory computer-readable medium of claim 21, wherein the type of model to train is selected from the group consisting of a generalized regression model and a neural network model.

### Claim 23 (depends on 21)

The non-transitory computer-readable medium of claim 21, wherein the type of model to train is a generalized regression model and a second selector associated with the first window of the display provides user selection of a response distribution method used by the generalized regression model.

### Claim 24 (depends on 21)

The non-transitory computer-readable medium of claim 21, wherein the type of model to train is a generalized regression model and a second selector associated with the first window of the display provides user selection of an estimation method used by the generalized regression model.

### Claim 25 (depends on 21)

The non-transitory computer-readable medium of claim 21, wherein the type of model to train is a generalized regression model and a second selector associated with the first window of the display provides user selection of a validation method used by the generalized regression model.

### Claim 26 (depends on 20)

The non-transitory computer-readable medium of claim 20, wherein, after (J), a selector associated with the window of the display provides user modification of a current factor variable value for one of a plurality of predefined factor variables, wherein (I) and (J) are automatically repeated in response to the user modification of the current factor variable value.

### Claim 27 (depends on 26)

The non-transitory computer-readable medium of claim 26, wherein the computer-readable instructions further cause the computing device to:
train a model using observation vector values associated with the explanatory variable, the response variable, and the user modified factor variable value;
read a second explanatory variable value from a scoring dataset;
compute a new response variable value using the trained model; and
output the computed new response variable value.

### Claim 28 (depends on 1)

The non-transitory computer-readable medium of claim 1, wherein, after (H), a selector associated with the window of the display provides user control of a number of functional principal components identified from the performed FPCA.

### Claim 29

A computing device comprising:
a processor; and
a non-transitory computer-readable medium operably coupled to the processor, the computer-readable medium having computer-readable instructions stored thereon that, when executed by the processor, cause the computing device to
read a dataset that includes a plurality of observation vectors defined for each value of a plurality of values of a group variable, wherein each observation vector of each plurality of observation vectors includes an explanatory variable value of an explanatory variable and a response variable value of a response variable;
(A) select a next group variable value from the plurality of values of the group variable;
(B) sort in ascending order the explanatory variable values of the plurality of observation vectors having the selected next group variable value;
(C) store the response variable value associated with each sorted explanatory variable value from (B) in a next row of a data matrix;
(D) repeat (A) through (C) until each unique group variable value is selected from the plurality of values of the group variable in (A);
(E) increment an eigenfunction index;
(F) perform a functional principal component analysis (FPCA) using the data matrix to define an eigenfunction for the incremented eigenfunction index;
(G) repeat (E) and (F) until a predefined number of eigenfunctions is defined in (E) or an eigenfunctions complete value indicates that defining eigenfunctions is done; and
(H) present FPCA results from the performed FPCA within a window of a display, wherein the FPCA results include an eigenvalue and an eigenfunction associated with the eigenvalue for each functional principal component identified from the performed FPCA in (F).

### Claim 30

A method of providing direct functional principal component analysis, the method comprising:
reading, by a computing device, a dataset that includes a plurality of observation vectors defined for each value of a plurality of values of a group variable, wherein each observation vector of each plurality of observation vectors includes an explanatory variable value of an explanatory variable and a response variable value of a response variable;
(A) selecting, by the computing device, a next group variable value from the plurality of values of the group variable;
(B) sorting, by the computing device, in ascending order the explanatory variable values of the plurality of observation vectors having the selected next group variable value;
(C) storing, by the computing device, the response variable value associated with each sorted explanatory variable value from (B) in a next row of a data matrix;
(D) repeating, by the computing device, (A) through (C) until each unique group variable value is selected from the plurality of values of the group variable in (A);
(E) incrementing, by the computing device, an eigenfunction index;
(F) performing, by the computing device, a functional principal component analysis (FPCA) using the data matrix to define an eigenfunction for the incremented eigenfunction index;
(G) repeating, by the computing device, (E) and (F) until a predefined number of eigenfunctions is defined in (E) or an eigenfunctions complete value indicates that defining eigenfunctions is done; and
(H) presenting, by the computing device, FPCA results from the performed FPCA within a window of a display, wherein the FPCA results include an eigenvalue and an eigenfunction associated with the eigenvalue for each functional principal component identified from the performed FPCA in (F).
