DETAILED ACTION
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA . This action is responsive to pending claims1-3, 5-11, 13-20 filed 6/29/2026.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claim(s) 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The 35 U.S.C. 101 subject matter eligibility analysis first asks whether the claim is directed to one of the four statutory categories (Step 1). It next asks whether the claim is directed to an abstract idea (Step 2A), via Prong 1, whether an abstract idea (e.g., mathematical concept, mental process, certain methods of organizing human activity) is recited, and Prong 2, whether it is integrated into a practical application. It finally asks whether the claim as a whole includes additional elements that amount to significantly more than the judicial exception (Step 2B). See MPEP 2106.
STEP 1: The claims falls within one of the four statutory categories:
As the claims recite methods, hardware apparatuses, and computer media that are not signals per se, the claims fall into statutory categories.
STEP 2A PRONG 1: The claims recite a judicial exception:
The claims recite a technique of reconciling hierarchical time series forecasts by holding certain nodes fixed relative to a base estimate and reconciling the remaining nodes. As such, it is directed to a mathematical process, hat of data reconciliation or constrained optimization, and a mental process, that of generating base forecast data and merging solution data. Additional elements are underlined below and subsequently analyzed. In particular:
For claim 1: A method of generating forecasts from time series data, the method comprising:
receiving, by a prediction module executed by one or more processors, a set of time series data organized according to a data structure having a plurality of nodes (This defines the hierarchical data structure for reconciliation / constrained optimization), wherein the data structure is stored in a time series database and a plurality of nodes are organized according to a hierarchy (This defines a data structure which may be considered mentally);
generating, by a base forecasting module, a plurality of base forecasts, including a base forecast for each node (The generating a plurality of base forecasts for the optimization problem may be performed in the mind); and
selecting a sub-set of the plurality of nodes as fixed nodes (data is selected for the mathematical optimization technique), wherein selecting the sub-set includes selecting only those nodes having time series that are statistically stable based on at least one of a variability below a threshold and an ability to be fit into a linear function, excluding leaf nodes (statistical stability, according to some threshold, and adherence to a linear fit, is a mental judgment that may be performed in the mind);
performing a reconciliation process to generate reconciled forecasts, wherein the reconciliation process includes reconciling only the base forecasts of non-fixed nodes (reconciliation of forecasts of a subset of nodes may be performed mentally);
merging the base forecasts of the fixed nodes and the reconciled forecasts of the non-fixed nodes to generate an overall forecast (the is combining the fixed and reconciled nodes into an overall solution, this merging or data rearranging being a mental process), and outputting the overall forecast as predictions by a prediction module (ibid: overall forecasts are output for evaluation, such as via prediction module, see e.g. Table 2-3), wherein selecting the sub-set of the nodes as fixed nodes reduces the intensity and complexity of computations by excluding nodes considered to be more stable from reconciliation (Forming selections based on judgments, stability judgments, may be performed mentally).
For claim 2: The method of claim 1, wherein the time series data is organized as at least one of a hierarchical time series and a grouped time series (This merely limits the type of mathematical structures for input).
For claim 3: The method of claim 1, wherein selecting the sub-set includes at least one of:
randomly selecting nodes from the nodes of the data structure, excluding bottom layer nodes and leaf nodes (The random selection of nodes to hold fixed may be performed mentally);
selecting nodes based on knowledge relating to stability of a domain of each node (Selecting based on judgments of stability is a mental process); and
selecting nodes based on a statistical analysis of time series data in each node (selecting based on observations of statistical characteristics is a mental process).
For claim 5: The method of claim 1, wherein the reconciliation process is based on a constrained optimization problem based on a Lagrange function having a Lagrange multiplier, and includes selecting a computationally stable formulation or a computationally efficient formulation for solving the optimization problem (Performing constrained optimization reconciliation on a Lagrange form is a mathematical concept).
For claim 6: The method of claim 1, wherein the reconciliation process includes splitting the nodes into a plurality of sets of nodes, and performing reconciliation separately for each set of nodes (Dividing a group of noes for separate reconciliation processes a mental process).
For claim 7: The method of claim 5, wherein the computationally stable formulation includes solving the optimization problem by solving for a vector of forecasts of the non-fixed nodes concatenated with the Lagrange multiplier (Solving a constrained optimization problem via concatenation with the Lagrangian is a mathematical concept).
For claim 8: The method of claim 5, wherein the computationally efficient formulation includes solving the optimization problem by solving for the Lagrange multiplier, and subsequently solving for a vector of forecasts of the non-fixed node forecasts (Solving a constrained optimization problem via sequential determination of Lagrangian multiplier followed by vector forecasts is a mathematical process).
The remaining limitations recite analogous apparatuses and computer media and hence are likewise rejected.
STEP 2A PRONG 2: The claims do not integrate the exception into a practical application:
The additional elements are:
Claims 1, 9, 17 recite modules, a prediction module and a forecasting module, implemented via processors, as well as a database for storing data structures. However, this is mere instructions to implement the abstract idea on a computer and database environment and hence does not constitute an integration into a practical application.
For claim 9: An apparatus for generating forecasts from time series data, comprising one or more computer processors that comprise: a processing unit including a processor. However, this is mere instructions to implement an abstract idea on a computer and hence does not constitute an integration into a practical application.
For claim 17: A computer program product comprising a storage medium readable by one or more processing circuits, the storage medium storing instructions executable by the one or more processing circuits to perform a method. However, this is mere instructions to implement an abstract idea on a computer and hence does not constitute an integration into a practical application.
STEP 2B: The claim as a whole do not include additional elements that amount to significantly more than the abstract idea:
Claims 1, 9, 17 recite modules, a prediction module and a forecasting module, implemented via processors, as well as a database for storing data structures. However, the use of general purpose computer hardware, including processors, memory and a database environment, etc. to implement an abstract idea is well-understood, routine, and conventional in the field of data processing and hence does not constitute significantly more.
For claim 9: An apparatus for generating forecasts from time series data, comprising one or more computer processors that comprise: a processing unit including a processor. The use of a computer with processor and memory medium to implement optimization techniques is well-understood, routine and conventional (WURC) and hence does not constitute significantly more (2b).
For claim 17: A computer program product comprising a storage medium readable by one or more processing circuits, the storage medium storing instructions executable by the one or more processing circuits to perform a method. The use of a computer with processor and memory medium to implement optimization techniques is well-understood, routine and conventional (WURC) and hence does not constitute significantly more (2b).
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1-3, 5-6, 8-11, 13-14, 16-19 are rejected under 35 U.S.C. 103 as being unpatentable over Zhang ("Optimal reconciliation with immutable forecasts", published 11/24/2022), with Wickramasuriya ("Optimal non-negative forecast reconciliation", published 2020) incorporate by reference, in view of Leonard (US 20140257778 A1).
For claim 1, Zhang discloses: a method of generating forecasts from time series data, the method (§2.1 gives problem overview, with §3: “Forecast reconciliation with immutability constraints” giving an overview of the technique of generating reconciled forecasts given time series data) comprising:
receiving, by a prediction module (§4-6 gives overview of generating and evaluating predictions), a set of time series data organized according to a data structure having a plurality of nodes (§3 ¶1: hierarchical time series data is received having a plurality of immutable and mutable nodes, see also fig.1 showing possible hierarchies), and the plurality of nodes are organized according to a hierarchy (§3 ¶1, fig.1);
generating, by a base forecasting module, a plurality of base forecasts, including a base forecast for each node (§3.1 ¶1: y-hat = (w-hat, v-hat, u-hat) is generated as base forecasts);
selecting a sub-set of the plurality of nodes as fixed nodes (§3.1 ¶1: u_t is selected as immutable), wherein selecting the sub-set includes selecting only those nodes having time series that are statistically stable based on at least one of a variability below a threshold and an ability to be fit into a linear function, excluding leaf nodes (§4 ¶1-3 discloses selecting non-leaf aggregate nodes having less noise compared to unaggregated nodes, with §4.2.1 disclosing generating bottom level series with increased noise, hence, the top levels have a higher signal-to-noise ratio due to the increased noise; based on the bottom level noises averaging out, the variability, of, say, the selected BA, BB branch will have the gamma noise canceled out and the AA, AB will have the epsilon noise cancelled out, hence, selecting aggregate series based on variability below an original threshold), and checking whether the selected nodes are evenly distributed across the hierarchy to avoid concentrating the fixed nodes on one level (§4: ¶1-3: nodes are evenly distributed across top nodes to avoid concentration on any particular level, e.g., bottom level, hence, checking adherence to the strategy before performing the analysis);
performing a reconciliation process to generate reconciled forecasts, wherein the reconciliation process includes reconciling only the base forecasts of non-fixed nodes (§3 eq.3 gives general form of a reconciled forecast, with the immutable nodes u held constant, with eq.4-5 giving a closed-formed solution for v-tilde (as u is immutable and w is derived from u and v via S, the aggregation matrix), hence, reconciliation is performed on the base forecasts of the non-fixed nodes); and
merging the base forecasts of the fixed nodes and the reconciled forecasts of the non-fixed nodes to generate an overall forecast (eq.3, eq5: fixed nodes forecasts u-tilde = u-hat and reconciled forecasts (v-tilde, w-tilde) are merged for an overall forecast), and outputting the overall forecast as predictions by a prediction module (ibid: overall forecasts are output for evaluation, such as via prediction module, see e.g. Table 2-3), wherein selecting the sub-set of the nodes as fixed nodes reduces the intensity and complexity of computations by excluding nodes considered to be more stable from reconciliation (§3 eq.4-5 discloses an overall algorithm for determining v-tilde based on the unreconciled forecasts w-hat, u-hat, with §4 ¶3 contemplating selecting fixed nodes as less noisy or more stable nodes, hence, said nodes being excluded from reconciliation, hence, reducing complexity and intensity, for example, based on a smaller overall size of the reconciliation matrix, based on smaller set to be reconciled (v rather than v and u), etc., see §2.1 eq.1).
Zhang does not disclose the hardware implementation: wherein the modules are executed by one or more processors; wherein the data structure is stored in a time series database.
Leonard discloses: wherein the modules are executed by one or more processors (0025-26); wherein the data structure is stored in a time series database (0026, 0062, hence combination with Zhang disclosing use of database to store data structures defining time series hierarchies).
It would have been obvious before the effective filing date to one of ordinary skill in the art to modify the method of Zhang by incorporating the time series database of Leonard. Both concern the art of data forecasting, and the incorporation would have, according to Leonard, provided a common computer data environment for processing the data (0025).
For claim 2, Zhang modified by Leonard discloses the method of claim 1, as described above. Zhang further discloses: wherein the time series data is organized as at least one of a hierarchical time series and a grouped time series (p.651 col.1 last ¶ contemplates application to grouped time series, see also §6 for application to grouped data).
For claim 3, Zhang modified by Leonard discloses the method of claim 1, as described above. Zhang further discloses: wherein selecting the sub-set includes at least one of:
randomly selecting nodes from the nodes of the data structure, excluding bottom layer nodes and leaf nodes;
selecting nodes based on knowledge relating to stability of a domain of each node (p.651 col.1 ¶3: sufficiently long history reflects stability of predictions); and
selecting nodes based on a statistical analysis of time series data in each node (§5 ¶2: intermittent time series or promotional peaked series constitutes a statistical feature analyzed from the data).
For claim 5, Zhang modified by Leonard discloses the method of claim 3, as described above. Zhang further discloses, in p.653 c.1 ¶2, using various algorithms disclosed by Wickramasuriya 2020 such as BPV, PCG, scaled gradient projection, hence, Zhang discloses: wherein the reconciliation process is based on a constrained optimization problem based on a Lagrange function having a Lagrange multiplier (Wickramasuriya §2.3 goes discloses a Lagrange formulation of the constrained optimization problem for non-negative reconciliation, hence, the reconciliation being based on the Lagrangian function), and includes selecting a computationally stable formulation or a computationally efficient formulation for solving the optimization problem (Wickramasuriya §3.1-3 goes over several numerical algorithms for solving said problem, with BPV (§3.1) being computationally unstable, i.e., causing a cycle, see p.1171 col.1 ¶2, and the remaining 3.2-3 not having the issue and using a computationally efficient algorithms such as by finding a piecewise-linear path to the solution, see §3.2 ¶1 ).
For claim 6, Zhang modified by Leonard discloses the method of claim 1, as described above. Zhang further discloses: wherein the reconciliation process includes splitting the nodes into a plurality of sets of nodes, and performing reconciliation separately for each set of nodes (§3 ¶1 discloses separating nodes into an immutable, mutable set, and non-basis set, the reconciliation being performed separately on each set such that the immutable set is unchanged, the mutable set is optimized, and the basis set is adjusted accordingly ).
For claim 8, Zhang modified by Leonard discloses the method of claim 5, as described above. Zhang further discloses: wherein the computationally efficient formulation includes solving the optimization problem by solving for the Lagrange multiplier, and subsequently solving for a vector of forecasts of the non-fixed node forecasts (Wickramasuriya p.1173 alg.3.3 (scaled gradient projection) algorithm includes iteratively solving for the Lagrangian multiplier lambda at step 8-10 and subsequently solving for base b corresponding to iterative base reconciliation forecasts corresponding to non-fixed nodes).
Claims 9-11, 13-14, 16-19 recite apparatuses and computer media corresponding to the above methods and are hence likewise rejected.
Claim(s) 7, 15, 20 are rejected under 35 U.S.C. 103 as being unpatentable over Zhang ("Optimal reconciliation with immutable forecasts", published 11/24/2022), with Wickramasuriya ("Optimal non-negative forecast reconciliation", published 2020) incorporate by reference, in view of Leonard (US 20140257778 A1) in view of Won ("Newton's Method for Constrained Optimization", published 12/8/2021).
For claim 7, Zhang modified by Leonard discloses the method of claim 5, as described above. Zhang does not disclose: wherein the computationally stable formulation includes solving the optimization problem by solving for a vector of forecasts of the non-fixed nodes concatenated with the Lagrange multiplier.
Won discloses: wherein the computationally stable formulation includes solving the optimization problem by solving for a vector concatenated with the Lagrange multiplier (p.4-5 “Feasible start Newton”, “Infeasible Start Newton” shows solving for the concatenation of the vector x concatenated with the Lagrangian estimator v / w via sequential updates, hence, combination with Zhang yielding a computationally stable (e.g., no cycling issues) formulation for solving the constrained hierarchical distillation problem).
It would have been obvious before the effective filing date to a person of ordinary skill in the art to modify the method of Zhang modified by Leonard by incorporating the numerical method of Won. Both concern the art of numerical solving of constrained optimization problems, and the incorporation would have been, according to Won, to use a fundamental and theoretically grounded numerical convergence method to solve constrained optimization problems (p.2: “Course Objectives”).
Claim 15 recites an apparatus corresponding to the above method and is hence likewise rejected.
For claim 20, Zhang modified by Leonard discloses the method of claim 19, as described above. Zhang further discloses: wherein the computationally efficient formulation includes solving the optimization problem by solving for the Lagrange multiplier, and subsequently solving for a vector of forecasts of the non-fixed node forecasts (Wickramasuriya p.1173 alg.3.3 (scaled gradient projection) algorithm includes iteratively solving for the Lagrangian multiplier lambda at step 8-10 and subsequently solving for base b corresponding to iterative base reconciliation forecasts corresponding to non-fixed nodes).
Zhang modified by Leonard does not disclose: wherein the computationally stable formulation includes solving the optimization problem by solving for a vector of forecasts of the non-fixed nodes concatenated with the Lagrange multiplier.
Won discloses: wherein the computationally stable formulation includes solving the optimization problem by solving for a vector concatenated with the Lagrange multiplier (p.4-5 “Feasible start Newton”, “Infeasible Start Newton” shows solving for the concatenation of the vector x concatenated with the Lagrangian estimator v / w via sequential updates, hence, combination with Zhang yielding a computationally stable (e.g, no cycling issues) formulation for solving the constrained hierarchical distillation problem).
It would have been obvious before the effective filing date to a person of ordinary skill in the art to modify the method of Zhang modified by Leonard by incorporating the numerical method of Won. Both concern the art of numerical solving of constrained optimization problems, and the incorporation would have been, according to Won, to use a fundamental and theoretically grounded numerical convergence method to solve constrained optimization problems (p.2: “Course Objectives”).
Response to Arguments
Applicant’s arguments have been fully considered. In the remarks, Applicant argued:
1. The claims are not directed merely to mathematical manipulation of forecast data or to a mental process. Rather, the claims recite a computer-implemented forecasting process that operates on time series data stored and organized in a time series database environment, selects statistically stable fixed nodes according to defined node-selection criteria, performs reconciliation only on non-fixed nodes, and outputs an overall forecast.
Examiner respectfully disagrees. Outside of computer implementation, all the recited steps may be performed mentally, and would be part of a general mental forecasting technique independent of the computer implementation, see rejection above. Hence, the steps do not recite nor are meaningfully limited by a technical environment.
2. Under Step 2A, Prong 1, Applicant respectfully submits that the claims, considered as a whole, are directed to a specific improvement in computer-implemented time series forecasting and reconciliation. The specification explains that time series databases are used for device monitoring, production line management, financial analysis, server metrics, performance monitoring data, network data, sensor data, events, clicks, trades, and analytics data.
Examiner respectfully disagrees and submits that claims at present do not incorporate or recite such potential applications to particular technical fields.
3. The claims do not merely recite a result of "forecasting" or "optimization." Instead, the amended claims specify how a prediction module receives grouped and/or hierarchical time series data and hierarchy or grouping information, how a base forecasting module generates base forecasts for nodes, and how fixed nodes are selected based on statistical stability while excluding leaf nodes and checking distribution across the hierarchy.
Examiner respectfully disagrees. At present, the recitation of hardware modules comprise mere instructions to implement the abstract idea on a computer and hence does not constitute an integration into a practical application. The recited technical computing components are extra solutional and are separable from the main forecasting steps which may be performed mentally.
4. Furthermore, the claims integrate any alleged abstract idea into a practical application under Step 2A, Prong 2. The claimed reconciliation is tied to the technical structure of hierarchical and/or grouped time series data and is performed to preserve relationships between nodes.
Examiner respectfully disagrees. Grouped time series data cannot be themselves said to be a technical structure.
5. The claims require selecting only statistically stable nodes, excluding leaf nodes, and checking whether selected nodes are evenly distributed across the hierarchy. These limitations improve the operation of the forecasting process itself by reducing the scope of reconciliation to non-fixed nodes while preserving coherent relationships across the data structure.
Examiner respectfully disagrees. The forecasting process is a mental process and such limitations serve only to improve the mental process itself.
5. The specification states that excluding stable nodes from reconciliation reduces the intensity and complexity of computations, makes reconciliation quicker and more efficient, and addresses limitations of existing approaches that are computationally intensive and require a large number of assumptions. Thus, the claims recite a concrete technical solution to a technical problem in computer-implemented forecast reconciliation.
Examiner respectfully disagrees. Such considerations constitute optimizations to the forecasting mental process and not to particular technical implementations or technical improvements in incorporating the mental process.
Turning no to the prior art rejections:
7. In contrast to the method of claim 1, Zhang and Wickramasuriya do not disclose or suggest at least the feature of "selecting a sub-set of the plurality of nodes as fixed nodes, wherein selecting the sub-set includes selecting only nodes having time series that are statistically stable based on at least one of a variability below a threshold and an ability to be fit to a linear function, excluding leaf nodes, and checking whether the selected nodes are evenly distributed across the hierarchy to avoid concentrating the fixed nodes on one level."
Zhang does not disclose or suggest the fixed node selection process recited in claim 1. The Office Action relies on Zhang's disclosure, in which a time series is selected as immutable and that immutable forecasts are held constant during reconciliation. At most, the cited description corresponds to the general concept of keeping a chosen or pre-specified subset of forecasts unchanged. Zhang does not disclose selecting fixed nodes based on a statistical stability criteria, excluding leaf nodes, and checking whether the selected nodes are evenly distributed.
Examiner respectfully disagrees, based on the portions of Zhang cited above.
8. Zhang's examples confirm the distinction. Zhang identifies immutable series based on circumstances such as intermittent bottom-level series, bottom-level series with more than 365 days of training observations, or a top-level series whose aggregate signal is strong. Those examples do not teach selecting only nodes having time series that satisfy the claimed statistical stability criteria, much less excluding leaf nodes and checking even distribution of the selected nodes across the hierarchy.
Examiner respectfully disagrees. The cited portions disclose selecting portions with favorable statistical properties such as noise level, SNR, etc., and furthermore the exclusion of leaf nodes. Furthermore, the nodes are selected evenly over the higher nodes to avoid concentration on any particular level (e.g., the bottom level).
9. Indeed, Zhang's use of bottom-level series as immutable is inconsistent with the amended claim's exclusion of leaf nodes in a hierarchy where the lowest level contains leaf nodes. Wickramasuriya does not address the deficiencies of Zhang.
Examiner respectfully disagrees as top level nodes are held immutable in Zhang.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Yanchenko (US 20240211835 A1) discloses hierarchical time series forecasting reconciliation.
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any extension fee pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to LIANG LI whose telephone number is (303)297-4263. The examiner can normally be reached Mon-Fri 9-12p, 3-11p MT (11-2p, 5-1a ET).
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice. The examiner is available for interviews Mon-Fri 6-11a, 2-7p MT (8-1p, 4-9p ET).
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor Jennifer Welch can be reached on (571)272-7212. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from Patent Center and the Private Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from Patent Center or Private PAIR. Status information for unpublished applications is available through Patent Center or Private PAIR to authorized users only. Should you have questions about access to Patent Center or the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free).
/LIANG LI/
Primary examiner AU 2143
Prosecution Insights