DETAILED ACTION
Claims 1-20 are pending.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
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.
Claims 1-4, 6, 8-13, and 15-19 are rejected under 35 U.S.C. 103 as being unpatentable over Seeger et al. (US 10,748,072 B1).
Regarding claim 1, Amazon teaches a computer-implemented method (Col. 2, lines 48-49: methods and apparatus for intermittent demand forecasting for large data sets are described) comprising:
segmenting at least one resource demand time series data, into multiple segments based on at least one resource demand level threshold (Col. 3, lines 4-11: a forecaster may determine that a particular input data set indicating demand observations for one or more items over a period of time meets an intermittency criterion that makes the data set suitable for a composite latent state model. The criterion may be based, for example, on the fraction of entries in the data set that are zeroes, the temporal distribution of the non-zero entries, and/or on other factors.; Col. 5, lines 34-35: intermittent bursty data sets—where many zt are zero and a few zt are large.; Col. 8, lines 24-46: Data sources 105 may include at least one source from observed item demand time series 110 for some set of items may be obtained, and one source from which feature metadata 120 may be obtained… In other embodiments, fresh demand information and/or fresh feature metadata may be provided to the forecaster as soon as it becomes available, and the forecaster 150 may be responsible for discretizing or batching the data to make its predictions.);
determining at least one probability distribution that fits at least a plurality of the multiple segments using one or more statistical analyses (Col. 3, lines 24-37: In various embodiments, the forecaster may generate, with respect to the input data set, a statistical model which utilizes a likelihood function comprising one or more latent functions. At least one latent function may be a combination of a deterministic function and a random process. One such approach may combine, for example, a generalized linear model and probabilistic smoothing. In other embodiments, nonlinear deterministic functions such as various types of neural networks may be used. Generally speaking, the statistical model may utilize either non-Gaussian or Gaussian likelihood functions, with the non-Gaussian likelihoods being most useful for dealing with intermittent data sets. Free parameters of the statistical model may be fitted in some embodiments using approximate Bayesian inference; Col. 3, lines 42-47: After the model parameters have been fitted, the model may be run to produce probabilistic demand forecasts—e.g., corresponding to various future times or time intervals, a probability distribution (and not just point predictions) of the demand for a given item may be generated.);
generating forecasts for two or more of the multiple segments for at least one future time period based at least in part on the at least one probability distribution (Col. 3, lines 42-47: After the model parameters have been fitted, the model may be run to produce probabilistic demand forecasts—e.g., corresponding to various future times or time intervals, a probability distribution (and not just point predictions) of the demand for a given item may be generated.; Col. 1, lines 56-60: FIG. 1 illustrates an example forecasting system in which probabilistic forecasts for intermittent demand data sets may be generated using composite latent state models, according to at least some embodiments.; Col. 8, lines 47-53: Generally speaking, the forecaster 150 may comprise one or more computing devices collectively configured to implement a set of forecasting algorithms and models. A variety of statistical models may be supported in different embodiments, including for example composite latent state models 152 of the kind described earlier, which incorporate both deterministic components and random processes.);
generating a resource demand forecast for the at least one resource for the at least one future time period by aligning the forecasts for the two or more of the multiple segments to at least one time index associated with the at least one future time period using one or more machine learning techniques (Col. 3, lines 50-60: The results provided by the forecaster may be used to make various types of business decisions more intelligently—e.g., to generate purchase orders for appropriate quantities of items at appropriate times, to decide whether it is worthwhile to continue stocking various items or not, to plan ahead for warehouse space expansions, and so on. In at least some embodiments, the forecasts may be provided as input to an automated ordering system, which may in turn transmit orders for various inventory items to meet business requirements of the organization on behalf of which the forecasts were produced.; Col. 9, lines 23-29: A number of iterations of training, testing and evaluation may be performed for the composite model in some embodiments, with initial settings and/or other model parameters or hyper-parameters being adjusted between the iterations as needed based on the accuracy of the predictions with respect to test data sets that were not used for training.; Col. 9, lines 30-33: After the model has been fitted by the forecaster 150, probabilistic predictions for future demand may be generated.; Col. 15, lines 2-7: In some embodiments, some or all of the forecasting algorithms for intermittent data described above may be implemented at a machine learning service of a provider network. FIG. 8 illustrates example components of a machine learning service which may be used for generating forecasts for time series data, according to at least some embodiments.); and
performing one or more automated actions based at least in part on the resource demand forecast (Col. 9, lines 36-43: The probabilistic forecasts 180 may be provided programmatically to one or more forecast consumers 185. In some embodiments, for example, the forecasts may be transmitted to an automated ordering system via one or more application programming interface (APIs), and the projections of the forecast may be used to place orders for various items.);
wherein the method is performed by at least one processing device comprising a processor coupled to a memory (In at least some embodiments, a server that implements a portion or all of one or more of the technologies described herein, including the composite forecasting techniques, as well as various components of a machine learning service may include a general-purpose computer system that includes or is configured to access one or more computer-accessible media. FIG. 10 illustrates such a general-purpose computing device 9000. In the illustrated embodiment, computing device 9000 includes one or more processors 9010 coupled to a system memory 9020).
While Seeger teaches at least one resource, into multiple segments based on at least one resource demand level, Seeger does not specify a resource demand level threshold.
However, Seeger as cited does teach that time series data are divided between zeroes and non-zero entries. Accordingly, one of ordinary skill in the art would have interpreted this as a threshold as it is an evaluation criteria that results in the dissection of time series data. Therefore, Seeger reasonably teaches the limitation as claimed.
Regarding claim 2, Seeger teaches wherein aligning the forecasts for the two or more of the multiple segments to at least one time index associated with the at least one future time period comprises using at least one non-linear regression model (Col. 3, lines 24-32: In various embodiments, the forecaster may generate, with respect to the input data set, a statistical model which utilizes a likelihood function comprising one or more latent functions. At least one latent function may be a combination of a deterministic function and a random process. One such approach may combine, for example, a generalized linear model and probabilistic smoothing. In other embodiments, nonlinear deterministic functions such as various types of neural networks may be used.).
Regarding claim 3, Seeger teaches wherein segmenting resource demand time series data comprises segmenting the resource demand time series data into at least one segment containing resource demand data above the at least one resource demand level threshold and at least one segment containing resource demand data below the at least one resource demand level threshold, and wherein the at least one resource demand level threshold is derived from one or more descriptive statistical techniques (Col. 3, lines 4-11: a forecaster may determine that a particular input data set indicating demand observations for one or more items over a period of time meets an intermittency criterion that makes the data set suitable for a composite latent state model. The criterion may be based, for example, on the fraction of entries in the data set that are zeroes, the temporal distribution of the non-zero entries, and/or on other factors.; Col. 5, lines 34-35).
Regarding claim 4, Seeger teaches wherein generating forecasts for two or more of the multiple segments comprises extrapolating distribution values, in accordance with the at least one probability distribution, for the at least one future time period for each of the two or more of the multiple segments (Col. 3, lines 42-47: After the model parameters have been fitted, the model may be run to produce probabilistic demand forecasts—e.g., corresponding to various future times or time intervals, a probability distribution (and not just point predictions) of the demand for a given item may be generated).
Regarding claim 6, Seeger teaches wherein determining at least one probability distribution that fits at least a plurality of the multiple segments comprises determining at least one of multiple probability distributions that fits the at least a plurality of the multiple segments, wherein the multiple probability distributions comprises at least one gamma distribution, at least one Weibull distribution, at least one log-normal distribution, at least one normal distribution, and at least one logistic distribution (Col. 6, lines 20-33 : In some embodiments, a link function referred to as a “twice logistic” link function may be used with the GLM. The logistic function (on which this twice logistic function is based) may be defined as:
g(x)=log(1+e.sup.x) In the logistic function, the value of g(x) is positive for all x, and g(x) approaches x for large positive values of x. The twice logistic link function may be defined as:)
λ(y)=g(1+κ*g(y))) where κ is a constant (e.g., 0.0005) whose value may be selected based on experimentation. Commonly-employed link functions, such as λ(y)=g(y), or λ(y)=e.sup.y may not work well for intermittent data sets, and may even lead to the failure of the model in some cases.).
Regarding claim 8, Seeger teaches wherein performing one or more automated actions comprises automatically training at least a portion of the one or more machine learning techniques using feedback related to at least a portion of the resource demand forecast (Col. 4, lines 57-64: Free parameters may be fitted using the training subset and the client-preferred metrics may be used to determine the accuracy of the forecasts for the test subset. If the evaluations indicate that a given model does not meet a desired quality/accuracy criterion, the model may be adjusted in some embodiments—e.g., various initial parameters and/or features may be modified and the model may be retrained.).
Regarding claim 9, Seeger teaches wherein performing one or more automated actions comprises automatically initiating one or more resource implementation actions in accordance with the resource demand forecast (Col. 9, lines 36-43).
Regarding claim 10, it is a media/product type claim having similar limitations as claim 1 above. Therefore, it is rejected under the same rationale above.
Regarding claim 11, it is a media/product type claim having similar limitations as claim 2 above. Therefore, it is rejected under the same rationale above.
Regarding claim 12, it is a media/product type claim having similar limitations as claim 3 above. Therefore, it is rejected under the same rationale above.
Regarding claim 13, it is a media/product type claim having similar limitations as claim 4 above. Therefore, it is rejected under the same rationale above.
Regarding claim 15, it is a media/product type claim having similar limitations as claim 6 above. Therefore, it is rejected under the same rationale above.
Regarding claim 16, it is a system type claim having similar limitations as claim 1 above. Therefore, it is rejected under the same rationale above.
Regarding claim 17, it is a system type claim having similar limitations as claim 2 above. Therefore, it is rejected under the same rationale above.
Regarding claim 18, it is a system type claim having similar limitations as claim 3 above. Therefore, it is rejected under the same rationale above.
Regarding claim 19, it is a system type claim having similar limitations as claim 4 above. Therefore, it is rejected under the same rationale above.
Claims 5, 14, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Seeger et al. (US 10,748,072 B1) in further view of Rodrigues Vistulo De Abreu et al. (US 2015/0100525 A1) hereinafter “Rodrigues”.
Regarding claim 5, Seeger as cited above teaches probability distributions that fit at least a plurality of multiple segments but does not explicitly teach wherein determining at least one probability distribution that fits at least a plurality of the multiple segments comprises using one or more of at least one Kolmogorov-Smirnov test, at least one Anderson-Darling test, and at least one Cramer-von-Mises test.
However, Rodrigues teaches wherein determining at least one probability distribution that fits at least a plurality of the multiple segments comprises using one or more of at least one Kolmogorov-Smirnov test, at least one Anderson-Darling test ([0005] Another possible approach consists in using non-parametric tests--for instance, Kolmogorov-Smirnov, or Anderson-Darling--to evaluate if a sample from the signal deviates significantly from the behavior that could be predicted from the probability distribution characterizing the system's typical behavior.), and at least one Cramer-von-Mises test.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Rodrigues with the teachings of Seeger to use a K-S test or an Anderson-Darling test to evaluate deviations. The modification would have been motivated by the desire of combining known methods to yield predictable results.
Regarding claim 14, it is a media/product type claim having similar limitations as claim 5 above. Therefore, it is rejected under the same rationale above.
Regarding claim 20, it is a system type claim having similar limitations as claim 5 above. Therefore, it is rejected under the same rationale above
Claim 7 is rejected under 35 U.S.C. 103 as being unpatentable over Seeger et al. (US 10,748,072 B1) in further view of Yin et al. (US 2022/0058669 A1).
Regarding claim 7, Seeger does not explicitly teach wherein generating forecasts for two or more of the multiple segments comprises using at least one weighted average approach a weighted average approach in connection with two or more probability distributions.
However, Yin teaches wherein generating forecasts for two or more of the multiple segments comprises using at least one weighted average approach a weighted average approach in connection with two or more probability distributions ([0067] Overall, the present ensemble steps 1 and 2 refer an ensemble of machine learning models to generate an empirical cumulative probability distribution of the forecast. Thereafter, an optimal range of percentile is chosen based on Table 3 and the forecasts of different time scales are computed through Table 4 by a weighted average of the different percentiles of the empirical cumulative probability distribution.).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Yin with the teachings of Seeger to generate forecasted demand using a weighted average. The modification would have been motivated by the desire of combining known methods to yield predictable results.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JORGE A CHU JOY-DAVILA whose telephone number is (571)270-0692. The examiner can normally be reached Monday-Friday, 6:00am-5:00pm.
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.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Aimee J Li can be reached at (571)272-4169. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/JORGE A CHU JOY-DAVILA/Primary Examiner, Art Unit 2195
Prosecution Insights