# US 12,001,840 B1 · Likelihood ratio test-based approach for detecting data entry errors > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,001,840 B1 - Document ID: US12001840B1 - Status: Granted - Kind: B1 - Application number: 18/122,627 - Filed: 2023-03-16 - Granted: 2024-06-04 - Earliest claimed: 2023-03-16 - Assignees: Intuit Inc. - Applicants: Intuit, Inc. - Inventors: Arkadeep Banerjee; Vignesh T. Subrahmaniam - Web page: https://patentgrep.com/patent/US12001840B1 ## CPC - G06F 9/3001: {Arithmetic instructions} - G06F 9/545: {where tasks reside in different layers, e.g. user- and kernel-space} - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) ## Abstract Certain aspects of the present disclosure provide techniques for detecting data errors. A method generally includes receiving a new value as user input for a data field, generating histogram-based approximations of kernel density estimates generated based on valid data and invalid data associated with the data field, determining a likelihood that the new value is a valid value and a likelihood that the new value is an invalid value, computing a likelihood ratio test statistic based on a ratio of the likelihoods, and classifying the new value as valid or invalid based on comparing the likelihood ratio test statistic to a likelihood ratio test threshold. ## Claims ### Claim 1 A computer-implemented method of detecting data errors, comprising: receiving a new value as user input for a data field; generating a first histogram-based approximation of a first kernel density estimate generated based on valid data associated with the data field and a second histogram-based approximation of a second kernel density estimate generated based on invalid data associated with the data field; determining a first likelihood that the new value is valid, wherein the first likelihood is equal to a first probability density of a first bin of the first histogram-based approximation that includes a log ratio of the new value to a mean value associated with the data field; determining a second likelihood that the new value is invalid, wherein the second likelihood is equal to a second probability density of a second bin of the second histogram-based approximation that includes the log ratio of the new value to the mean value associated with the data field; computing a likelihood ratio test statistic based on a ratio of the first likelihood that the new value is valid to the second likelihood that the new value is invalid; classifying the new value as valid or invalid based on comparing the likelihood ratio test statistic to a likelihood ratio test threshold; and when the new value is classified as invalid, taking one or more actions to correct the new value. ### Claim 2 (depends on 1) The computer-implemented method of claim 1, further comprising, prior to receiving the new value as the user input for the data field: determining mean values for the valid data associated with the data field; determining a log ratio of each individual valid datum in the valid data to a corresponding mean value of the mean values determined for the valid data; generating invalid data associated with the data field; determining a log ratio of each individual invalid datum in the invalid data to a corresponding mean value of the mean values determined for the valid data; and generating the first kernel density estimate based on applying a kernel function to the log ratio of each individual valid datum in the valid data; and generating the second kernel density estimate based on applying the kernel function to the log ratio of each individual invalid datum in the invalid data. ### Claim 3 (depends on 2) The computer-implemented method of claim 2, wherein the mean values for the valid data and the mean values for the invalid data are determined via a simple moving average technique, a weighted moving average technique, an exponential weighting technique, or an exponential smoothing technique. ### Claim 4 (depends on 2) The computer-implemented method of claim 2, wherein the kernel function comprises a Gaussian kernel. ### Claim 5 (depends on 2) The computer-implemented method of claim 2, further comprising: determining the mean value associated with the data field; and computing the log ratio of the new value to the mean value. ### Claim 6 (depends on 1) The computer-implemented method of claim 1, wherein at least one of the first histogram-based approximation or the second histogram-based approximation comprises variable bin width sizes. ### Claim 7 (depends on 1) The computer-implemented method of claim 1, wherein the likelihood ratio test statistic is a type 1 error value. ### Claim 8 (depends on 1) The computer-implemented method of claim 1, wherein the new value, the valid data, and the invalid data are associated with at least one of a same person, a same company, or a same item. ### Claim 9 (depends on 1) The computer-implemented method of claim 1, wherein taking the one or more actions to correct the new value comprises at least one of: indicating, to a user, that the new value is invalid; or correcting the new value based on the valid data associated with the data field. ### Claim 10 An apparatus, comprising: a memory comprising computer-executable instructions; and a processor in data communication with the memory and configured to execute the computer-executable instructions and cause the apparatus to perform a method for detecting data entry errors, the method comprising: receiving a new value as user input for a data field; generating a first histogram-based approximation of a first kernel density estimate generated based on valid data associated with the data field and a second histogram-based approximation of a second kernel density estimate generated based on invalid data associated with the data field; determining a first likelihood that the new value is valid, wherein the first likelihood is equal to a first probability density of a first bin of the first histogram-based approximation that includes a log ratio of the new value to a mean value associated with the data field; determining a second likelihood that the new value is invalid, wherein the second likelihood is equal to a second probability density of a second bin of the second histogram-based approximation that includes the log ratio of the new value to the mean value associated with the data field; computing a likelihood ratio test statistic based on a ratio of the first likelihood that the new value is valid to the second likelihood that the new value is invalid; classifying the new value as valid or invalid based on comparing the likelihood ratio test statistic to a likelihood ratio test threshold; and when the new value is classified as invalid, taking one or more actions to correct the new value. ### Claim 11 (depends on 10) The apparatus of claim 10, wherein the method further comprises, prior to receiving the new value as the user input for the data field: determining mean values for the valid data associated with the data field; determining a log ratio of each individual valid datum in the valid data to a corresponding mean value of the mean values determined for the valid data; generating invalid data associated with the data field; determining a log ratio of each individual invalid datum in the invalid data to a corresponding mean value of the mean values determined for the valid data; and generating the first kernel density estimate based on applying a kernel function to the log ratio of each individual valid datum in the valid data; and generating the second kernel density estimate based on applying the kernel function to the log ratio of each individual invalid datum in the invalid data. ### Claim 12 (depends on 11) The apparatus of claim 11, wherein the mean values for the valid data and the mean values for the invalid data are determined via a simple moving average technique, a weighted moving average technique, an exponential weighting technique, or an exponential smoothing technique. ### Claim 13 (depends on 11) The apparatus of claim 11, wherein the kernel function comprises a Gaussian kernel. ### Claim 14 (depends on 11) The apparatus of claim 11, wherein the method further comprises: determining the mean value associated with the data field; and computing the log ratio of the new value to the mean value. ### Claim 15 (depends on 10) The apparatus of claim 10, wherein taking the one or more actions to correct the new value comprises at least one of: indicating, to a user, that the new value is invalid; or correcting the new value based on the valid data associated with the data field. ### Claim 16 (depends on 10) The apparatus of claim 10, wherein at least one of the first histogram-based approximation or the second histogram-based approximation comprises variable bin width sizes. ### Claim 17 (depends on 10) The apparatus of claim 10, wherein the likelihood ratio test statistic is a type 1 error value. ### Claim 18 (depends on 10) The apparatus of claim 10, wherein the new value, the valid data, and the invalid data are associated with at least one of a same person, a same company, or a same item.