Automated return evaluation with anomoly detection
Media, methods, and systems are disclosed for applying a computer-implemented model to a table of computed values to identify one or more anomalies. One or more input forms having a plurality of input form field values is received. The input form field values are automatically parsed into a set of computer-generated candidate standard field values. The set of candidate standard field values are automatically normalized into a corresponding set of normalized field values, based on a computer-automated input normalization model. An automated review model controller is applied to automatically identify a review model to apply to the set of normalized field values, based on certain predetermined target field values. The automatically identified review model is then applied to the set of normalized inputs, and in response to detecting an anomaly, a field value is flagged accordingly.
1. One or more non-transitory computer-readable media storing computer-executable instructions that, when executed by a processor, perform a method for applying a computer-implemented model to a table of computed values to identify one or more anomalies in the table of the computed values, the method comprising:
receiving one or more input forms, each of the one or more input forms comprising a plurality of input form field values,
wherein each of the one or more input forms is text-based ad-hoc, arbitrary, non-preprocessed tax data corresponding to a particular set of input forms belonging to a particular taxpayer;
automatically parsing the plurality of input form field values into a set of computer-generated candidate standard field values;
automatically normalizing the set of computer-generated candidate standard field values into a set of normalized field values, based on a computer-automated input normalization model,
wherein the computer-automated input normalization model normalizes the set of computer-generated candidate standard field values based on one or more of: metadata based on existing known-correct tax returns or regulatory agency data sources;
applying an automated review model controller to automatically select a review model to apply to the set of normalized field values, based on a predetermined one or more target field values in the set of normalized field values, the review model being selected based on a best-fitting neural network providing a best performance based on iteratively applying known training data and evaluating performance of the review model in connection with the known training data;
applying the review model to the set of normalized field values;
in response to detecting the one or more anomalies in the set of normalized field values, flagging a detected anomalous field value in the set of normalized field value; and
responsive to flagging the detected anomalous field value, predicting a subject matter expert recommendation based on the detected anomalous field value.
2. The one or more non-transitory computer-readable media of claim 1 , wherein the best-fitting neural network is selected from a plurality of candidate neural networks, the best-fitting neural network providing the best performance based on using the known training data in connection with the plurality of candidate neural networks.
3. The one or more non-transitory computer-readable media of claim 1 , the method further comprising:
providing one or more predetermined known target field values to the automated review model controller; and
in response to determining that the review model is not a known correct review model, updating review model metadata associated with the automated review model controller to enable the automated review model controller to identify the known correct review model.
4. The one or more non-transitory computer-readable media of claim 1 , wherein the method further comprises:
collapsing the set of normalized field values into one or more data frames; and
applying the review model to the one or more data frames to flag an alternative possible anomaly.
5. The one or more non-transitory computer-readable media of claim 1 , wherein applying the review model to the set of normalized field values further comprises:
further applying a set of subsequent review models to the set of normalized field values to determine whether a subsequent review model in the set of subsequent review models detects a subsequent anomaly.
6. The one or more non-transitory computer-readable media of claim 5 , wherein the set of subsequent review models are trained on actual anomalies received as training data.
7. A method for applying a computer-implemented model to a table of computed values to identify one or more anomalies in the table of the computed values, the method comprising:
receiving one or more input forms, each of the one or more input forms comprising a plurality of input form field values,
wherein each of the one or more input forms is text-based ad-hoc, arbitrary, non-preprocessed tax data corresponding to a particular set of input forms belonging to a particular taxpayer;
automatically parsing the plurality of input form field values into a set of computer-generated candidate standard field values;
automatically normalizing the set of computer-generated candidate standard field values into a set of normalized field values, based on a computer-automated input normalization model,
wherein the computer-automated input normalization model normalizes the set of computer-generated candidate standard field values based on one or more of: metadata based on existing known-correct tax returns or regulatory agency data sources;
applying an automated review model controller to automatically select a review model to apply to the set of normalized field values, based on a predetermined one or more target field values in the set of normalized field values, the review model being selected based on a best-fitting neural network providing a best performance based on iteratively applying known training data and evaluating performance of the review model in connection with the known training data;
applying the review model to the set of normalized field values;
in response to detecting the one or more anomalies in the set of normalized field values, flagging a detected anomalous field value in the set of normalized field value; and
responsive to flagging the detected anomalous field value, predicting a subject matter expert recommendation based on the detected anomalous field value.
8. The method of claim 7 , wherein the best-fitting neural network is selected from a plurality of candidate neural networks, the best-fitting neural network providing the best performance based on using the known training data in connection with the plurality of candidate neural networks.
9. The method of claim 7 further comprising:
providing one or more predetermined known target field values to the automated review model controller; and
in response to determining that the review model is not a known correct review model, updating review model metadata associated with the automated review model controller to enable the automated review model controller to identify the known correct review model.
10. The method of claim 7 , further comprising:
collapsing the set of normalized field values into one or more data frames; and
applying the review model to the one or more data frames to flag an alternative possible anomaly.
11. The method of claim 7 , wherein applying the review model to the set of normalized field values further comprises:
further applying a set of subsequent review models to the set of normalized field values to determine whether a subsequent review model in the set of subsequent review models detects a subsequent anomaly.
12. The method of claim 11 , wherein the set of subsequent review models are trained on actual anomalies received as training data.
13. A system comprising at least one processor and at least one non-transitory memory storing computer executable instructions that when executed by the at least one processor cause the system to carry out actions comprising:
receiving one or more input forms, each of the one or more input forms comprising a plurality of input form field values,
wherein each of the one or more input forms is text-based ad-hoc, arbitrary, non-preprocessed tax data corresponding to a particular set of input forms belonging to a particular taxpayer;
automatically parsing the plurality of input form field values into a set of computer-generated candidate standard field values;
automatically normalizing the set of computer-generated candidate standard field values into a set of normalized field values, based on a computer-automated input normalization model,
wherein the computer-automated input normalization model normalizes the set of computer-generated candidate standard field values based on one or more of: metadata based on existing known-correct tax returns or regulatory agency data sources;
applying an automated review model controller to automatically select a review model to apply to the set of normalized field values, based on a predetermined one or more target field values in the set of normalized field values, the review model being selected based on a best-fitting neural network providing a best performance based on iteratively applying known training data and evaluating performance of the review model in connection with the known training data;
applying the review model to the set of normalized field values;
in response to detecting the one or more anomalies in the set of normalized field values, flagging a detected anomalous field value in the set of normalized field value; and
responsive to flagging the detected anomalous field value, predicting a subject matter expert recommendation based on the detected anomalous field value.
14. The system of claim 13 , wherein the best-fitting neural network is selected from a plurality of candidate neural networks, the best-fitting neural network providing the best performance based on using the known training data in connection with the plurality of candidate neural networks.
15. The system of claim 14 , the actions further comprising:
providing one or more predetermined known target field values to the automated review model controller; and
in response to determining that the review model is not a known correct review model, updating review model metadata associated with the automated review model controller to enable the automated review model controller to identify the known correct review model.
16. The system of claim 15 , wherein applying the review model to the set of normalized field values further comprises:
further applying a set of subsequent review models to the set of normalized field values to determine whether a subsequent review model in the set of subsequent review models detects a subsequent anomaly.
17. The system of claim 16 , wherein the set of subsequent review models are trained on actual anomalies received as training data.
18. The one or more non-transitory computer-readable media of claim 1 , further comprising:
responsive to detecting the one or more anomalies in the set of normalized field values, determining a certainty factor for the one or more anomalies, wherein the certainty factor is a ratio or percentage associated with a level of certainty of detection of an anomaly.
19. The method of claim 7 further comprising:
responsive to detecting the one or more anomalies in the set of normalized field values, determining a certainty factor for the one or more anomalies, wherein the certainty factor is a ratio or percentage associated with a level of certainty of detection of an anomaly.
20. The system of claim 13 , the actions further comprising:
responsive to detecting the one or more anomalies in the set of normalized field values, determining a certainty factor for the one or more anomalies, wherein the certainty factor is a ratio or percentage associated with a level of certainty of detection of an anomaly.