Automatic machine learning driven compliance classification and determination
Automatic compliance using custom integrated machine learning is provided. For example, a system integrates one or more processors, coupled with memory, to receive a data file including digital images corresponding to statements. The system determines, using a text classification model, a type of the statements. The system identifies a section identification model relating to the type of the statements and annotated statements with labeled sections. The system determines, using the section identification model, a location and a label of a section in a first digital image corresponding to a first statement. The system extracts, based on the location and the label of the section, information from the first digital image. The system detects, based on a comparison of the extracted information with threshold information established for the label, a non-compliance with a procedure. The system performs an action responsive to detection of the non-compliance with the procedure.
1 . A system, comprising:
one or more processors, coupled with memory, to:
receive a data file comprising a plurality of digital images respectively corresponding to a plurality of statements;
determine, using one or more machine learning models, a type of a first pay statement of the plurality of statements in the data file;
identify a section identification model trained with machine learning relating to annotated statements with labeled sections for the type of the first pay statement;
determine, using the section identification model, a location and a label of a predetermined section in a first digital image of the first pay statement based on the type of the first pay statement;
extract, based on the location and the label of the predetermined section, information from the first digital image;
detect, based on a comparison of the extracted information with threshold information established for the label, a non-compliance with a procedure; and
perform an action responsive to detection of the non-compliance with the procedure.
2 . The system of claim 1 , wherein to determine the location, the one or more processors are further configured to:
determine coordinates in the first digital image corresponding to the predetermined section.
3 . The system of claim 1 , wherein the one or more processors are further configured to:
construct a data structure comprising the extracted information from the first digital image; and
annotate the extracted information in the data structure with the label.
4 . The system of claim 1 , wherein to extract the information, the one or more processors are further configured to:
use a support vector machine to identify a column in the predetermined section; and
extract the information based on the identified column.
5 . The system of claim 1 , wherein to detect the non-compliance, the one or more processors are further configured to:
detect an erroneous value in the extracted information.
6 . The system of claim 1 , wherein to detect the non-compliance, the one or more processors are further configured to:
determine a missing value in the extracted information.
7 . The system of claim 1 , wherein the one or more processors are further configured to:
extract second information from a second section at a second location in the first image having a second label; and
determine the non-compliance based on the extracted second information.
8 . The system of claim 1 , wherein to detect the non-compliance, the one or more processors are further configured to:
select the threshold information for the procedure based on the type of the plurality of statements and a geographic location relating to the first statement.
9 . The system of claim 1 , wherein to detect the non-compliance, the one or more processors are further configured to:
select the threshold information for the procedure based on at least one of a date of the first statement, a region code of the first statement, or a company code of the first statement.
10 . The system of claim 1 , wherein to perform the action, the one or more processors are further configured to:
provide, for display, a notification of the non-compliance.
11 . The system of claim 1 , wherein to perform the action, the one or more processors are further configured to:
generate a spreadsheet comprising an indication of the first statement or the predetermined section with the non-compliance.
12 . The system of claim 1 , wherein the one or more processors are further configured to:
generate a metric indicative of a level of compatibility of the plurality of statements.
13 . A method, comprising:
receiving, by one or more processors coupled with memory, a data file comprising a plurality of digital images respectively corresponding to a plurality of statements;
determining, by the one or more processors, using one or more machine learning models, a type of the plurality of statements;
identifying, by the one or more processors, a section identification model trained with machine learning relating to annotated statements with labeled sections for the type of the first pay statement;
determining, by the one or more processors using the section identification model, a location and a label of a predetermined section in a first digital image of the plurality of digital images corresponding to a first statement of the plurality of statements;
extracting, by the one or more processors based on the location and the label of the predetermined section, information from the first digital image;
detecting, by the one or more processors, based on a comparison of the extracted information with threshold information established for the label, a non-compliance with a procedure; and
performing, by the one or more processors, an action responsive to detection of the non-compliance with the procedure.
14 . The method of claim 13 , wherein determining the location comprises:
determining, by the one or more processors, coordinates in the first digital image corresponding to the predetermined section.
15 . The method of claim 13 , comprising:
constructing, by the one or more processors, a data structure comprising the extracted information from the first digital image; and
annotating, by the one or more processors, the extracted information in the data structure with the label.
16 . The method of claim 13 , wherein detecting the non-compliance comprises:
determining, by the one or more processors, a missing value in the extracted information.
17 . The method of claim 13 , wherein detecting the non-compliance comprises:
selecting, by the one or more processors, the threshold information for the procedure based on the type of the plurality of statements and a geographic location relating to the first statement.
18 . The method of claim 13 , comprising:
generating, by the one or more processors, a metric indicative of a level of compatibility of the plurality of statements.
19 . A non-transitory computer-readable medium storing processor executable instructions, that upon execution by one or more processors, cause the one or more processors to:
receive a data file comprising a plurality of digital images respectively corresponding to a plurality of statements;
determine, using one or more machine learning models, a type of the plurality of statements;
determine, using a section identification model, a location and a label of a section in a first digital image of the plurality of digital images corresponding to a first statement of the plurality of statements;
extract, based on the location and the label of the section, information from the first digital image;
detect, based on a comparison of the extracted information with threshold information established for the label, a non-compliance with a procedure; and
perform an action responsive to detection of the non-compliance with the procedure.
20 . The non-transitory computer-readable medium of claim 19 , wherein the instructions, upon execution, further cause the one or more processors to:
determine coordinates in the first digital image corresponding to the section.