Agile iteration for data mining using artificial intelligence

Abstract

Techniques for training an entity resolution model are presented. The techniques include receiving a minimum viable data product (MVDP) scope, a product scope, and a data mining scope from a user. A data mining goal is determined based on the MVDP scope, product scope, and data mining scope. One or more proof of concept (PoC) models are defined based on the data mining goal, and one of the PoC models is selected for training. A trained deep learning model is generated by iteratively training the selected PoC model. The trained deep learning model is then tested and validated against a predefined achievable loss metric using a sample labelled dataset for testing.

What is claimed is: 
     
         1 . A computer-implemented method performed by a server, the method comprising:
 receiving a minimum viable data product (MVDP) scope, a product scope, and a data mining scope from a user;   determining a data mining goal based on the MVDP scope, product scope, and data mining scope;   defining one or more proof of concept (PoC) models based on the data mining goal;   selecting one of the PoC models;   generating a trained deep learning model by iteratively training the selected PoC model via a deep learning model training program;   testing the trained deep learning model against a predefined achievable loss metric using a sample labelled dataset for testing; and   validating the trained deep learning using the achievable loss metric.   
     
     
         2 . The computer-implemented method in accordance with  claim 1 ,
 said determining the data mining goal comprises sourcing the training data and verifying the training data quality.   
     
     
         3 . The computer-implemented method in accordance with  claim 2 ,
 said determining the data mining goal further comprises receiving input data, the input data including an initial data collection report, a data description report, and a data exploration report.   
     
     
         4 . The computer-implemented method in accordance with  claim 1 ,
 said defining the one or more PoC models comprises:
 receiving the MVDP scope, product scope, data mining scope, the data mining goal, and a sample labelled dataset; and 
 outputting a finalized modelling technical approach for data mining, and one or more timelines associated with creation, training, evaluation, and deployment of the trained deep learning. 
   
     
     
         5 . The computer-implemented method in accordance with  claim 1 ,
 said iteratively training the selected PoC model via a deep learning model training program comprising:
 performing data preparation of an input dataset; 
 after the data preparation, building the trained deep learning using the input dataset; 
 evaluating the trained deep learning; 
 record the evaluation of the trained deep learning; 
 receive feedback used to inform one or more of a next iteration and the testing of the trained deep learning model. 
   
     
     
         6 . The computer-implemented method in accordance with  claim 1 ,
 said testing the trained deep learning model comprises:
 defining testing criteria; 
 receiving a sample labelled dataset for testing; 
 applying the sample labelled dataset to the trained deep learning model; and 
 outputting a performance testing report. 
   
     
     
         7 . The computer-implemented method in accordance with  claim 1 , further comprising determining the predefined achievable loss metric during iteratively training the selected PoC model via a deep learning model training program. 
     
     
         8 . A server comprising:
 one or more processors; and   a memory storing computer-executable instructions, that when executed by the one or more processors, cause the one or more processors to:
 receive a minimum viable data product (MVDP) scope, a product scope, and a data mining scope from a user; 
 determine a data mining goal based on the MVDP scope, product scope, and data mining scope; 
 define one or more proof of concept (PoC) models based on the data mining goal; 
 select one of the PoC models; 
 generate a trained deep learning model by iteratively training the selected PoC model via a deep learning model training program; 
 test the trained deep learning model against a predefined achievable loss metric using a sample labelled dataset for testing; and 
 validate the trained deep learning using the achievable loss metric. 
   
     
     
         9 . The server in accordance with  claim 8 ,
 said determining the data mining goal comprises the one or more processors sourcing the training data and verifying the training data quality.   
     
     
         10 . The server in accordance with  claim 9 ,
 said determining the data mining goal further comprises the one or more processors receiving input data, the input data including an initial data collection report, a data description report, and a data exploration report.   
     
     
         11 . The server in accordance with  claim 8 ,
 said defining the one or more PoC models comprises the one or more processors:
 receiving the MVDP scope, product scope, data mining scope, the data mining goal, and a sample labelled dataset; and 
 outputting a finalized modelling technical approach for data mining, and one or more timelines associated with creation, training, evaluation, and deployment of the trained deep learning. 
   
     
     
         12 . The server in accordance with  claim 8 ,
 said iteratively training the selected PoC model via a deep learning model training program comprising the one or more processors:
 performing data preparation of an input dataset; 
 after the data preparation, building the trained deep learning using the input dataset; 
 evaluating the trained deep learning; 
 record the evaluation of the trained deep learning; 
 receive feedback used to inform one or more of a next iteration and the testing of the trained deep learning model. 
   
     
     
         13 . The server in accordance with  claim 8 ,
 said testing the trained deep learning model comprises the one or more processors:
 defining testing criteria; 
 receiving a sample labelled dataset for testing; 
 applying the sample labelled dataset to the trained deep learning model; and 
 outputting a performance testing report. 
   
     
     
         14 . The server in accordance with  claim 8 ,
 said computer-executable instructions causing the one or more processors to determine the predefined achievable loss metric during iteratively training the selected PoC model via a deep learning model training program.   
     
     
         15 . A non-transitory computer-readable storage medium having computer-executable instructions stored thereon, the computer-executable instructions, when executed by one or more processors, causing the one or more processors to:
 receive a minimum viable data product (MVDP) scope, a product scope, and a data mining scope from a user;   determine a data mining goal based on the MVDP scope, product scope, and data mining scope;   define one or more proof of concept (PoC) models based on the data mining goal;   select one of the PoC models;   generate a trained deep learning model by iteratively training the selected PoC model via a deep learning model training program;   test the trained deep learning model against a predefined achievable loss metric using a sample labelled dataset for testing; and   validate the trained deep learning using the achievable loss metric.   
     
     
         16 . The non-transitory computer-readable storage medium in accordance with  claim 15 ,
 said determining the data mining goal comprises the one or more processors sourcing the training data and verifying the training data quality.   
     
     
         17 . The non-transitory computer-readable storage medium in accordance with  claim 16 ,
 said determining the data mining goal further comprises the one or more processors receiving input data, the input data including an initial data collection report, a data description report, and a data exploration report.   
     
     
         18 . The non-transitory computer-readable storage medium in accordance with  claim 15 ,
 said defining the one or more PoC models comprises the one or more processors:
 receiving the MVDP scope, product scope, data mining scope, the data mining goal, and a sample labelled dataset; and 
 outputting a finalized modelling technical approach for data mining, and one or more timelines associated with creation, training, evaluation, and deployment of the trained deep learning. 
   
     
     
         19 . The non-transitory computer-readable storage medium in accordance with  claim 15 ,
 said iteratively training the selected PoC model via a deep learning model training program comprising the one or more processors:
 performing data preparation of an input dataset; 
 after the data preparation, building the trained deep learning using the input dataset; 
 evaluating the trained deep learning; 
 record the evaluation of the trained deep learning; 
 receive feedback used to inform one or more of a next iteration and the testing of the trained deep learning model. 
   
     
     
         20 . The non-transitory computer-readable storage medium in accordance with  claim 15 ,
 said testing the trained deep learning model comprises the one or more processors:
 defining testing criteria; 
 receiving a sample labelled dataset for testing; 
 applying the sample labelled dataset to the trained deep learning model; and 
 outputting a performance testing report.