Feature enhancement via unsupervised learning of external knowledge embedding

1 . A processor-implemented method for improving feature engineering utilizing unsupervised learning of associated external knowledge embedding without manual feature engineering from a data scientist, the method comprising:

receiving, by a processor, input data as a table and a name of a column;

identifying within the column multisets of concepts or sequences of concepts represented in natural language within the column;

creating by the processor an expanded column by automatically expanding by the processor the column of the table by linking the identified natural language multisets of concepts or the sequences of the concepts with corresponding concepts in an external knowledge graph associated with an external knowledge base of specialized knowledge;

training by the processor a neural network to learn embedding vectors associated with the concepts in the expanded column of the tables, wherein the training is unsupervised without provision of labels of data when the neural network learns the embedding vectors of the concepts in the expanded columns, the learned embedding vectors minimizing a reconstruction error of the concepts in the expanded columns; and

utilizing by the processor the trained neural network and the learnt embedding vectors in a predictive analytic task or a data mining task associated with the natural language multisets or the sequences of the concepts.

2 . The method of claim 1 , further comprising:

prompting a user to select a column from the one or more columns that the user wants to expand.

3 . The method of claim 1 , further comprising:

receiving the external knowledge graph based on classification of the input data and training dataset into a concept.

4 . The method of claim 1 , further comprising:

extracting related concepts from the external knowledge graph.

5 . The method of claim 1 , further comprising:

utilizing a neural network structure to take a multi-set of data as input and output the embedding vector of the multi-set of the data.

6 . The method of claim 1 , further comprising:

utilizing an unsupervised learning method that trains a neural network structure on the expanded column by forcing the neural network structure to predict likelihood of items presenting in a given data set.

7 . The method of claim 6 , further comprising:

applying the unsupervised learning method to data clustering, anomaly detection and supervised problems, wherein the unsupervised learning method utilizes numerical vector representations of each set of the multi-set as an additional feature in addition to existing numerical feature in the table.

8 . A computer system for improving feature engineering utilizing unsupervised learning of associated external knowledge embedding without manual feature engineering from a data scientist, the computer system comprising:

one or more processors, one or more computer-readable memories, one or more computer-readable tangible storage media, and program instructions stored on at least one of the one or more tangible storage media for execution by at least one of the one or more processors via at least one of the one or more memories, wherein the computer system is capable of performing a method comprising:

receiving, by a processor, input data as a table and a name of a column;

identifying within the column multisets of concepts or sequences of concepts represented in natural language within the column;

creating by the processor an expanded column by automatically expanding by the processor the column of the table by linking the identified natural language multisets of concepts or the sequences of the concepts with corresponding concepts in an external knowledge graph associated with an external knowledge base of specialized knowledge;

training by the processor a neural network to learn embedding vectors associated with the concepts in the expanded column of the tables, wherein the training is unsupervised without provision of labels of data when the neural network learns the embedding vectors of concepts in the expanded columns, the learned embedding vectors minimizing reconstruction error of the of concepts in the expanded columns; and

utilizing by the processor the trained neural network and the learnt embedding vectors in a predictive analytic task or a data mining task associated with the natural language multisets or the sequences of the concepts.

9 . The computer system of claim 8 , further comprising:

prompting a user to select a column from the one or more columns that the user wants to expand.

10 . The computer system of claim 8 , further comprising:

receiving the external knowledge graph based on classification of the input data and training dataset into a concept.

11 . The computer system of claim 8 , further comprising:

extracting related concepts from the external knowledge graph.

12 . The computer system of claim 8 , further comprising:

utilizing a neural network structure to take a multi-set of data as input and output an embedding vector of the multi-set of the data.

13 . The computer system of claim 8 , further comprising:

utilizing an unsupervised learning method that trains a neural network structure on the expanded column by forcing the neural network structure to predict likelihood of items presenting in a given data set.

14 . The computer system of claim 13 , further comprising:

applying the unsupervised learning method to data clustering, anomaly detection and supervised problems, wherein the unsupervised learning method utilizes numerical vector representations of each set of the multi-set as an additional feature in addition to existing numerical feature in the table.

15 . A computer program product for improving feature engineering utilizing unsupervised learning of associated external knowledge embedding without manual feature engineering from a data scientist, the computer program product comprising:

one or more computer-readable tangible storage media and program instructions stored on at least one of the one or more tangible storage media, the program instructions executable by a processor of a computer to perform a method, the method comprising:

receiving, by a processor, input data as a table and a name of a column;

identifying within the column multisets of concepts or sequences of concepts represented in natural language within the column;

creating by the processor an expanded column by automatically expanding by the processor the column of the table by linking the identified natural language multisets of concepts or the sequences of the concepts with corresponding concepts in an external knowledge graph associated with an external knowledge base of specialized knowledge;

training by the processor a neural network to learn embedding vectors associated with the concepts of in the expanded column of the tables, wherein the training is unsupervised without provision of labels of data when the neural network learns the embedding vectors of the concepts in the expanded columns, the learned embedding vectors minimizing a reconstruction error of the of concepts in the expanded columns; and

utilizing by the processor the trained neural network and the learnt embedding vectors in a predictive analytic task or a data mining task associated with the natural language multisets or the sequences of the concepts.

16 . The computer program product of claim 15 , further comprising:

prompting a user to select a column from the one or more columns that the user wants to expand.

17 . The computer program product of claim 15 , further comprising:

receiving the external knowledge graph based on classification of the input data and training dataset into a concept.

18 . The computer program product of claim 15 , further comprising:

utilizing a neural network structure to take a multi-set of data as input and output an embedding vector of the multi-set of the data.

19 . The computer program product of claim 15 , further comprising:

utilizing an unsupervised learning method that trains a neural network structure on the expanded column by forcing the neural network structure to predict likelihood of items presenting in a given data set.

20 . The computer program product of claim 19 , further comprising:

applying the unsupervised learning method to data clustering, anomaly detection and supervised problems, wherein the unsupervised learning method utilizes numerical vector representations of each set of the multi-set as an additional feature in addition to existing numerical feature in the table.