Feature learning on customer journey using categorical sequence data

1. A method, comprising:

converting, by a computer having a processor and a memory, categorical sequence data for an actual customer journey taken by an actual customer into a numerical similarity matrix; and

learning, by the computer, features of the customer journey by applying a distance metric learning based matrix factorization approach to the numerical similarity matrix, the distance metric learning based matrix factorization approach involving forming a n*d binary matrix using an objective function, where n is a number of categorical sequence patterns, and d is a dimensionality of the binary matrix, wherein each dimension in the binary matrix represents a hidden factor from a set of hidden factors,

wherein the categorical sequence data is non-numerical and lacks an explicit ordering.

2. The method of claim 1 , wherein the features are numerical.

3. The method of claim 1 , further comprising forming a computer learning system by applying a machine learning process to the features.

4. The method of claim 1 , further comprising classifying an object by applying a classification model to the features.

5. The method of claim 1 , further comprising clustering a set of objects by applying a clustering model to the features.

6. The method of claim 1 , further comprising performing a regression operation to generate a relationship estimate between at least two variables by applying a regression model to the features.

7. The method of claim 1 , wherein the distance metric learning based matrix factorization approach automatically learns hidden factors that influence patterns in the categorical sequence data based on the numerical similarity matrix.

8. The method of claim 1 , wherein the distance metric learning based matrix factorization approach automatically learns relationships between different patterns in the categorical sequence data based on the numerical similarity matrix.

9. The method of claim 1 , wherein the distance metric learning based matrix factorization approach automatically learns implicit sequence information in the categorical sequence data based on the numerical similarity matrix.

10. The method of claim 1 , wherein the numerical similarity matrix is formed based on respective numbers of common subsequences between patterns in the categorical sequence data for the customer journey of two users.

11. The method of claim 10 , wherein the common subsequences comprise a plurality of subsequences having different lengths.

12. The method of claim 11 , wherein a similarity is determined for the numerical similarity matrix based on a weighted average of the plurality of subsequences having the different lengths.

13. The method of claim 1 , wherein each entry in the binary matrix indicates whether a given respective hidden factor from the set influences a particular pattern from among a plurality of patterns in the categorical sequence data.

14. The method of claim 13 , wherein an objective function measures a relationship between the given respective hidden factor and the particular pattern.

15. A non-transitory computer readable storage medium comprising a computer readable program for feature learning, wherein the computer readable program when executed on a computer causes the computer to perform the steps of:

converting, by a computer having a processor and a memory, categorical sequence data for an actual customer journey taken by an actual customer into a numerical similarity matrix; and

learning, by the computer, features of the customer journey by applying a distance metric learning based matrix factorization approach to the numerical similarity matrix, the distance metric learning based matrix factorization approach involving forming a n*d binary matrix using an objective function, where n is a number of categorical sequence patterns, and d is a dimensionality of the binary matrix, wherein each dimension in the binary matrix represents a hidden factor from a set of hidden factors,

wherein the categorical sequence data is non-numerical and lacks an explicit ordering.

16. The non-transitory computer readable storage medium of claim 15 , wherein the distance metric learning based matrix factorization approach automatically learns hidden factors that influence patterns in the categorical sequence data based on the numerical similarity matrix.

17. The non-transitory computer readable storage medium of claim 15 , wherein the distance metric learning based matrix factorization approach automatically learns relationships between different patterns in the categorical sequence data based on the numerical similarity matrix.

18. The non-transitory computer readable storage medium of claim 15 , wherein the numerical similarity matrix is formed based on respective numbers of common subsequences between patterns in the categorical sequence data for the customer journey of two users.

19. The non-transitory computer readable storage medium of claim 18 , wherein the common subsequences comprise a plurality of subsequences having different lengths.

20. The non-transitory computer readable storage medium of claim 19 , wherein a similarity is determined for the numerical similarity matrix based on a weighted average of the plurality of subsequences having the different lengths.

21. The non-transitory computer readable storage medium of claim 15 , wherein each entry in the binary matrix indicates whether a given respective hidden factor from the set influences a particular pattern from among a plurality of patterns in the categorical sequence data.

22. A system, comprising:

a computer, having a processor and a memory, configured to:

convert categorical sequence data for an actual customer journey taken by an actual customer into a numerical similarity matrix; and

learn features of the customer journey by applying a distance metric learning based matrix factorization approach to the numerical similarity matrix, the distance metric learning based matrix factorization approach involving forming a n*d binary matrix using an objective function, where n is a number of categorical sequence patterns, and d is a dimensionality of the binary matrix, wherein each dimension in the binary matrix represents a hidden factor from a set of hidden factors,

wherein the categorical sequence data is non-numerical and lacks an explicit ordering.

23. The system of claim 22 , wherein the computer is implemented as a server using a cloud computing configuration.