The FACTMAC Procedure

The FACTMAC procedure implements the factorization machine model in SAS Viya. The flexible factorization machine model has applications in predictive modeling and recommendation (Rendle 2012). Factorization machines generalize matrix factorization, among other techniques. You can use the FACTMAC procedure to read and write data in distributed form, and to perform factorization in parallel by making full use of multicore computers or distributed computing environments.

The FACTMAC procedure estimates factors for each of the nominal input variables you specify, in addition to estimating a global bias and a bias for each level of those nominal input variables. You also specify an interval target variable. The procedure computes the biases and factors by using the stochastic gradient descent (SGD) algorithm, which minimizes the root mean square error (RMSE) criterion on the input data table that you provide. In this method, each iteration attempts to reduce the RMSE. The SGD algorithm proceeds until the maximum number of iterations is reached.

PROC FACTMAC stores the results of the factorization an output data table, which is produced by the OUTMODEL= option. This data table contains the factors in addition to the global bias and the biases for all the levels of the input variables, in addition to the factors. The corresponding level names are listed for ease of reference. The biases and factors are used for scoring.