Let matrix A be a user item matrix.Upon performing UV decomposition , I get a user factor matrix and factor entity matrix. The company I am interning at doesn't keep track of the user factor matrix.I also have access to the sales of the entities very month.
The company has to put new categories ( such people who like coke, people who like kfc) on their website for companies like (kfc,coke etc) to purchase audience data for advertising to them.
I get new factor entities matrix every week. The factors may mean completely different every week or may be similar as the entities they track vary slightly every week.I am expecting there to be at least some similarity as at least 50% of the entities they track do not change.
Therefore I have a entity factor matrix (say 1500 * 100) which varies every week and the sales of the entities every month.
My plan was to aggregate the matrices I get every week for a month and merge the sales if the entities with it. i.e. I will have a 1500*(400+1) (+1 for sales).Can I use transfer learning to build model such that if given an input of 1500*400 matrix of the next month it should be able to predict the sales.