Abstract
Our starting motivation is a user visiting an e-shop. E-shops usually offer conjunction of sharp filter conditions and one attribute ordering of results.
We use a top-k query system where results are ordered by a multi-criterial monotone combination of soft filter conditions. For prediction of users' behavior, we introduce a class of basis functions with positive Linear combination of Triangular (soft) filters (LT).
We prove that LT gives a unique representation of preferences. From database point of view LT act as a source for choosing pivots.
From business perspective LT reflect aggregation of users' (soft) ideal values (choice points). Our experiments use artificial data and are organized along variants of user's search habits, learning algorithms and evaluation measures.
We argue that LT recommendations behave better with respect to order sensitive measures. This gives raise a problem of pivot based indexing with order sensitive metrics.