In this paper, a novel method is proposed to solve the regression-based binary recommendation problem (RBR). The RBR is a univariate minimization problem, which has been solved recently using the Trichotomy approach w...
详细信息
In this paper, a novel method is proposed to solve the regression-based binary recommendation problem (RBR). The RBR is a univariate minimization problem, which has been solved recently using the Trichotomy approach with a logarithmic time complexity. The Trichotomy can obtain the global optimal minimum of univariate quasi-convex function. In this paper, it has been asserted that the objective function of the RBR problem is not necessarily quasi-convex;therefore, the Trichotomy method cannot always find its optimal solution that can be ultimately achieved by using the full search. Nonetheless, the difficulty of the full search method is that for each possible value of the variable of the problem, the objective function of the problem must be computed which involves the user-item tables to be reviewed entirely for each possible value of the variable. Since these tables are usually large, the runtime of the full search method is very high. In this paper, to unravel this difficulty, we made use of a recursive method to compute the objective function. Our proposed algorithm requires only one review of the user-item tables. Hence, the runtime of our proposed method is far less than the two other methods. (C) 2019 Elsevier Inc. All rights reserved.
暂无评论