Non-parametric optimal service pricing: a simulation study
Citations
12 citations
Cites background from "Non-parametric optimal service pric..."
...Muzaffar, Deng, and Rashid (2016) also designed a nonparametric optimal service pricing....
[...]
2 citations
References
2,141 citations
"Non-parametric optimal service pric..." refers background in this paper
...Kiefer and Wolfowitz (1952) estimate the point of maximum when the function is unknown and stochastic approximation algorithm proposed uses simulation-based optimization....
[...]
1,537 citations
"Non-parametric optimal service pric..." refers background in this paper
...These assumptions are standard in literature and such demand functions are called regular demand functions (Gallego & Ryzin, 1994)....
[...]
229 citations
"Non-parametric optimal service pric..." refers background in this paper
...For extensive research related to pricing and/or inventory decisions in Bayesian framework, see Araman and Caldentey (2009), etc. and the reference therein....
[...]
...For extensive research related to pricing and/or inventory decisions in Bayesian framework, see Araman and Caldentey (2009), etc. and the reference therein. There is plethora of literature on non-parametric approach in dynamic pricing. Ryzin and Mc Gill (2000) analyse adaptive algorithms for determining airline seat protection levels. In their model, the distributions of customers’ requests for each class are non-parametric. Besbes and Zeevi (2006) introduce a blind non-parametric approach for searching near optimal price setting under unknown demand pattern. Their proposed algorithm is asymptotically optimal. The decision-maker observes demand over time but is otherwise blind to the demand function which maps prices. Structural non-parametric assumptions are made for demand function and demand rate at any instant is estimated by multivariate Poisson process and intensity is determined by price vectors which denotes the set of feasible prices. The authors use non-parametric price algorithm which experiments with several feasible prices and selects price that is expected to optimize revenue based on the demand observed in the previous step. The advantage is the use of multiple products which are linked with various resource constraints contrary to previous approaches which were restricted to single product problem. Rusmevichientong, Roy, and Glynn (2006) formulate a multiproduct single resource pricing problem as a static optimization problem in which a car manufacturer seeks fixed prices that maximizes revenues based on historical preference data. Lim and Shanthikumar (2007) use a max–min approach to a single-product revenue management problem where the uncertainty arises at the level of the point process distribution characterizing the customers’ requests. Other min–max approaches to modeling uncertainty include minimax regret (see Eren, Maglaras, and Ryzin (2006) and Ball and Queyranne (2006) for references in the revenue management literature)....
[...]
...For extensive research related to pricing and/or inventory decisions in Bayesian framework, see Araman and Caldentey (2009), etc. and the reference therein. There is plethora of literature on non-parametric approach in dynamic pricing. Ryzin and Mc Gill (2000) analyse adaptive algorithms for determining airline seat protection levels. In their model, the distributions of customers’ requests for each class are non-parametric. Besbes and Zeevi (2006) introduce a blind non-parametric approach for searching near optimal price setting under unknown demand pattern. Their proposed algorithm is asymptotically optimal. The decision-maker observes demand over time but is otherwise blind to the demand function which maps prices. Structural non-parametric assumptions are made for demand function and demand rate at any instant is estimated by multivariate Poisson process and intensity is determined by price vectors which denotes the set of feasible prices. The authors use non-parametric price algorithm which experiments with several feasible prices and selects price that is expected to optimize revenue based on the demand observed in the previous step. The advantage is the use of multiple products which are linked with various resource constraints contrary to previous approaches which were restricted to single product problem. Rusmevichientong, Roy, and Glynn (2006) formulate a multiproduct single resource pricing problem as a static optimization problem in which a car manufacturer seeks fixed prices that maximizes revenues based on historical preference data....
[...]
...For extensive research related to pricing and/or inventory decisions in Bayesian framework, see Araman and Caldentey (2009), etc. and the reference therein. There is plethora of literature on non-parametric approach in dynamic pricing. Ryzin and Mc Gill (2000) analyse adaptive algorithms for determining airline seat protection levels. In their model, the distributions of customers’ requests for each class are non-parametric. Besbes and Zeevi (2006) introduce a blind non-parametric approach for searching near optimal price setting under unknown demand pattern. Their proposed algorithm is asymptotically optimal. The decision-maker observes demand over time but is otherwise blind to the demand function which maps prices. Structural non-parametric assumptions are made for demand function and demand rate at any instant is estimated by multivariate Poisson process and intensity is determined by price vectors which denotes the set of feasible prices. The authors use non-parametric price algorithm which experiments with several feasible prices and selects price that is expected to optimize revenue based on the demand observed in the previous step. The advantage is the use of multiple products which are linked with various resource constraints contrary to previous approaches which were restricted to single product problem. Rusmevichientong, Roy, and Glynn (2006) formulate a multiproduct single resource pricing problem as a static optimization problem in which a car manufacturer seeks fixed prices that maximizes revenues based on historical preference data. Lim and Shanthikumar (2007) use a max–min approach to a single-product revenue management problem where the uncertainty arises at the level of the point process distribution characterizing the customers’ requests. Other min–max approaches to modeling uncertainty include minimax regret (see Eren, Maglaras, and Ryzin (2006) and Ball and Queyranne (2006) for references in the revenue management literature). Kiefer and Wolfowitz (1952) estimate the point of maximum when the function is unknown and stochastic approximation algorithm proposed uses simulation-based optimization. The proposed Kiefer–Wolfowitz (KW) algorithm uses sequential procedure for stochastic approximation depending on four conditions. The algorithm does not require existence of derivative but optimality of tuning sequence in some reasonable sense is required for the convergence of the algorithm. Though sequence converges stochastically but the proposed algorithm says nothing about the optimal values of the sequence parameters and the stopping rule. Broadie, Deniz, and Assaf (2009) derive general bound on the mean squared error and most importantly the optimal choice of tuning sequences....
[...]
...For extensive research related to pricing and/or inventory decisions in Bayesian framework, see Araman and Caldentey (2009), etc. and the reference therein. There is plethora of literature on non-parametric approach in dynamic pricing. Ryzin and Mc Gill (2000) analyse adaptive algorithms for determining airline seat protection levels. In their model, the distributions of customers’ requests for each class are non-parametric. Besbes and Zeevi (2006) introduce a blind non-parametric approach for searching near optimal price setting under unknown demand pattern. Their proposed algorithm is asymptotically optimal. The decision-maker observes demand over time but is otherwise blind to the demand function which maps prices. Structural non-parametric assumptions are made for demand function and demand rate at any instant is estimated by multivariate Poisson process and intensity is determined by price vectors which denotes the set of feasible prices. The authors use non-parametric price algorithm which experiments with several feasible prices and selects price that is expected to optimize revenue based on the demand observed in the previous step. The advantage is the use of multiple products which are linked with various resource constraints contrary to previous approaches which were restricted to single product problem. Rusmevichientong, Roy, and Glynn (2006) formulate a multiproduct single resource pricing problem as a static optimization problem in which a car manufacturer seeks fixed prices that maximizes revenues based on historical preference data. Lim and Shanthikumar (2007) use a max–min approach to a single-product revenue management problem where the uncertainty arises at the level of the point process distribution characterizing the customers’ requests. Other min–max approaches to modeling uncertainty include minimax regret (see Eren, Maglaras, and Ryzin (2006) and Ball and Queyranne (2006) for references in the revenue management literature). Kiefer and Wolfowitz (1952) estimate the point of maximum when the function is unknown and stochastic approximation algorithm proposed uses simulation-based optimization. The proposed Kiefer–Wolfowitz (KW) algorithm uses sequential procedure for stochastic approximation depending on four conditions. The algorithm does not require existence of derivative but optimality of tuning sequence in some reasonable sense is required for the convergence of the algorithm. Though sequence converges stochastically but the proposed algorithm says nothing about the optimal values of the sequence parameters and the stopping rule. Broadie, Deniz, and Assaf (2009) derive general bound on the mean squared error and most importantly the optimal choice of tuning sequences. They present an adaptive version of the KW algorithm, i.e. scale and shifted KW (SSKW) which improved finite time behaviour dramatically. The key idea is to dynamically shift and scale tuning sequences to get better results for the unknown function and noise level. Although problem of oscillation and noisy gradient never affects the asymptotic convergence, SSKW algorithm greatly improves the rate of convergence for finite time than truncated KW algorithm. Xu and Yu-Hong (2008) give a stochastic approximation frame algorithm which uses the noisy evaluation of the negative gradient direction as the iterative direction....
[...]
214 citations
"Non-parametric optimal service pric..." refers background in this paper
...Geraghty and Johnson (1997) mention National car rentals increased revenue by $56 million in the first year of use....
[...]
205 citations
"Non-parametric optimal service pric..." refers background in this paper
...Farias and Roy (2009) analyse dynamic pricing with a prior on market response with maximizing expected discounted revenue over infinite horizon of time....
[...]