Personalized Pricing              Recommender System         Multi-Stage Epsilon-Greedy Approach               Toshihiro K...
Personalized Pricing                  Recommender System             Multi-Stage Epsilon-Greedy Approach                  ...
IntroductionSeventeen years has passed after the birth of Grouplens...  But, recommender systems still have many limitatio...
Introduction    Seventeen years has passed after the birth of Grouplens...      But, recommender systems still have many l...
Introduction    Seventeen years has passed after the birth of Grouplens...      But, recommender systems still have many l...
Introduction    Seventeen years has passed after the birth of Grouplens...      But, recommender systems still have many l...
OutlineIntroductionPrice Personalization and Its Merits  price personalization, resale, commercial viability of RS, merits...
Price Personalization and Its Merits                                       4
Price Discrimination &              Price Personalization                      Price DiscriminationA pricing scheme where ...
Resale        Resale: obstacle to implement price discrimination                 Customers buy items at low prices        ...
Commercial Viability of RS Commercial viability is important for reliable recommendation                        Cost for m...
Commercial Viability of RS Commercial viability is important for reliable recommendation                        Cost for m...
Merits of Price PersonalizationCommercial viability is improved by introducing PP     Recommendation becomes more reliable...
Merits of Price Personalization       Commercial viability is improved by introducing PP            Recommendation becomes...
Formalization of PPRS                        9
Setting of PPRS     Personalized Pricing Recommender System (PPRS)Recommender system having the functionality of price per...
Objective of PPRS    Objective of a Personalized Pricing Recommender Systemmaximize the cumulative rewards by Iterating th...
Objective of PPRS    Objective of a Personalized Pricing Recommender Systemmaximize the cumulative rewards by Iterating th...
Customer Type                    There are three customer typesStandard: Customers who will buy an item regardless of whet...
Customers’ Actions and RewardsPredicted customer type and rewards brought by customer’s action             profit gained by...
Implementation of PPRS                         14
Implementation of PPRS     The I / O of the prediction model for the customer type                  A customer-item pair t...
Three Technical Problems   Three technical problems in the prediction of customer typesAmbiguity in Observation  System ca...
Ambiguity in Observation           The impossibility to detect true customer type                by observing customers’ r...
Multistage Classification        To solve the problem of the ambiguity in classification,            we take a multi-stage c...
Exploitation-Exploration Trade-OffA system has to collect training data by offering non-optimal prices   Too frequent non-...
Multi-Armed Bandit: ε-Greedy         A Multi-armed bandit problem treats the adjustment of exploitation-exploration trade-...
Class Imbalance ProblemThe decline in accuracy when the class distribution is highly skewed         A class wighting appro...
Experiments              22
Experimental ConditionQuasi-synthetic data from MovieLens’ 1M dataset   Preference data and customers’ demographics are im...
Main Experimental ResultsOur PPRS could successfully obtain additional rewards byadopting a personalized pricing schemeWe ...
ConclusionContributions  We added the function to take actions other than recommendation,  i.e., price personalization, to...
Upcoming SlideShare
Loading in …5
×

Personalized Pricing Recommender System — Multi-Stage Epsilon-Greedy Approach

1,623 views

Published on

Personalized Pricing Recommender System — Multi-Stage Epsilon-Greedy Approach
2nd Intternational Workshop on Information Heterogeneity and Fusion in Recommender Systems (HetRec 2011), in conjunction with RecSys 2011

Article @ Official Site: http://dx.doi.org/10.1145/2039320.2039329
Article @ Personal Site: http://www.kamishima.net/archive/2011-ws-recsys_hetrec.pdf
Handnote : http://www.kamishima.net/archive/2011-ws-recsys_hetrec-HN.pdf
Workshop Homepage: http://ir.ii.uam.es/hetrec2011/

Abstract:
Many e-commerce sites use recommender systems, which suggest items that customers prefer. Though recommender systems have achieved great success, their potential is not yet fulfilled. One weakness of current systems is that the actions of the system toward customers are restricted to simply showing items. We propose a system that relaxes this restriction to offer price discounting as well as recommendations. The system can determine whether or not to offer price discounting for individual customers, and such a pricing scheme is called price personalization. We discuss how the introduction of price personalization improves the commercial viability of managing a recommender system, and thereby improving the customers’ sense of the system’s reliability. We then propose a method for adding price personalization to standard recommendation algorithms which utilize two types of customer data: preferential data and purchasing history. Based on the analysis of the experimental results, we reveal further issues in designing a personalized pricing recommender system.

0 Comments
4 Likes
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total views
1,623
On SlideShare
0
From Embeds
0
Number of Embeds
9
Actions
Shares
0
Downloads
29
Comments
0
Likes
4
Embeds 0
No embeds

No notes for slide

Personalized Pricing Recommender System — Multi-Stage Epsilon-Greedy Approach

  1. 1. Personalized Pricing Recommender System Multi-Stage Epsilon-Greedy Approach Toshihiro Kamishima and Shotaro AkahoNational Institute of Advanced Industrial Science and Technology (AIST), Japan2nd Intl Ws on Information Heterogeneity and Fusion in Recommender Systems In conjunction with RecSys 2011 @ Chicago, U.S.A., Oct. 27, 2011 1
  2. 2. Personalized Pricing Recommender System Multi-Stage Epsilon-Greedy Approach Toshihiro Kamishima and Shotaro Akaho National Institute of Advanced Industrial Science and Technology (AIST), Japan 2nd Intl Ws on Information Heterogeneity and Fusion in Recommender Systems In conjunction with RecSys 2011 @ Chicago, U.S.A., Oct. 27, 2011START 1
  3. 3. IntroductionSeventeen years has passed after the birth of Grouplens... But, recommender systems still have many limitations 2
  4. 4. Introduction Seventeen years has passed after the birth of Grouplens... But, recommender systems still have many limitationsOne of such limitations is that a RS is a system only to recommend and cannot behave like clerks in real store 2
  5. 5. Introduction Seventeen years has passed after the birth of Grouplens... But, recommender systems still have many limitationsOne of such limitations is that a RS is a system only to recommend and cannot behave like clerks in real storeA RS that can take an action other than a simple recommendation 2
  6. 6. Introduction Seventeen years has passed after the birth of Grouplens... But, recommender systems still have many limitationsOne of such limitations is that a RS is a system only to recommend and cannot behave like clerks in real storeA RS that can take an action other than a simple recommendation As such an action, we chose Price PersonalizationA pricing scheme that allows sellers to adjust the price for an item depending on the customer or transaction 2
  7. 7. OutlineIntroductionPrice Personalization and Its Merits price personalization, resale, commercial viability of RS, meritsFormalization of a PPRS setting, objective, customer typeImplementation of a PPRS I/O, Ambiguity in observation, exploitation-exploration trade-off, class imbalance problemExperimentsConclusion 3
  8. 8. Price Personalization and Its Merits 4
  9. 9. Price Discrimination & Price Personalization Price DiscriminationA pricing scheme where different prices are charged for the same item hamburger chain stores region A region B $1.20 $1.00 Price Personalization (Price Customization / Dynamic Pricing) A pricing scheme that allows sellers to adjust the price for an item depending on the customer or transaction Personalized coupons in real retail stores Air tickets are sold at personalized price based on the past behavior Sellers can obtain the additional profit by offering a discount only to customers who will not buy at a standard price but will buy at a discounted price 5
  10. 10. Resale Resale: obstacle to implement price discrimination Customers buy items at low prices and then resell them at higher prices Resale activity must be blockedTraditional price discrimination Sales regions are changed for the items that are difficult to transportPrice personalization Targeting e-commerce where the sales volumes of individual customers are precisely controlled Dealing with personalized items, such as registered air tickets or subscription services Our approach: predict whether customers will resale or not 6
  11. 11. Commercial Viability of RS Commercial viability is important for reliable recommendation Cost for managing recommender systems Profit obtained by the increase of customer customer loyalty seller Because the effect of loyalty on the profit is indirect and uncertain,the additional profit might be inadequate to compensate for the cost 7
  12. 12. Commercial Viability of RS Commercial viability is important for reliable recommendation Cost for managing recommender systems Profit obtained by the increase of customer customer loyalty seller Because the effect of loyalty on the profit is indirect and uncertain,the additional profit might be inadequate to compensate for the costA dark recommender system may recommend more expensive itemsinstead of offering lower-cost items that will satisfy customers’ needs 7
  13. 13. Merits of Price PersonalizationCommercial viability is improved by introducing PP Recommendation becomes more reliableWhat can we do for such a dark recommender system? 8
  14. 14. Merits of Price Personalization Commercial viability is improved by introducing PP Recommendation becomes more reliable What can we do for such a dark recommender system? Use the personalization Additional profit brought by introducing price personalization enhances the commercial viability of RS Decreasing the sellers’ incentive of making dark recommendationCustomers have the additional benefit of being offered price discounts 8
  15. 15. Formalization of PPRS 9
  16. 16. Setting of PPRS Personalized Pricing Recommender System (PPRS)Recommender system having the functionality of price personalization The simplest PPRS A PPRS is passively invoked for an item that a customer is currently viewing or accessing There are only two levels of prices: a standard and a discounted A system offers discounts when the customer is expected to buy the item only if a discounted price is offered For each target item, a specific customer can be offered a discounted price only when the customer first views the item (This rule blocks the repetition of revisiting until discount prices are offered) 10
  17. 17. Objective of PPRS Objective of a Personalized Pricing Recommender Systemmaximize the cumulative rewards by Iterating the process below 1) select an item 2) predict a customer type for the pair of the customer and the item 3) determine whether to offer a discounted or a standard pricecustomer based on the predicted customer type PPRS 4) decide whether to buy the item 5) receives a reward based on the customer’s decision and the customer type 11
  18. 18. Objective of PPRS Objective of a Personalized Pricing Recommender Systemmaximize the cumulative rewards by Iterating the process below 1) select an item 2) predict a customer type for the pair of the customer and the item 3) determine whether to offer a discounted or a standard pricecustomer based on the predicted customer type PPRS 4) decide whether to buy the item 5) receives a reward based on the customer’s decision and the customer type 11
  19. 19. Customer Type There are three customer typesStandard: Customers who will buy an item regardless of whether theprice is standard or discounted A standard price should be offered to obtain more profitDiscount: Price-sensitive customers who will buy an item only if adiscounted price is offered. A discounted price should be offered so that a customer to buy an itemIndifferent: Customers who will not buy an item whether or not it isdiscounted A standard price should be offered to block the customers to resale, because these customers will not consume the item for oneself 12
  20. 20. Customers’ Actions and RewardsPredicted customer type and rewards brought by customer’s action profit gained by selling itemsCType Standard Discount Indifferent offer standard price discount price standard priceBuy α β 0NotBuy 0 0 γ α>β≫γ>0 potential profit by blocking resale 13
  21. 21. Implementation of PPRS 14
  22. 22. Implementation of PPRS The I / O of the prediction model for the customer type A customer-item pair to predict its customer typeINPUTS customer & item ID preference to features of log of customers’the items in DB customers purchases preference DB customer data purchasing historyRecommendationmodel features target values model parameter classification model OUTPUTSpreference to item customer type 15
  23. 23. Three Technical Problems Three technical problems in the prediction of customer typesAmbiguity in Observation System cannot detect true customer type only by observing customers’ behaviorExploitation–Exploration Trade-Off System must offer non-best prices occasionally to collect purchase dataClass Imbalance Problem The decline in accuracy when the class distribution is highly skewed 16
  24. 24. Ambiguity in Observation The impossibility to detect true customer type by observing customers’ responses true customer type standard Buy cannot differentiateoffer discount discount Buy indifferent Not Buy unknown to a PPRS must be guessed from customers’ responses 17
  25. 25. Multistage Classification To solve the problem of the ambiguity in classification, we take a multi-stage classification approach learned from the customers’ customer-item pair responses to offers at a stadard price standard classifierlearned from the customers’ responses to offers non-standard type standard type at a discounted price discount classifier indifferent type discount type In our experiment, a prescreening stage is added 18
  26. 26. Exploitation-Exploration Trade-OffA system has to collect training data by offering non-optimal prices Too frequent non-optimal actions may reduce the total rewardA PPRS collects purchasing histories while predicting customer type Exploitation Trade-Off Exploration take the best action take non-best actions to earn rewards to collect data too frequent non-best current prediction of actions will reduce customer types the total rewards might be incorrect 19
  27. 27. Multi-Armed Bandit: ε-Greedy A Multi-armed bandit problem treats the adjustment of exploitation-exploration trade-offs ε-Greedy: the most naive approach Exploitation Exploration Pr = 1 - ε Pr = ε standard classifier prediction = a standard typeExploitation Exploration standard type non-standard typeoffer at a standard price pass to a discount classifier A parameter ε must be tuned by hand 20
  28. 28. Class Imbalance ProblemThe decline in accuracy when the class distribution is highly skewed A class wighting approach alleviates this problem In a case of a standard classifier,many standard customers wrongly classified as non-standard ones major class minor class non-standard type standard type smaller larger High Recall High Precision decision threshold for the minor class probability 21
  29. 29. Experiments 22
  30. 30. Experimental ConditionQuasi-synthetic data from MovieLens’ 1M dataset Preference data and customers’ demographics are imported from Movielens dataset Customers’ purchasing histories are artificially generated so that satisfy the following conditions: 1.Preference for the target items would become stronger in the order of a standard customer, a discount customer, and an indifferent customer 2.The determination of purchasing activities was assumed to depend on the customers’ preference for the target items and their demographics 3.Almost all customers are indifferent, and the number of discount customers is slightly larger than that of standard customers Though this purchasing history is simple, it is not trivial for a system to be able to obtain additional reward because of three technical problems Please refer to our manuscripts about the details of conditions 23
  31. 31. Main Experimental ResultsOur PPRS could successfully obtain additional rewards byadopting a personalized pricing schemeWe could adjust parameters by observing customers’ behaviors,even though true customer types could not be observedHigher-weighting on standard customers than non-standards in astandard classifier, because it is important not to miss loyalcustomersLower-weighing on discount customers than indifferent ones in adiscount classifier, because discounts should be offered forcustomers who are certainly discount typesExploration probability ε heavily affects the total rewards 24
  32. 32. ConclusionContributions We added the function to take actions other than recommendation, i.e., price personalization, to a RS We discussed how it improves the commercial viability of managing a RS, and thereby improving the reliability to a RS We implemented a simple system and tested on a quasi-synthetic data setFuture Work If utilities other than prices can be considered as rewards, a framework of a PPRS could be made applicable to broader actions Recommender systems have started to provide not only simple recommendations but also more sophisticated actions; such evolved systems could be called Attendant Systems 25

×