WO2014173237A1 - 一种推荐的方法及服务器 - Google Patents
一种推荐的方法及服务器 Download PDFInfo
- Publication number
- WO2014173237A1 WO2014173237A1 PCT/CN2014/075183 CN2014075183W WO2014173237A1 WO 2014173237 A1 WO2014173237 A1 WO 2014173237A1 CN 2014075183 W CN2014075183 W CN 2014075183W WO 2014173237 A1 WO2014173237 A1 WO 2014173237A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- recommendation
- user
- combined
- recommendation list
- selection result
- Prior art date
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/45—Management operations performed by the client for facilitating the reception of or the interaction with the content or administrating data related to the end-user or to the client device itself, e.g. learning user preferences for recommending movies, resolving scheduling conflicts
- H04N21/466—Learning process for intelligent management, e.g. learning user preferences for recommending movies
- H04N21/4668—Learning process for intelligent management, e.g. learning user preferences for recommending movies for recommending content, e.g. movies
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
Definitions
- the invention belongs to the field of data processing, and in particular relates to a recommended method and server.
- the business system to the recommendation system generally imports data offline, and the user feedback is the user's real-time data feedback to the recommendation system for updating the recommendation model to improve future prediction accuracy.
- Commonly used user feedback methods such as collecting, clicking, browsing (time), purchasing, scoring, commenting, etc.
- a disadvantage of the prior art 1 is that the technology is relatively simple, and it is difficult to balance both the accuracy of the recommendation and the real-time of the calculation.
- the recommended combination technique obtains the selection model through offline training, and is essentially a single recommendation system.
- the recommended combination and the back-end recommendation technology are generally strongly related and deployed together.
- Recommended combination techniques neural networks, case-based reasoning (CBR), decision trees, etc.
- CBR case-based reasoning
- a disadvantage of the prior art 2 is that the recommendation technique is not easy to expand. For each additional recommendation technique, the combined model needs to be re-offline training, and the current interest of the user cannot be fed back in real time.
- SUMMARY OF THE INVENTION It is an object of the present invention to provide a recommended method and server for solving multiple recommendation models Or the system, how to use the real-time evaluation effect to update the recommendation list, and reflect the user's current interests and historical interests.
- a preferred method comprising:
- the recommendation list is combined according to a preset combination policy, and the combined recommendation list is presented to the user, so that the user selects according to the combined recommendation list;
- the combining, according to the preset combination policy, the combining the recommended list includes:
- the receiving the feedback result of the user feedback, updating the combination according to the selecting result A list of recommendations including:
- the update coefficient is preset, and the proportion of each recommended system after updating is calculated according to the selection probability, the update coefficient, and the proportion of each recommendation system;
- the combined recommendation list is updated according to the updated weight.
- the receiving the feedback result of the user feedback, and calculating the selection probability according to the selection result including:
- the weight selected by the user is set in advance, and the weight of the result selected by the user for each recommendation system is obtained according to the weight selected by the user, and the ratio of the weight of all the results of the combined recommendation list is obtained, and the ratio is the selection probability.
- the method further includes: The selection result of the user's regular feedback is received, and the recommendation list of each recommendation system is updated according to the selection result of the periodic feedback.
- a server the server includes:
- a receiving unit configured to receive a recommendation list sent by each recommendation system
- a combination unit configured to combine the recommended list according to a preset combination policy, and present the combined recommendation list to the user, so that the user selects according to the combined recommendation list; and the update unit is configured to receive the user The result of the selection of the feedback updates the combined recommendation list according to the selection result.
- the combining unit is specifically configured to:
- the updating unit is specifically configured to:
- the update coefficient is preset, and the proportion of each recommended system after updating is calculated according to the selection probability, the update coefficient, and the proportion of each recommendation system;
- the combined recommendation list is updated according to the updated weight.
- the performing step in the updating unit receives a selection result that is fed back by the user, and calculates a selection probability according to the selection result, Includes:
- the weight selected by the user is set in advance, and the weight of the result selected by the user for each recommendation system is obtained according to the weight selected by the user, and the ratio of the weight of all the results of the combined recommendation list is obtained, and the ratio is the selection probability.
- the server further includes a periodic feedback unit, configured to: The selection result of the user's regular feedback is received, and the recommendation list of each recommendation system is updated according to the selection result of the periodic feedback.
- the present invention combines recommendation lists sent by each recommendation system by using a preset combination strategy, implements multiple recommendation models or systems to feed back recommendation results to the user, and updates the description according to the selection probability of the user feedback.
- the combined recommendation list realizes real-time evaluation, and receives the selection result of the user's regular feedback, and updates the combined recommendation list, because the real-time evaluation can reflect the current interest of the user, and the regular feedback can reflect the historical interest of the user, so the present invention It can simultaneously reflect the user's current interests and historical interests.
- FIG. 1 is an application scenario diagram of a recommended method according to an embodiment of the present invention
- FIG. 2 is a flowchart of a method for recommending a method according to an embodiment of the present invention
- FIG. 3 is a schematic diagram of a method for recommending a method according to an embodiment of the present invention.
- FIG. 4 is a schematic diagram of a method for recommending a method according to an embodiment of the present invention.
- FIG. 5 is a structural diagram of a device according to an embodiment of the present invention.
- FIG. 6 is a structural diagram of a device of a server according to an embodiment of the present invention.
- DETAILED DESCRIPTION OF THE EMBODIMENTS In order to make the objects, technical solutions and advantages of the present invention more comprehensible, the present invention will be further described in detail below with reference to the accompanying drawings. It is understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
- FIG. 1 is an application scenario diagram of a recommended method according to an embodiment of the present invention.
- the user 101 selects an interest from the recommendation list according to the recommendation list provided by the server 102.
- the user 101 feeds the selected result to the server 102 in real time, and the server 102 updates the system according to the selection result of the user's real-time feedback.
- the server 102 pushes the recommended list pushed by the user 101, the user 101 can be reflected in time.
- the server 102 periodically receives the selection result fed back by the user 101, and the server 102 updates the system according to the selection result of the periodic feedback, so that the server 102 can simultaneously reflect the user's historical interest and current interest in the recommendation list pushed to the user 101 each time.
- FIG. 2 is a flowchart of a method for recommending a method according to an embodiment of the present invention. As shown in FIG. 2, the method includes the following steps:
- Step 201 Receive a recommendation list sent by each recommendation system.
- FIG. 3 is a schematic diagram of a method for recommending a method according to an embodiment of the present invention.
- the recommended front-end system receives the recommendation list sent by the recommendation system 1, the recommendation system 2, and the recommendation system 3, and the recommended front-end system sends the recommendation system 1, the recommendation system 2, and the recommendation system 3 according to the combination policy.
- the recommendation list is combined, and the recommended front-end system sends the combined recommendation list to the service system, so that the service system presents the combined recommendation list to the user.
- the recommended front-end system receives the selection probability of the user's real-time feedback, and is used to update the combined recommendation list.
- the recommendation system 1, the recommendation system 2, and the recommendation system 3 receive the selection result of the user's periodic feedback, and is used to update the recommendation system.
- Recommended system 2 recommendation system 3 recommendation list, each recommendation system is different according to the database calculation recommendation results, for example, recommendation system 1 is more recommended for children's products, recommendation system 2 more recommended electronic products or books or clothes And other fields.
- Step 202 Combine the recommendation list according to a preset combination policy, and present the combined recommendation list to the user, so that the user selects according to the combined recommendation list;
- the combining, according to the preset combination policy, the recommendation list is combined, including: predefining a ratio of results sent by each recommendation system to results sent by all recommendation systems;
- the recommendation list A sent by the recommendation system 1 to the recommended front end system is ⁇ al, a2, a3 ⁇
- the recommendation list B sent by the recommendation system 2 to the recommended front end system is ⁇ bl, b2, b3 ⁇
- the recommendation system 3 sends
- the recommended list C for the recommended front-end system is ⁇ cl, c2, c3 ⁇ .
- the combined recommendation list may be ⁇ al, bl, c2 ⁇ , hypothetical push Recommended system 1, recommendation system 2, recommendation system 3 recommended results accounted for 3/5, 1/5, 1/5 of the recommended results of all recommended systems, then the combined recommendation list can be ⁇ al, a2, A3, b2, c3 ⁇ .
- the proportion of the combination strategy can be freely defined. 4
- the proportion of the recommendation system 1 can be increased, because most of the recommended systems 1 are recommended for children's products.
- Step 203 Receive a selection result fed back by the user, and update the combined recommendation list according to the selection result.
- the receiving the result of the selection of the user feedback, and updating the combined recommendation list according to the selection result includes:
- the update coefficient is preset, and the proportion of each recommended system after updating is calculated according to the selection probability, the update coefficient, and the proportion of each recommendation system;
- the combined recommendation list is updated according to the updated weight.
- the receiving the result of the selection of the user feedback, and calculating the selection probability according to the selection result including:
- the weight selected by the user is set in advance, and the weight of the result selected by the user for each recommendation system is obtained according to the weight selected by the user, and the ratio of the weight of all the results of the combined recommendation list is obtained, and the ratio is the selection probability.
- each recommendation system in FIG. 3 is initialized, and the proportions of each recommendation system are respectively corpse 2 ( ), P 3 ⁇ t),
- P i (t + 1) represents the proportion of the first recommendation system at the next moment after the moment
- ⁇ (0 represents the probability that the result recommended by the first recommendation system is selected by the user at the moment)
- the recommendation list ⁇ sent by the recommendation system 1 is ⁇ il, i2, i3 ⁇
- the recommendation list B sent by the recommendation system 2 is ⁇ i2, i3, i4, i5 ⁇
- the recommendation system 3 does not send the recommendation list, and the final combination is presented to the user.
- the list of recommended lists is ⁇ il, i2, i3, i4, i5 ⁇ .
- the user selects ⁇ i2,i4 ⁇ , in fact, the user chooses ⁇ i2(A), i2(B), i4(B)
- the selection probability that each recommendation subsystem recommendation result is selected by the user needs to consider the weight of each user's operation behavior, and the obvious purchase behavior weight is more important than the browsing behavior, assuming that the purchase behavior weight is 0.3 and the browsing behavior weight is 0.2.
- the user selection probability is ⁇ i2(A)*0.3, i2(B)*0.3, i4(B)*0.21, so
- recommendation system 1 is better than the first case in the second case, because the recommendation system 1 recommended i2 was purchased by the user.
- the method further includes:
- the selection result of the user's regular feedback is received, and the recommendation list of each recommendation system is updated according to the selection result of the periodic feedback.
- the recommendation system 1.
- the recommendation system 3 periodically receives the selection result of the user's regular feedback, and updates the recommendation list sent by the user according to the result.
- the invention combines the recommendation lists sent by the respective recommendation systems by using a combination strategy set in advance, and implements a plurality of recommendation models or systems to feed back the recommendation results to the user, and updates the combined recommendation list according to the selection probability fed back by the user, thereby realizing The real-time evaluation, and receiving the selection result of the user's regular feedback, updating the combined recommendation list, thereby simultaneously reflecting the user's current interests and historical interests.
- FIG. 4 is a schematic diagram of a method for recommending a method according to an embodiment of the present invention. The method includes the following steps:
- Step 401 If the user logs in for the first time, and the system has no user information, the recommended front-end system selects the recommendation result to the user in a recommended amount in each recommendation subsystem, or the system assigns an initial value to each recommendation subsystem according to the combination policy. ;
- Step 402 The front-end system is recommended to obtain a recommendation list of each recommendation system for combined display; step 403, the user selects browsing or purchasing or collecting the recommended result;
- Step 404 recommending that the front-end system captures the behavior of the user in real time, and gives different users different behaviors.
- the weight represents the current preference of the user, and the selection probability calculation of the recommendation result is performed;
- Step 405 recommending that the front-end system adjusts the proportion of the output of each recommendation system in the recommended final result list according to a user's current preference according to a certain rule;
- Step 406 the user can see a recommendation result different from the previous one at this time;
- Step 407 the user logs out to end the session, and each recommendation subsystem obtains the overall behavior of the user during the session;
- Step 408 Each recommendation subsystem adjusts the recommendation list according to the historical preference of the user
- Step 409 the user logs in to the system again, and the process goes to step 401.
- FIG. 5 is a structural diagram of a device of a server according to an embodiment of the present invention. As shown in Figure 4, the server includes the following units:
- the receiving unit 501 is configured to receive a recommendation list sent by each recommendation system
- FIG. 3 is a schematic diagram of a method for recommending a method according to an embodiment of the present invention.
- the recommended front-end system receives the recommendation list sent by the recommendation system 1, the recommendation system 2, and the recommendation system 3, and the recommended front-end system sends the recommendation system 1, the recommendation system 2, and the recommendation system 3 according to the combination policy.
- the recommendation list is combined, and the recommended front-end system sends the combined recommendation list to the service system, so that the service system presents the combined recommendation list to the user.
- the recommended front-end system receives the selection probability of the user's real-time feedback, and is used to update the combined recommendation list.
- the recommendation system 1, the recommendation system 2, and the recommendation system 3 receive the selection result of the user's periodic feedback, and is used to update the recommendation system. 1.
- the combining unit 502 is configured to combine the recommended list according to a preset combination policy, and present the combined recommendation list to the user, so that the user selects according to the combined recommendation list;
- the combining unit 502 is specifically configured to:
- the recommendation list A sent by the recommendation system 1 to the recommended front end system is ⁇ al, a2, a3 ⁇
- the recommendation list B sent by the recommendation system 2 to the recommended front end system is ⁇ bl, b2, b3 ⁇
- the recommendation system 3 sends
- the recommended list C for the recommended front-end system is ⁇ cl, c2, c3 ⁇ .
- the recommended results of recommendation system 1, recommendation system 2, recommendation system 3 account for all recommended system recommendation results
- the recommended list after the combination can be ⁇ al, bl, c2 ⁇ , assuming that the recommended results of the recommendation system 1, the recommendation system 2, and the recommendation system 3 account for the recommendation results of all the recommended systems, respectively.
- the combined recommendation list can be ⁇ al, a2, a3, b2, c3 ⁇ .
- the proportion of the combination strategy can be freely defined. 4
- the proportion of the recommendation system 1 can be increased, because most of the recommended systems 1 are recommended for children's products.
- the updating unit 503 is configured to receive a selection result of the user feedback, and update the combined recommendation list according to the selection result.
- the updating unit 503 is specifically configured to:
- the update coefficient is preset, and the proportion of each recommended system after updating is calculated according to the selection probability, the update coefficient, and the proportion of each recommendation system;
- the combined recommendation list is updated according to the updated weight.
- the performing step of the updating unit receives the selection result of the user feedback, and calculates the selection probability according to the selection result, including:
- the weight selected by the user is set in advance, and the weight of the result selected by the user for each recommendation system is obtained according to the weight selected by the user, and the ratio of the weight of all the results of the combined recommendation list is obtained, and the ratio is the selection probability.
- each recommendation system is respectively corpse 2 ( ), P 3 ⁇ t) , l + ⁇ ⁇ ⁇
- the recommendation list ⁇ sent by the recommendation system 1 is ⁇ il, i2, i3 ⁇
- the recommendation list B sent by the recommendation system 2 is ⁇ i2, i3, i4, i5 ⁇
- the recommendation system 3 does not send the recommendation list, and the final combination is presented to the user.
- the list of recommended lists is ⁇ il, i2, i3, i4, i5 ⁇ .
- the user's operational behavior is different, for example: the user's behavior on i2 is purchased, and for i4
- the behavior is only when browsing.
- the selection probability of each recommendation subsystem recommendation result selected by the user needs to consider the weight of each user's operation behavior, and the obvious purchase behavior weight is more important than the browsing behavior, assuming that the purchase behavior weight is 0.3 and the browsing behavior weight is 0.2.
- the user selection probability is ⁇ i2(A)*0.3, i2(B)*0.3, i4(B)*0.2 ⁇ , so
- the server further includes a periodic feedback unit 504, configured to: receive a selection result periodically reported by the user, and update a recommendation list of each recommendation system according to the selection result of the periodic feedback.
- a periodic feedback unit 504 configured to: receive a selection result periodically reported by the user, and update a recommendation list of each recommendation system according to the selection result of the periodic feedback.
- the recommendation system 1.
- the recommendation system 3 periodically receives the selection result of the user's regular feedback, and updates the recommendation list sent by the user according to the result.
- the invention combines the recommendation lists sent by the respective recommendation systems by using a combination strategy set in advance, and implements a plurality of recommendation models or systems to feed back the recommendation results to the user, and updates the combined recommendation list according to the selection probability fed back by the user, thereby realizing The real-time evaluation, and receiving the selection result of the user's regular feedback, updating the combined recommendation list, thereby simultaneously reflecting the user's current interests and historical interests.
- FIG. 6 is a structural diagram of a device of a server according to an embodiment of the present invention.
- FIG. 6 is a server 600 according to an embodiment of the present invention.
- the specific embodiment of the present invention does not limit the specific implementation of the server.
- the server 600 includes:
- the processor 601, the communication interface 602, and the memory 603 complete communication with each other via the bus 604.
- the processor 601 is configured to execute a program.
- the program can include program code, the program code including computer operating instructions.
- the processor 601 may be a central processing unit CPU, or an Application Specific Integrated Circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention.
- ASIC Application Specific Integrated Circuit
- the memory 603 is used to store the program.
- the memory 603 may include a high speed RAM memory and may also include a non-volatile memory such as at least one disk memory.
- the program specific can include:
- the recommendation list is combined according to a preset combination policy, and the combined recommendation list is presented to the user, so that the user selects according to the combined recommendation list;
- the combining the recommendation lists according to the preset combination policy includes: predefining a ratio of results sent by each recommendation system to results sent by all recommendation systems;
- Receiving the selection result of the user feedback, and updating the combined recommendation list according to the selection result including:
- the update coefficient is preset, and the proportion of each recommended system after updating is calculated according to the selection probability, the update coefficient, and the proportion of each recommendation system;
- the combined recommendation list is updated according to the updated weight.
- the calculating the selection result according to the selection result, and calculating the selection probability according to the selection result comprising: calculating a ratio of the selection result to all the results of the combined recommendation list, wherein the ratio is a selection probability; or
- the weight selected by the user is set in advance, and the weight of the result selected by the user for each recommendation system is obtained according to the weight selected by the user, and the ratio of the weight of all the results of the combined recommendation list is obtained, and the ratio is the selection probability.
- the method further includes:
- the selection result of the user's regular feedback is received, and the recommendation list of each recommendation system is updated according to the selection result of the periodic feedback.
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
Abstract
本发明实施例公布了一种推荐的方法及服务器,所述方法通过预先设置的组合策略,对各个推荐系统发送的推荐列表进行组合,实现多个推荐模型或者系统给用户反馈推荐结果,根据用户反馈的选择概率,更新所述组合后的推荐列表,实现实时评估,并接收用户定期反馈的选择结果,更新所述组合后的推荐列表,进而同时体现用户的当前兴趣和历史兴趣爱好。
Description
说 明 书 一种推荐的方法 艮务器 技术领域
本发明属于数据处理领域, 尤其涉及一种推荐的方法及服务器。
背景技术 在信息爆炸的今天, 越来越多的商业系统引入推荐技术, 从以前人找内容 的模式转变成内容找人, 满足用户个性化需求。 单一的推荐系统推荐结果的效 果有限, 这点特别是越来越多的推荐竟赛中得到体现, 竟赛最终的获奖者往往 是釆用多个推荐技术 /模型或者评分结果进行融合集成。 以往的推荐更多类比成 预测评分问题, 但有学者认为推荐列表形式可能更合适, 以往的推荐系统效果 不好评估, 一般也是离线进行, 实时的效果评估未被加以利用。 历史喜好代表 的是用户一直以来的兴趣爱好, 在很长一段时间内一般是不会改变的, 可以通 过分析用户的历史行为得到; 而当前喜好代表的是用户当前临时的兴趣爱好, 一般也是随时间和外界环境而易变的。
通常的现有技术一中, 业务系统到推荐系统一般通过离线导入数据, 用户 反馈是用户实时的数据反馈到推荐系统, 用于更新推荐模型以提高将来的预测 准确性。 常用的用户反馈方式, 有收藏、 点击、 浏览 (时间) 、 购买、 打分、 评论等行为。 现有技术一的缺点在于, 技术比较单一, 在推荐的准确性和计算 的实时性方面难以兼顾。
通常的现有技术二中, 推荐组合技术通过离线训练得到选择模型, 本质上 还是单一的推荐系统。 推荐组合与后端推荐技术一般都是强相关, 一起部署的。 推荐组合技术: 神经网络、 案例式推理(Case-based reasoning, CBR ) 、 决策树 等。 现有技术二的缺点在于, 推荐技术不易扩展, 每增加一种推荐技术则组合 模型需要重新离线训练, 并且无法实时反馈用户的当前兴趣。 发明内容 本发明的目的在于提供一种推荐的方法及服务器, 解决存在多个推荐模型
或系统时, 如何利用实时的评估效果对推荐列表进行更新, 同时体现用户的当 前兴趣和历史兴趣爱好。
第一方面, 一种推荐的方法, 所述方法包括:
接收各个推荐系统发送的推荐列表;
根据预先设置的组合策略, 将所述推荐列表进行组合, 将组合后的推荐列 表呈现给用户, 使得用户根据所述组合后的推荐列表进行选择;
接收用户反馈的选择结果, 根据所述选择结果更新所述组合后的推荐列表。 结合第一方面, 在第一方面的第一种可能的实现方式中, 所述根据预先设 置的组合策略, 将所述推荐列表进行组合, 包括:
预先定义各个推荐系统发送的结果在所有推荐系统发送的结果中占的比 重;
根据所述比重, 对各个推荐系统发送的推荐列表进行组合。
结合第一方面或者第一方面的第一种可能的实现方式, 在第一方面的第二 种可能的实现方式中, 所述接收用户反馈的选择结果, 根据所述选择结果更新 所述组合后的推荐列表, 包括:
接收用户反馈的选择结果, 根据所述选择结果计算选择概率;
预先设置更新系数, 根据所述选择概率、 更新系数和各个推荐系统的比重 计算各个推荐系统更新后的比重;
根据所述更新后的比重更新组合后的推荐列表。
结合第一方面的第二种可能的实现方式, 在第一方面的第三种可能的实现 方式中, 所述接收用户反馈的选择结果, 根据所述选择结果计算选择概率, 包 括:
计算所述选择结果占所述组合后的推荐列表所有的结果的比例, 所述比例 为选择概率; 或者,
预先设置用户选择的权重, 根据所述用户选择的权重得到用户选择各个推 荐系统推荐的结果的权重占组合后的推荐列表所有的结果的权重的比例, 所述 比例为选择概率。
结合第一方面或者第一方面的第一种可能的实现方式或者第一方面的第二 种可能的实现方式或者第一方面的第三种可能的实现方式, 在第一方面的第四 种可能的实现方式中, 所述方法还包括:
接收用户定期反馈的选择结果, 根据所述定期反馈的选择结果, 更新各个 推荐系统的推荐列表。
第二方面, 一种服务器, 所述服务器包括:
接收单元, 用于接收各个推荐系统发送的推荐列表;
组合单元, 用于根据预先设置的组合策略, 将所述推荐列表进行组合, 将 组合后的推荐列表呈现给用户, 使得用户根据所述组合后的推荐列表进行选择; 更新单元, 用于接收用户反馈的选择结果, 根据所述选择结果更新所述组 合后的推荐列表。
结合第二方面, 在第二方面的第一种可能的实现方式中, 所述组合单元具 体用于:
预先定义各个推荐系统发送的结果在所有推荐系统发送的结果中占的比 重;
根据所述比重, 对各个推荐系统发送的推荐列表进行组合。
结合第二方面或者第二方面的第一种可能的实现方式, 在第二方面的第二 种可能的实现方式中, 所述更新单元具体用于:
接收用户反馈的选择结果, 根据所述选择结果计算选择概率;
预先设置更新系数, 根据所述选择概率、 更新系数和各个推荐系统的比重 计算各个推荐系统更新后的比重;
根据所述更新后的比重更新组合后的推荐列表。
结合第二方面的第二种可能的实现方式, 在第二方面的第三种可能的实现 方式中, 所述更新单元中执行步骤接收用户反馈的选择结果, 根据所述选择结 果计算选择概率, 包括:
计算所述选择结果占所述组合后的推荐列表所有的结果的比例, 所述比例 为选择概率; 或者,
预先设置用户选择的权重, 根据所述用户选择的权重得到用户选择各个推 荐系统推荐的结果的权重占组合后的推荐列表所有的结果的权重的比例, 所述 比例为选择概率。
结合第二方面或者第二方面的第一种可能的实现方式或者第二方面的第二 种可能的实现方式或者第二方面的第三种可能的实现方式, 在第二方面的第四 种可能的实现方式中, 所述服务器还包括定期反馈单元, 用于:
接收用户定期反馈的选择结果, 根据所述定期反馈的选择结果, 更新各个 推荐系统的推荐列表。
与现有技术相比, 本发明通过预先设置的组合策略, 对各个推荐系统发送 的推荐列表进行组合, 实现多个推荐模型或者系统给用户反馈推荐结果, 根据 用户反馈的选择概率, 更新所述组合后的推荐列表, 实现实时评估, 并接收用 户定期反馈的选择结果, 更新所述组合后的推荐列表, 因为实时评估可以体现 用户的当前兴趣, 定期反馈可以体现用户的历史兴趣, 因此本发明可以同时体 现用户的当前兴趣和历史兴趣爱好。 附图说明 为了更清楚地说明本发明实施例中的技术方案, 下面将对实施例中所需要 使用的附图作简单地介绍, 显而易见地, 下面描述中的附图仅仅是本发明的一 些实施例, 对于本领域普通技术人员来讲, 在不付出创造性劳动性的前提下, 还可以根据这些附图获得其他的附图。
图 1是本发明实施例提供的一种推荐的方法的应用场景图;
图 2是本发明实施例提供的一种推荐的方法的方法流程图;
图 3是本发明实施例提供的一种推荐的方法的方法示意图;
图 4是本发明实施例提供的一种推荐的方法的方法示意图;
图 5是本发明实施例提供的一种服务器的装置结构图;
图 6是本发明实施例提供的一种服务器的装置结构图。 具体实施方式 为了使本发明的目的、 技术方案及优点更加清楚明白, 以下结合附图及实 施例, 对本发明进行进一步详细说明。 应当理解, 此处所描述的具体实施例仅 仅用以解释本发明, 并不用于限定本发明。
以上所述仅为本发明的较佳实施例而已, 并不用以限制本发明, 凡在本发 明的精神和原则之内所作的任何修改、 等同替换和改进等, 均应包含在本发明 的保护范围之内。
参考图 1, 图 1是本发明实施例提供的一种推荐的方法的应用场景图。如图 1所示, 用户 101根据服务器 102提供的推荐列表,从推荐列表中选择感兴趣的
物品等, 同时, 用户 101将选择的结果实时反馈给服务器 102, 服务器 102根据 用户实时反馈的选择结果更新系统, 服务器 102下一次给用户 101推送的推荐 列表中能及时反应出用户 101上一次的喜好, 并且服务器 102会定期接收用户 101反馈的选择结果,服务器 102根据定期反馈的选择结果更新系统, 使得服务 器 102每次给用户 101推送的推荐列表中能同时反应用户的历史兴趣和当前兴 趣。
参考图 2, 图 2是本发明实施例提供的一种推荐的方法的方法流程图。如图 2所示, 所述方法包括以下步骤:
步骤 201, 接收各个推荐系统发送的推荐列表;
具体的, 如图 3所示, 图 3是本发明实施例提供的一种推荐的方法的方法 示意图。 如图 3所示, 推荐前端系统接收推荐系统 1、 推荐系统 2、 推荐系统 3 发送的推荐列表, 所述推荐前端系统根据组合策略将所述推荐系统 1、推荐系统 2、 推荐系统 3发送的推荐列表进行组合, 所述推荐前端系统将组合后的推荐列 表发送给业务系统, 使得所述业务系统将组合后的推荐列表呈现给用户。 所述 推荐前端系统接收用户实时反馈的选择概率, 用于更新所述组合后的推荐列表, 同时, 推荐系统 1、 推荐系统 2、 推荐系统 3接收用户定期反馈的选择结果, 用 于更新推荐系统 1、 推荐系统 2、 推荐系统 3的推荐列表, 每个推荐系统根据数 据库计算的推荐结果不同, 例如推荐系统 1更多推荐的是儿童用品,推荐系统 2 更多推荐的电子产品或者书籍或者衣服等领域。
步骤 202, 根据预先设置的组合策略, 将所述推荐列表进行组合, 将组合后 的推荐列表呈现给用户, 使得用户根据所述组合后的推荐列表进行选择;
可选地, 所述根据预先设置的组合策略, 将所述推荐列表进行组合, 包括: 预先定义各个推荐系统发送的结果在所有推荐系统发送的结果中占的比 重;
根据所述比重, 对各个推荐系统发送的推荐列表进行组合。
具体的,假设推荐系统 1发送给推荐前端系统的推荐列表 A为 {al, a2, a3 }, 推荐系统 2发送给推荐前端系统的推荐列表 B为 {bl, b2, b3 }, 推荐系统 3发 送给推荐前端系统的推荐列表 C 为 {cl, c2, c3 }。 根据预先设置的推荐结果的 组合策略, 4叚设推荐系统 1、 推荐系统 2、 推荐系统 3推荐的结果占所有推荐系 统推荐结果的比重都是 1/3时, 则组合后的推荐列表可为 {al, bl, c2}, 假设推
荐系统 1、 推荐系统 2、 推荐系统 3推荐的结果占所有推荐系统推荐结果的比重 分别是 3/5、 1/5、 1/5时, 则组合后的推荐列表可为 { al, a2, a3, b2, c3 }。
同时, 组合策略的比重可以自由定义, 4叚设儿童节时, 可以将推荐系统 1 的比重提高, 因为推荐系统 1推荐的大多是儿童用品。
步骤 203,接收用户反馈的选择结果,根据所述选择结果更新所述组合后的 推荐列表。
可选地, 所述接收用户反馈的选择结果, 根据所述选择结果更新所述组合 后的推荐列表, 包括:
接收用户反馈的选择结果, 根据所述选择结果计算选择概率;
预先设置更新系数, 根据所述选择概率、 更新系数和各个推荐系统的比重 计算各个推荐系统更新后的比重;
根据所述更新后的比重更新组合后的推荐列表。
可选地, 所述接收用户反馈的选择结果, 根据所述选择结果计算选择概率, 包括:
计算所述选择结果占所述组合后的推荐列表所有的结果的比例, 所述比例 为选择概率; 或者,
预先设置用户选择的权重, 根据所述用户选择的权重得到用户选择各个推 荐系统推荐的结果的权重占组合后的推荐列表所有的结果的权重的比例, 所述 比例为选择概率。
具体的, 初始化图 3中 3个推荐系统, 4叚设各个推荐系统的比重分别为 尸 2( )、 P3 {t) ,
/^ + ι) = ρ''(0 Γ (0 ,∑/^) = ι且∑ = 1
l + η ι ι
表示时刻第个推荐系统的占比
Pi (t + 1)表示时刻之后的下一时刻第个推荐系统的占比
//是更新系数
Λ(0代表时刻第个推荐系统推荐的结果被用户选中的机率
假设推荐系统 1发送的推荐列表 Α为 {il,i2,i3 }, 推荐系统 2发送的推荐列表 B 为 {i2,i3,i4,i5 }, 推荐系统 3未发送推荐列表, 最后组合呈现给用户的推荐列表 list 为 {il,i2,i3,i4,i5 }, 为了避免推荐结果组合排列的位置对用户选择影响,我们把结 果组合后随机安排顺序。 假设用户选择了 {i2,i4}, 其实用户选择的是
{i2(A),i2(B),i4(B)| , 所以 ^( ) = 和 ( = 具体的, 当用户操作行为不同时, 比如:, 用户对 i2的行为是购买, 而对 i4 的行为只是浏览时, 则各个推荐子系统推荐结果被用户选中的选择概率需要考 虑各用户操作行为的权重, 明显购买行为权重要大于浏览行为, 假设购买行为 权 重 为 0.3 和 浏 览 行 为 权 重 为 0.2 , 则 用 户 选择概 率 是 {i2(A)*0.3,i2(B)*0.3,i4(B)*0.21 , 所以
0.3 0.3 + 0.2
0.3 + 0.3 + 0.2 0.3 + 0.3 + 0.2 对 AA (t)来说, 3/8>1/3, 推荐系统 1在第二种情况下的推荐效果要好于第一 种情况, 原因就是推荐系统 1推荐的 i2被用户购买了。
作为另一种可选的方法, 所述方法还包括:
接收用户定期反馈的选择结果, 根据所述定期反馈的选择结果, 更新所述 各个推荐系统的推荐列表。
具体的, 推荐系统 1、 推荐系统 2、 推荐系统 3定期接收用户定期反馈的选择 结果, 根据结果更新自己发送的推荐列表。
本发明通过预先设置的组合策略, 对各个推荐系统发送的推荐列表进行组 合, 实现多个推荐模型或者系统给用户反馈推荐结果, 根据用户反馈的选择概 率, 更新所述组合后的推荐列表, 实现实时评估, 并接收用户定期反馈的选择 结果, 更新所述组合后的推荐列表, 进而同时体现用户的当前兴趣和历史兴趣 爱好。
参考图 4, 图 4是本发明实施例提供的一种推荐的方法的方法示意图。 所述 方法包括以下步骤:
步骤 401, 用户首次登录, 系统无用户任何信息, 则所述推荐前端系统在各 个推荐子系统的推荐列表中等量地选取推荐结果呈现给用户, 或者系统根据组 合策略为各推荐子系统赋初值;
步骤 402, 推荐前端系统获得各个推荐系统的推荐列表进行组合展示; 步骤 403, 用户对推荐的结果进行选择浏览或购买或收藏等操作;
步骤 404, 推荐前端系统实时捕获用户的行为, 赋予不同的用户行为不同的
权重代表用户当前的喜好, 并进行推荐结果的选择概率计算;
步骤 405, 推荐前端系统根据用户当前喜好按一定的规则调整各个推荐系统 的输出在推荐最终结果列表中的比重;
步骤 406, 用户此时可以看到不同于之前的推荐结果;
步骤 407, 用户注销结束此次会话, 各推荐子系统获得此次会话期间用户的 总体行为;
步骤 408, 各推荐子系统根据用户的历史喜好调整推荐列表;
步骤 409, 用户再次登录系统, 跳转到步骤 401。
参考图 5, 图 5是本发明实施例提供的一种服务器的装置结构图。如图 4所 示, 所述服务器包括以下单元:
接收单元 501, 用于接收各个推荐系统发送的推荐列表;
具体的, 如图 3所示, 图 3是本发明实施例提供的一种推荐的方法的方法 示意图。 如图 3所示, 推荐前端系统接收推荐系统 1、 推荐系统 2、 推荐系统 3 发送的推荐列表, 所述推荐前端系统根据组合策略将所述推荐系统 1、推荐系统 2、 推荐系统 3发送的推荐列表进行组合, 所述推荐前端系统将组合后的推荐列 表发送给业务系统, 使得所述业务系统将组合后的推荐列表呈现给用户。 所述 推荐前端系统接收用户实时反馈的选择概率, 用于更新所述组合后的推荐列表, 同时, 推荐系统 1、 推荐系统 2、 推荐系统 3接收用户定期反馈的选择结果, 用 于更新推荐系统 1、 推荐系统 2、 推荐系统 3的推荐列表。
组合单元 502, 用于根据预先设置的组合策略, 将所述推荐列表进行组合, 将组合后的推荐列表呈现给用户, 使得用户根据所述组合后的推荐列表进行选 择;
可选地, 所述组合单元 502, 具体用于:
预先定义各个推荐系统发送的结果在所有推荐系统发送的结果中占的比 重;
根据所述比重, 对各个推荐系统发送的推荐列表进行组合。
具体的,假设推荐系统 1发送给推荐前端系统的推荐列表 A为 {al, a2, a3 }, 推荐系统 2发送给推荐前端系统的推荐列表 B为 {bl, b2, b3 }, 推荐系统 3发 送给推荐前端系统的推荐列表 C 为 {cl, c2, c3 }。 根据预先组合的推荐策略, 4叚设推荐系统 1、 推荐系统 2、 推荐系统 3推荐的结果占所有推荐系统推荐结果
的比重都是 1/3时, 则组合后的推荐列表可为 {al, bl, c2}, 假设推荐系统 1、 推荐系统 2、推荐系统 3推荐的结果占所有推荐系统推荐结果的比重分别是 3/5、 1/5、 1/5时, 则组合后的推荐列表可为 { al, a2, a3, b2, c3 }。
同时, 组合策略的比重可以自由定义, 4叚设儿童节时, 可以将推荐系统 1 的比重提高, 因为推荐系统 1推荐的大多是儿童用品。
更新单元 503, 用于接收用户反馈的选择结果,根据所述选择结果更新所述 组合后的推荐列表。
可选地, 所述更新单元 503具体用于:
接收用户反馈的选择结果, 根据所述选择结果计算选择概率;
预先设置更新系数, 根据所述选择概率、 更新系数和各个推荐系统的比重 计算各个推荐系统更新后的比重;
根据所述更新后的比重更新组合后的推荐列表。
所述更新单元中执行步骤接收用户反馈的选择结果, 根据所述选择结果计 算选择概率, 包括:
计算所述选择结果占所述组合后的推荐列表所有的结果的比例, 所述比例 为选择概率; 或者,
预先设置用户选择的权重, 根据所述用户选择的权重得到用户选择各个推 荐系统推荐的结果的权重占组合后的推荐列表所有的结果的权重的比例, 所述 比例为选择概率。
具体的, 初始化图 3中 3个推荐系统, 4叚设各个推荐系统的比重分别为 尸 2( )、 P3 {t) , l + η ι ι
A. (0表示时刻第个推荐系统的占比
( + 1)表示时刻之后的下一时刻第个推荐系统的占比
是更新系数
代表时刻第个推荐系统推荐的结果被用户选中的机率
假设推荐系统 1发送的推荐列表 Α为 {il,i2,i3 }, 推荐系统 2发送的推荐列表 B 为 {i2,i3,i4,i5 }, 推荐系统 3未发送推荐列表, 最后组合呈现给用户的推荐列表 list 为 {il,i2,i3,i4,i5 }, 为了避免推荐结果组合排列的位置对用户选择影响,我们把结 果组合后随机安排顺序。 假设用户选择了 {i2,i4}, 其实用户选择的是
; i2(A),i2(B),i4(B) } , 所以^ ( ) = ^ ( = | 具体的, 当用户操作行为不同时, 比如: 用户对 i2的行为是购买, 而对 i4的 行为只是浏览时。 , 则各个推荐子系统推荐结果被用户选中的选择概率需要考 虑各用户操作行为的权重, 明显购买行为权重要大于浏览行为, 假设购买行为 权 重 为 0.3 和 浏 览 行 为 权 重 为 0.2 , 则 用 户 选 择 概 率 是 {i2(A)*0.3,i2(B)*0.3,i4(B)*0.2 } , 所以
AA (t) = 03 = -3 ( = 03 + 02 =~5
Α 0.3 + 0.3 + 0.2 8 Β 0.3 + 0.3 + 0.2 8 对 AA (t)来说, 3/8>1/3, 推荐系统 1在第二种情况下的推荐效果要好于第一 种情况, 原因就是推荐系统 1推荐的 i2被用户购买了。
作为一种可选的实施例, 所述服务器还包括定期反馈单元 504, 用于: 接收用户定期反馈的选择结果, 根据所述定期反馈的选择结果, 更新所述 各个推荐系统的推荐列表。
具体的, 推荐系统 1、 推荐系统 2、 推荐系统 3定期接收用户定期反馈的选择 结果, 根据结果更新自己发送的推荐列表。
本发明通过预先设置的组合策略, 对各个推荐系统发送的推荐列表进行组 合, 实现多个推荐模型或者系统给用户反馈推荐结果, 根据用户反馈的选择概 率, 更新所述组合后的推荐列表, 实现实时评估, 并接收用户定期反馈的选择 结果, 更新所述组合后的推荐列表, 进而同时体现用户的当前兴趣和历史兴趣 爱好。
参考图 6,图 6是本发明实施例提供的一种服务器的装置结构图。参考图 6, 图 6是本发明实施例提供的一种服务器 600,本发明具体实施例并不对所述服务 器的具体实现做限定。 所述服务器 600包括:
处理器(processor)601, 通信接口(Communications Interface)602, 存者器 (memory)603 , 总线 604。
处理器 601, 通信接口 602, 存储器 603通过总线 604完成相互间的通信。 通信接口 602, 用于与其他设备进行通信;
处理器 601, 用于执行程序。
具体地, 程序可以包括程序代码, 所述程序代码包括计算机操作指令。
处理器 601 可能是一个中央处理器 CPU, 或者是特定集成电路 ASIC ( Application Specific Integrated Circuit ), 或者是被配置成实施本发明实施例的 一个或多个集成电路。
存储器 603, 用于存放程序 。 存储器 603可能包含高速 RAM存储器, 也 可能还包括非易失性存储器(non-volatile memory ), 例如至少一个磁盘存储器。 程序具体可以包括:
接收各个推荐系统发送的推荐列表;
根据预先设置的组合策略, 将所述推荐列表进行组合, 将组合后的推荐列 表呈现给用户, 使得用户根据所述组合后的推荐列表进行选择;
接收用户反馈的选择结果, 根据所述选择结果更新所述组合后的推荐列表。 所述根据预先设置的组合策略, 将所述推荐列表进行组合, 包括: 预先定义各个推荐系统发送的结果在所有推荐系统发送的结果中占的比 重;
根据所述比重, 对各个推荐系统发送的推荐列表进行组合。
所述接收用户反馈的选择结果, 根据所述选择结果更新所述组合后的推荐 列表, 包括:
接收用户反馈的选择结果, 根据所述选择结果计算选择概率;
预先设置更新系数, 根据所述选择概率、 更新系数和各个推荐系统的比重 计算各个推荐系统更新后的比重;
根据所述更新后的比重更新组合后的推荐列表。
所述接收用户反馈的选择结果, 根据所述选择结果计算选择概率, 包括: 计算所述选择结果占所述组合后的推荐列表所有的结果的比例, 所述比例 为选择概率; 或者,
预先设置用户选择的权重, 根据所述用户选择的权重得到用户选择各个推 荐系统推荐的结果的权重占组合后的推荐列表所有的结果的权重的比例, 所述 比例为选择概率。
所述方法还包括:
接收用户定期反馈的选择结果, 根据所述定期反馈的选择结果, 更新所述 各个推荐系统的推荐列表。
以上所述仅为本发明的优选实施方式, 并不构成对本发明保护范围的限定。
任何在本发明的精神和原则之内所作的任何修改、 等同替换和改进等, 均应包 含在本发明要求包含范围之内。
Claims
1、 一种推荐的方法, 其特征在于, 所述方法包括:
接收各个推荐系统发送的推荐列表;
根据预先设置的组合策略, 将所述推荐列表进行组合, 将组合后的推荐列 表呈现给用户, 使得用户根据所述组合后的推荐列表进行选择;
接收用户反馈的选择结果, 根据所述选择结果更新所述组合后的推荐列表。
2、 根据权利要求 1所述的方法, 其特征在于, 所述根据预先设置的组合策 略, 将所述推荐列表进行组合, 包括:
预先定义各个推荐系统发送的结果在所有推荐系统发送的结果中占的比 重;
根据所述比重, 对各个推荐系统发送的推荐列表进行组合。
3、 根据权利要求 1或 2所述的方法, 其特征在于, 所述接收用户反馈的选 择结果, 根据所述选择结果更新所述组合后的推荐列表, 包括:
接收用户反馈的选择结果, 根据所述选择结果计算选择概率;
预先设置更新系数, 根据所述选择概率、 更新系数和各个推荐系统的比重 计算各个推荐系统更新后的比重;
根据所述更新后的比重更新组合后的推荐列表。
4、 根据权利要求 3所述的方法, 其特征在于, 所述接收用户反馈的选择结 果, 根据所述选择结果计算选择概率, 包括:
计算所述选择结果占所述组合后的推荐列表所有的结果的比例, 所述比例 为选择概率; 或者,
预先设置用户选择的权重, 根据所述用户选择的权重得到用户选择各个推 荐系统推荐的结果的权重占组合后的推荐列表所有的结果的权重的比例, 所述 比例为选择概率。
5、 根据权利要求 1-4任意一项所述的方法, 其特征在于, 所述方法还包括: 接收用户定期反馈的选择结果, 根据所述定期反馈的选择结果, 更新所述 各个推荐系统的推荐列表。
6、 一种服务器, 其特征在于, 所述服务器包括:
接收单元, 用于接收各个推荐系统发送的推荐列表;
组合单元, 用于根据预先设置的组合策略, 将所述推荐列表进行组合, 将
组合后的推荐列表呈现给用户, 使得用户根据所述组合后的推荐列表进行选择; 更新单元, 用于接收用户反馈的选择结果, 根据所述选择结果更新所述组 合后的推荐列表。
7、 根据权利要求 6所述的服务器, 其特征在于, 所述组合单元具体用于: 预先定义各个推荐系统发送的结果在所有推荐系统发送的结果中占的比 重;
根据所述比重, 对各个推荐系统发送的推荐列表进行组合。
8、 根据权利要求 6或 7所述的服务器, 其特征在于, 所述更新单元具体用 于:
接收用户反馈的选择结果, 根据所述选择结果计算选择概率;
预先设置更新系数, 根据所述选择概率、 更新系数和各个推荐系统的比重 计算各个推荐系统更新后的比重;
根据所述更新后的比重更新组合后的推荐列表。
9、 根据权利要求 8所述的服务器, 其特征在于, 所述更新单元中执行步骤 接收用户反馈的选择结果, 根据所述选择结果计算选择概率, 包括:
计算所述选择结果占所述组合后的推荐列表所有的结果的比例, 所述比例 为选择概率; 或者,
预先设置用户选择的权重, 根据所述用户选择的权重得到用户选择各个推 荐系统推荐的结果的权重占组合后的推荐列表所有的结果的权重的比例, 所述 比例为选择概率。
10、 根据权利要求 6-9任意一项所述的服务器, 其特征在于, 所述服务器还 包括定期反馈单元, 用于:
接收用户定期反馈的选择结果, 根据所述定期反馈的选择结果, 更新所述 各个推荐系统的推荐列表。
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310145097.0A CN104123284B (zh) | 2013-04-24 | 2013-04-24 | 一种推荐的方法及服务器 |
CN201310145097.0 | 2013-04-24 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2014173237A1 true WO2014173237A1 (zh) | 2014-10-30 |
Family
ID=51768697
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2014/075183 WO2014173237A1 (zh) | 2013-04-24 | 2014-04-11 | 一种推荐的方法及服务器 |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN104123284B (zh) |
WO (1) | WO2014173237A1 (zh) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10718539B2 (en) | 2013-03-15 | 2020-07-21 | Google Llc | Controlling an HVAC system in association with a demand-response event |
Families Citing this family (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106776660A (zh) | 2015-11-25 | 2017-05-31 | 阿里巴巴集团控股有限公司 | 一种信息推荐方法及装置 |
CN106056231B (zh) * | 2016-03-29 | 2021-04-23 | 北京四海道达网络科技有限公司 | 基于多方位置和时间可用性的自动匹配推荐方法和装置 |
CN105913286B (zh) * | 2016-05-16 | 2020-02-11 | 达而观信息科技(上海)有限公司 | 一种自动融合多种个性化推荐模型的方法 |
CN107734395A (zh) * | 2017-11-07 | 2018-02-23 | 山东浪潮商用系统有限公司 | 一种节目推荐方法、装置、可读介质及存储控制器 |
CN108322501B (zh) * | 2017-12-22 | 2021-04-27 | 深圳创新科软件技术有限公司 | 一种消息推送方法和装置 |
CN108419134B (zh) * | 2018-02-05 | 2020-02-18 | 华南理工大学 | 基于个体历史与群体当前行为融合的频道推荐方法 |
CN108763502B (zh) * | 2018-05-30 | 2022-03-25 | 腾讯科技(深圳)有限公司 | 信息推荐方法和系统 |
CN109063104B (zh) * | 2018-07-27 | 2020-11-10 | 百度在线网络技术(北京)有限公司 | 推荐信息的刷新方法、装置、存储介质和终端设备 |
CN109299368B (zh) * | 2018-09-29 | 2020-11-24 | 北京思路创新科技有限公司 | 一种用于环境信息资源ai智能个性化推荐的方法及系统 |
CN109544276B (zh) * | 2018-10-30 | 2022-07-15 | 珠海格力电器股份有限公司 | 一种产品选型方法及处理器 |
CN111061956B (zh) * | 2019-12-24 | 2022-08-16 | 北京百度网讯科技有限公司 | 用于生成信息的方法和装置 |
CN113159810B (zh) * | 2020-01-22 | 2024-06-04 | 百度在线网络技术(北京)有限公司 | 策略评估方法、装置、设备及存储介质 |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101888515A (zh) * | 2010-06-30 | 2010-11-17 | 中山大学 | 一种数字电视节目预订播放方法及系统 |
CN102685565A (zh) * | 2012-05-18 | 2012-09-19 | 合一网络技术(北京)有限公司 | 一种点击反馈式个性化推荐系统 |
CN202711318U (zh) * | 2012-04-09 | 2013-01-30 | 唐慈津 | 综合型网络商品比较系统 |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10664889B2 (en) * | 2008-04-01 | 2020-05-26 | Certona Corporation | System and method for combining and optimizing business strategies |
-
2013
- 2013-04-24 CN CN201310145097.0A patent/CN104123284B/zh active Active
-
2014
- 2014-04-11 WO PCT/CN2014/075183 patent/WO2014173237A1/zh active Application Filing
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101888515A (zh) * | 2010-06-30 | 2010-11-17 | 中山大学 | 一种数字电视节目预订播放方法及系统 |
CN202711318U (zh) * | 2012-04-09 | 2013-01-30 | 唐慈津 | 综合型网络商品比较系统 |
CN102685565A (zh) * | 2012-05-18 | 2012-09-19 | 合一网络技术(北京)有限公司 | 一种点击反馈式个性化推荐系统 |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10718539B2 (en) | 2013-03-15 | 2020-07-21 | Google Llc | Controlling an HVAC system in association with a demand-response event |
Also Published As
Publication number | Publication date |
---|---|
CN104123284B (zh) | 2018-01-23 |
CN104123284A (zh) | 2014-10-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2014173237A1 (zh) | 一种推荐的方法及服务器 | |
US11659050B2 (en) | Discovering signature of electronic social networks | |
Liu et al. | Deep reinforcement learning based recommendation with explicit user-item interactions modeling | |
US10430464B1 (en) | Scalable graph propagation for knowledge expansion | |
Liu et al. | State representation modeling for deep reinforcement learning based recommendation | |
US9524472B2 (en) | Distributed scalable incrementally updated models in decisioning systems | |
CN104854619B (zh) | 对社交活动流软张贴 | |
US8527447B2 (en) | Collaborative networking with optimized information quality assessment | |
CN109313722A (zh) | 存储器高效的基于时间的反向传播 | |
TWI793170B (zh) | 使用深度神經語言模型於自動出價之系統、裝置與方法 | |
CN109636490A (zh) | 广告转化率的实时预测方法、广告评价方法与系统 | |
CN106663228A (zh) | 用于商业应用系统的规则到约束翻译器 | |
US10936601B2 (en) | Combined predictions methodology | |
US20160379122A1 (en) | Recommendation algorithm optimization method, device and system | |
CN104992348B (zh) | 一种信息展示的方法和装置 | |
CN108629608A (zh) | 用户数据处理方法及装置 | |
CN111125420B (zh) | 基于人工智能的对象推荐方法、装置及电子设备 | |
CN111492394A (zh) | 出席者参与度确定系统和方法 | |
CN107609958A (zh) | 行为引导策略确定方法以及装置、存储介质以及电子设备 | |
CN110727761A (zh) | 对象信息获取方法、装置及电子设备 | |
CN110472798A (zh) | 时间序列数据的预测方法、装置及计算机可读存储介质 | |
US20210374577A1 (en) | Cross-domain action prediction | |
WO2016172610A1 (en) | System and method for matching dynamically validated network data | |
Zhao et al. | A Retweet Number Prediction Model Based on Followers' Retweet Intention and Influence | |
WO2013131335A1 (zh) | 一种在网络社区中发布信息的方法与设备 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 14788638 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 14788638 Country of ref document: EP Kind code of ref document: A1 |