WO2019071897A1 - 实时推荐方法、电子设备及计算机可读存储介质 - Google Patents

实时推荐方法、电子设备及计算机可读存储介质 Download PDF

Info

Publication number
WO2019071897A1
WO2019071897A1 PCT/CN2018/076173 CN2018076173W WO2019071897A1 WO 2019071897 A1 WO2019071897 A1 WO 2019071897A1 CN 2018076173 W CN2018076173 W CN 2018076173W WO 2019071897 A1 WO2019071897 A1 WO 2019071897A1
Authority
WO
WIPO (PCT)
Prior art keywords
real
model
user
time
digital
Prior art date
Application number
PCT/CN2018/076173
Other languages
English (en)
French (fr)
Inventor
许开河
兰相如
王建明
肖京
Original Assignee
平安科技(深圳)有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 平安科技(深圳)有限公司 filed Critical 平安科技(深圳)有限公司
Publication of WO2019071897A1 publication Critical patent/WO2019071897A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/28Databases characterised by their database models, e.g. relational or object models
    • G06F16/289Object oriented databases
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/22Indexing; Data structures therefor; Storage structures
    • G06F16/2228Indexing structures
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/24Querying
    • G06F16/245Query processing
    • G06F16/2455Query execution

Definitions

  • the present application relates to the field of computer information technology, and in particular, to a real-time recommendation method, an electronic device, and a computer readable storage medium.
  • the data analysis tools of most big data engines can only process offline data.
  • the corresponding real-time processing components (such as the spark-streaming component) require cluster support of the big data engine, resulting in high cost and long time (100ms). -2s), can not meet the requirements of real-time recommendation low time consumption.
  • the recommendation model trained by the current big data engine cannot be updated on-line to generate benefits, and it is not convenient for multiple models to perform A/B testing online at the same time. Therefore, the design of the real-time recommendation method in the prior art is not reasonable enough and needs to be improved.
  • the present application proposes a real-time recommendation method, an electronic device, and a computer-readable storage medium, which shortens the prediction time by real-time computing background operation, and realizes online management of multiple models simultaneously by using the front-end interface system.
  • the present application provides an electronic device including a memory and a processor, wherein the memory stores a real-time recommendation system operable on the processor, and the real-time recommendation system is The processor implements the following steps when executed:
  • the user feature is stored into the cache area through the real-time computing background
  • mapping the user identifier to a number in a specified interval and assigning users in different digital intervals to different models
  • the following steps are also implemented:
  • the prediction effect of the specific model is higher than the preset threshold, the user number interval allocated by the specific model is increased.
  • the following steps are also implemented:
  • the user digital interval allocated by the specific model is increased by a second ratio.
  • the model class inherits a unified parent class, and all model implementation classes perform prediction recommendation through a unified parent class interface;
  • the different digital interval includes a first digital interval and a second digital interval, and the corresponding user in the first digital interval is assigned to the first model, and the corresponding user in the second digital interval is assigned to the second model.
  • the following steps are also implemented:
  • the prediction effects of different models are written into the cache area in real time, and the prediction effects of different models are displayed on the set front end interface system through a specific chart format;
  • the key data of each forecasting item is generated according to the chronological statistics, and the generated report is displayed on the set front-end interface system.
  • the present application further provides a real-time recommendation method, which is applied to an electronic device, and the method includes:
  • the user feature is stored into the cache area through the real-time computing background
  • mapping the user identifier to a number in a specified interval and assigning users in different digital intervals to different models
  • the method further comprises:
  • the prediction effect of the specific model is higher than the preset threshold, the user number interval allocated by the specific model is increased.
  • the method further comprises:
  • the user digital interval allocated by the specific model is increased by a second ratio.
  • the model class inherits a unified parent class, and all model implementation classes perform prediction recommendation through a unified parent class interface;
  • the different digital interval includes a first digital interval and a second digital interval, and the corresponding user in the first digital interval is assigned to the first model, and the corresponding user in the second digital interval is assigned to the second model.
  • the present application further provides a computer readable storage medium storing a real-time recommendation system, the real-time recommendation system being executable by at least one processor to enable the At least one processor performs the steps of the real-time recommendation method as described above.
  • the electronic device, the real-time recommendation method and the computer-readable storage medium proposed by the present application shorten the prediction time by real-time computing background operation, and realize the line of multiple models simultaneously by using the front-end interface system.
  • the application completes the closed loop from model training to online to effect evaluation.
  • the created model can quickly update the iterative model, and can also be easily scaled horizontally to access more recommended services.
  • 1 is a schematic diagram of an optional hardware architecture of an electronic device of the present application
  • FIG. 2 is a schematic diagram of a program module of an embodiment of a real-time recommendation system in an electronic device of the present application
  • FIG. 3 is a schematic flowchart of an implementation manner of an embodiment of a real-time recommendation method according to the present application
  • FIG. 4 is a diagram showing an example of a model effect graph outputted by the present application.
  • first, second and the like in the present application are for the purpose of description only, and are not to be construed as indicating or implying their relative importance or implicitly indicating the number of technical features indicated. .
  • features defining “first” and “second” may include at least one of the features, either explicitly or implicitly.
  • the technical solutions between the various embodiments may be combined with each other, but must be based on the realization of those skilled in the art, and when the combination of the technical solutions is contradictory or impossible to implement, it should be considered that the combination of the technical solutions does not exist. Nor is it within the scope of protection required by this application.
  • FIG. 1 it is a schematic diagram of an optional hardware architecture of the electronic device 2 of the present application.
  • the electronic device 2 may include, but is not limited to, a memory 21, a processor 22, and a network interface 23 that can communicate with each other through a system bus. It is pointed out that FIG. 1 only shows the electronic device 2 with the components 21-23, but it should be understood that not all illustrated components are required to be implemented, and more or fewer components may be implemented instead.
  • the electronic device 2 may be a computing device such as a rack server, a blade server, a tower server, or a rack server.
  • the electronic device 2 may be an independent server or a server cluster composed of multiple servers. .
  • the memory 21 includes at least one type of readable storage medium including a flash memory, a hard disk, a multimedia card, a card type memory (eg, SD or DX memory, etc.), a random access memory (RAM), a static Random access memory (SRAM), read only memory (ROM), electrically erasable programmable read only memory (EEPROM), programmable read only memory (PROM), magnetic memory, magnetic disk, optical disk, and the like.
  • the memory 21 may be an internal storage unit of the electronic device 2, such as a hard disk or memory of the electronic device 2.
  • the memory 21 may also be an external storage device of the electronic device 2, such as a plug-in hard disk equipped on the electronic device 2, a smart memory card (SMC), and a secure digital device. (Secure Digital, SD) card, flash card, etc.
  • the memory 21 may also include both an internal storage unit of the electronic device 2 and an external storage device thereof.
  • the memory 21 is generally used to store an operating system installed in the electronic device 2 and various types of application software, such as program codes of the real-time recommendation system 20. Further, the memory 21 can also be used to temporarily store various types of data that have been output or are to be output.
  • the processor 22 may be a Central Processing Unit (CPU), controller, microcontroller, microprocessor, or other data processing chip in some embodiments.
  • the processor 22 is typically used to control the overall operation of the electronic device 2, such as performing control and processing related to data interaction or communication with the electronic device 2.
  • the processor 22 is configured to run program code or process data stored in the memory 21, such as running the real-time recommendation system 20 or the like.
  • the network interface 23 may comprise a wireless network interface or a wired network interface, which is typically used to establish a communication connection between the electronic device 2 and other electronic devices.
  • the network interface 23 is configured to connect the electronic device 2 to an external data platform through a network, and establish a data transmission channel and a communication connection between the electronic device 2 and an external data platform.
  • the network may be an intranet, an Internet, a Global System of Mobile communication (GSM), a Wideband Code Division Multiple Access (WCDMA), a 4G network, or a 5G network.
  • Wireless or wired networks such as network, Bluetooth, Wi-Fi, etc.
  • the real-time recommendation system 20 may be divided into one or more program modules, and the one or more program modules are stored in the memory 21 and are processed by one or more processors (this The embodiment is executed by the processor 22) to complete the application.
  • the real-time recommendation system 20 can be divided into a training module 201, a storage module 202, an allocation module 203, and a recommendation module 204.
  • the program module referred to in the present application refers to a series of computer program instruction segments capable of performing a specific function, and is more suitable than the program to describe the execution process of the real-time recommendation system 20 in the electronic device 2.
  • the functions of each of the program modules 201-204 will be described in detail below.
  • the training module 201 is configured to create a model class by using a predetermined programming language (such as the Scala programming language), use the model class to train the model offline on a specific computing engine (such as the Spark big data engine), and host the trained model.
  • a predetermined programming language such as the Scala programming language
  • the model obtained by training the model class created by the Scala programming voice is managed by the Java real-time computing background, thereby realizing real-time prediction and recommendation, and reducing the prediction and recommendation time.
  • the model class inherits a unified parent class, and all model implementation classes perform prediction recommendation through a unified parent class interface, thereby facilitating horizontal expansion and accessing more recommended services.
  • parent class of the model class can be defined as the following format:
  • the real-time computing background (such as Java background) can call this method for real-time prediction.
  • the storage module 202 is configured to store user characteristics (such as user portrait features, which may be large data levels) to a cache area (such as a Redis database) through the real-time computing background.
  • user characteristics such as user portrait features, which may be large data levels
  • a cache area such as a Redis database
  • the user portrait feature query time can be greatly shortened, thereby shortening the prediction time (the average time of a single prediction is 2 ms).
  • the allocating module 203 is configured to map a user identifier (user ID) into a number in a specified interval (such as a number between 0-100) through the real-time computing background, and allocate a number in a different number interval.
  • Users to different models (ie trained models). For example, after the user identifier is hashed to a number between 0 and 100, it is divided into a first number interval "0-50" and a second number interval "51-100", and the corresponding user in the first number interval is assigned to The first model (such as the A model), the corresponding user in the second number interval is assigned to the second model (such as the B model).
  • the recommendation module 204 is configured to acquire, by using the real-time computing background, different models, user characteristics corresponding to users in different digital sections from the cache area, and perform user behavior prediction according to the acquired user features (eg, predicting users Whether to purchase property insurance) and recommend the forecast results to users in real time.
  • the acquired user features eg, predicting users Whether to purchase property insurance
  • calling the first model to obtain a user feature corresponding to the user in the first digital interval from the cache area performing user behavior prediction for the user in the first digital interval according to the acquired user feature, and recommending the prediction result to the real-time recommendation to a user in the first digital interval; invoking a second model to obtain a user feature corresponding to the user in the second digital interval from the cache area, and performing user behavior prediction on the user in the second digital interval according to the acquired user feature, and The predicted results are recommended in real time to users within the second digital interval.
  • the real-time recommendation system 20 is further configured to:
  • the size of different digital intervals is dynamically adjusted to flexibly control the predicted user traffic covered by each model, so that the model with better prediction effect is allocated more users.
  • the prediction effect of the specific model is higher than a preset threshold (eg, 80%)
  • the user number interval allocated by the specific model is increased (eg, by 20%).
  • the prediction effect of the first model is 85%
  • the first digital interval “0-50” allocated by the first model is increased by 20%
  • the adjusted first digital interval “0-60 is obtained.
  • the second digital interval "51-100" assigned by the second model is reduced by 20%, and the adjusted second digital interval "61-100” is obtained.
  • the prediction effect of the specific model is higher than the first preset threshold (eg, 80%), the user digital interval allocated by the specific model is increased by the first ratio (eg, increased by 20). %); If the prediction effect of a particular model is higher than the second preset threshold (such as 90%), then the user number interval allocated for the particular model is increased by a second ratio (eg, 50% increase).
  • the first preset threshold e.g, 80%
  • the second preset threshold such as 90%
  • the real-time recommendation system 20 is further configured to:
  • the prediction effects of different models are written to the cache area in real time, and the prediction effects of different models are displayed on the set front end interface system (such as the application interface system) through a specific chart format (refer to the graph shown in FIG. 4). ), so that you can compare the pros and cons of different models in real time, and facilitate online management of multiple models at the same time (such as A/B test management, etc.).
  • the real-time recommendation system 20 is further configured to:
  • the key data of each forecasting project (or forecasting business) (such as core user information) is statistically generated according to the chronological order (such as day/week/month/year), and the generated report is displayed on the set front-end interface. system.
  • the real-time recommendation system 20 proposed by the present application shortens the prediction time by real-time calculation of the background operation, and realizes online management of multiple models simultaneously by using the front-end interface system.
  • the application completes the closed loop from model training to online to effect evaluation.
  • the created model can quickly update the iterative model, and can also be easily scaled horizontally to access more recommended services.
  • the present application also proposes a real-time recommendation method.
  • FIG. 3 it is a schematic flowchart of an implementation process of an embodiment of the real-time recommendation method of the present application.
  • the order of execution of the steps in the flowchart shown in FIG. 3 may be changed according to different requirements, and some steps may be omitted.
  • Step S31 creating a model class by using a predetermined programming language (such as the Scala programming language), using the model class to train the model offline on a specific computing engine (such as the Spark big data engine), and hosting the trained model to the real-time computing background (such as Java background).
  • a predetermined programming language such as the Scala programming language
  • the model obtained by training the model class created by the Scala programming voice is managed by the Java real-time computing background, thereby realizing real-time prediction and recommendation, and reducing the prediction and recommendation time.
  • the model class inherits a unified parent class, and all model implementation classes perform prediction and recommendation through a unified parent class interface, thereby facilitating horizontal expansion and accessing more recommended services.
  • parent class of the model class can be defined as the following format:
  • the real-time computing background (such as Java background) can call this method for real-time prediction.
  • Step S32 through the real-time computing background, store user characteristics (such as user portrait features, which may be large data magnitude) to a cache area (such as a Redis database).
  • user characteristics such as user portrait features, which may be large data magnitude
  • a cache area such as a Redis database
  • the user portrait feature query time can be greatly shortened, thereby shortening the prediction time (the average time of a single prediction is 2 ms).
  • Step S33 by using the real-time computing background, mapping the user identifier (user ID) into a number in a specified interval (such as a hash to a number between 0-100), and assigning users in different digital intervals to different models.
  • a specified interval such as a hash to a number between 0-100
  • assigning users in different digital intervals to different models. ie the model obtained by training. For example, after the user identifier is hashed to a number between 0 and 100, it is divided into a first number interval "0-50" and a second number interval "51-100", and the corresponding user in the first number interval is assigned to The first model (such as the A model), the corresponding user in the second number interval is assigned to the second model (such as the B model).
  • Step S34 in the real-time computing background, calling different models to obtain user characteristics corresponding to users in different digital sections from the cache area, and performing user behavior prediction according to the acquired user characteristics (such as predicting whether the user purchases property insurance) And recommend the forecast results to the user in real time.
  • calling the first model to obtain a user feature corresponding to the user in the first digital interval from the cache area performing user behavior prediction for the user in the first digital interval according to the acquired user feature, and recommending the prediction result to the real-time recommendation to a user in the first digital interval; invoking a second model to obtain a user feature corresponding to the user in the second digital interval from the cache area, and performing user behavior prediction on the user in the second digital interval according to the acquired user feature, and The predicted results are recommended in real time to users within the second digital interval.
  • the real-time recommendation method further includes the following steps:
  • the size of different digital intervals is dynamically adjusted to flexibly control the predicted user traffic covered by each model, so that the model with better prediction effect is allocated more users.
  • the prediction effect of the specific model is higher than a preset threshold (eg, 80%)
  • the user number interval allocated by the specific model is increased (eg, by 20%).
  • the prediction effect of the first model is 85%
  • the first digital interval “0-50” allocated by the first model is increased by 20%
  • the adjusted first digital interval “0-60 is obtained.
  • the second digital interval "51-100" assigned by the second model is reduced by 20%, and the adjusted second digital interval "61-100” is obtained.
  • the prediction effect of the specific model is higher than the first preset threshold (eg, 80%), the user digital interval allocated by the specific model is increased by the first ratio (eg, increased by 20). %); If the prediction effect of a particular model is higher than the second preset threshold (such as 90%), then the user number interval allocated for the particular model is increased by a second ratio (eg, 50% increase).
  • the first preset threshold e.g, 80%
  • the second preset threshold such as 90%
  • the real-time recommendation method further includes the following steps:
  • the prediction effects of different models are written to the cache area in real time, and the prediction effects of different models are displayed on the set front end interface system (such as the application interface system) through a specific chart format (refer to the graph shown in FIG. 4). ), so that you can compare the pros and cons of different models in real time, and facilitate online management of multiple models at the same time (such as A/B test management, etc.).
  • the real-time recommendation method further includes the following steps:
  • the key data of each forecasting project (or forecasting business) (such as core user information) is statistically generated according to the chronological order (such as day/week/month/year), and the generated report is displayed on the set front-end interface. system.
  • the real-time recommendation method proposed by the present application shortens the prediction time by real-time computing background operation, and realizes online management of multiple models simultaneously by using the front-end interface system.
  • the application completes the closed loop from model training to online to effect evaluation.
  • the created model can quickly update the iterative model, and can also be easily scaled horizontally to access more recommended services.
  • the present application further provides a computer readable storage medium (such as a ROM/RAM, a magnetic disk, an optical disk), the computer readable storage medium storing a real-time recommendation system 20, the real-time recommendation system 20 may be executed by at least one processor 22 to cause the at least one processor 22 to perform the steps of the real-time recommendation method as described above.
  • a computer readable storage medium such as a ROM/RAM, a magnetic disk, an optical disk
  • the computer readable storage medium storing a real-time recommendation system 20
  • the real-time recommendation system 20 may be executed by at least one processor 22 to cause the at least one processor 22 to perform the steps of the real-time recommendation method as described above.
  • the foregoing embodiment method can be implemented by means of software plus a necessary general hardware platform, and can also be implemented by hardware, but in many cases, the former is A better implementation.
  • the technical solution of the present application which is essential or contributes to the prior art, may be embodied in the form of a software product stored in a storage medium (such as ROM/RAM, disk,
  • the optical disc includes a number of instructions for causing a terminal device (which may be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to perform the methods described in various embodiments of the present application.

Abstract

一种实时推荐方法,该方法包括步骤:通过预定编程语言创建模型类,利用该模型类在特定的计算引擎上离线训练模型,并将训练得到的模型托管至实时计算后台(S31);通过所述实时计算后台,将用户特征存储至高速缓存区(S32);将用户标识映射成指定区间内的数字,并分配不同数字区间内的用户至不同的模型(S33);调用不同的模型从所述高速缓存区获取不同数字区间内的用户对应的用户特征,根据获取的用户特征进行用户行为预测,并将预测结果实时推荐至用户(S34)。可以快速更新迭代模型,方便横向扩展。

Description

实时推荐方法、电子设备及计算机可读存储介质
本申请要求于2017年10月13日提交中国专利局、申请号为201710953455.9、发明名称为“实时推荐方法、电子设备及计算机可读存储介质”的中国专利申请的优先权,其全部内容通过引用结合在申请中。
技术领域
本申请涉及计算机信息技术领域,尤其涉及一种实时推荐方法、电子设备及计算机可读存储介质。
背景技术
目前,多数大数据引擎(如Spark引擎)的数据分析工具只能处理离线的数据,对应的实时处理组件(如spark-streaming组件)需要大数据引擎的集群支持,导致成本高、耗时长(100ms-2s),不能满足实时推荐低耗时的要求。另外,目前的大数据引擎训练出来的推荐模型不能及时更新到线上产生效益,并且多个模型同时在线上进行A/B测试也不方便进行管理。故,现有技术中的实时推荐方法设计不够合理,亟需改进。
发明内容
有鉴于此,本申请提出一种实时推荐方法、电子设备及计算机可读存储介质,通过实时计算后台的运作缩短了预测时间,并利用前端界面系统实现了同时对多个模型进行线上管理。
首先,为实现上述目的,本申请提出一种电子设备,所述电子设备包括存储器和处理器,所述存储器上存储有可在所述处理器上运行的实时推荐系统,所述实时推荐系统被所述处理器执行时实现如下步骤:
通过预定编程语言创建模型类,利用该模型类在特定的计算引擎上离线 训练模型,并将训练得到的模型托管至实时计算后台;
通过所述实时计算后台,将用户特征存储至高速缓存区;
通过所述实时计算后台,将用户标识映射成指定区间内的数字,并分配不同数字区间内的用户至不同的模型;及
通过所述实时计算后台,调用不同的模型从所述高速缓存区获取不同数字区间内的用户对应的用户特征,根据获取的用户特征进行用户行为预测,并将预测结果实时推荐至用户。
优选地,所述实时推荐系统被所述处理器执行时还用于实现如下步骤:
根据不同模型的预测效果,动态调节不同数字区间的大小;及
若特定模型的预测效果高于预设阈值,则增大该特定模型所分配的用户数字区间。
优选地,所述实时推荐系统被所述处理器执行时还用于实现如下步骤:
若特定模型的预测效果高于第一预设阈值,则将该特定模型所分配的用户数字区间增大第一比例;及
若特定模型的预测效果高于第二预设阈值,则将该特定模型所分配的用户数字区间增大第二比例。
优选地,所述模型类继承一个统一的父类,所有模型的实现类通过统一的父类接口进行预测推荐;及
所述不同数字区间包括第一数字区间和第二数字区间,并将第一数字区间内对应的用户分配至第一模型,第二数字区间内对应的用户分配至第二模型。
优选地,所述实时推荐系统被所述处理器执行时还用于实现如下步骤:
将不同模型的预测效果实时写入所述高速缓存区,并通过特定图表格式将不同模型的预测效果显示于设定的前端界面系统;及
将每个预测项目的关键数据依据时间顺序统计生成报表,并将生成的报表显示于设定的前端界面系统。
此外,为实现上述目的,本申请还提供一种实时推荐方法,该方法应用于电子设备,所述方法包括:
通过预定编程语言创建模型类,利用该模型类在特定的计算引擎上离线训练模型,并将训练得到的模型托管至实时计算后台;
通过所述实时计算后台,将用户特征存储至高速缓存区;
通过所述实时计算后台,将用户标识映射成指定区间内的数字,并分配不同数字区间内的用户至不同的模型;及
通过所述实时计算后台,调用不同的模型从所述高速缓存区获取不同数字区间内的用户对应的用户特征,根据获取的用户特征进行用户行为预测,并将预测结果实时推荐至用户。
优选地,该方法还包括:
根据不同模型的预测效果,动态调节不同数字区间的大小;及
若特定模型的预测效果高于预设阈值,则增大该特定模型所分配的用户数字区间。
优选地,该方法还包括:
若特定模型的预测效果高于第一预设阈值,则将该特定模型所分配的用户数字区间增大第一比例;及
若特定模型的预测效果高于第二预设阈值,则将该特定模型所分配的用户数字区间增大第二比例。
优选地,所述模型类继承一个统一的父类,所有模型的实现类通过统一的父类接口进行预测推荐;及
所述不同数字区间包括第一数字区间和第二数字区间,并将第一数字区间内对应的用户分配至第一模型,第二数字区间内对应的用户分配至第二模型。
进一步地,为实现上述目的,本申请还提供一种计算机可读存储介质,所述计算机可读存储介质存储有实时推荐系统,所述实时推荐系统可被至少一个处理器执行,以使所述至少一个处理器执行如上述的实时推荐方法的步骤。
相较于现有技术,本申请所提出的电子设备、实时推荐方法及计算机可读存储介质,通过实时计算后台的运作缩短了预测时间,并利用前端界面系统实现了同时对多个模型进行线上管理。本申请完成了从模型训练到上线到效果评估的闭环,通过创建的模型类可以快速更新迭代模型,同时也可以方便地横向扩展,接入更多的推荐业务。
附图说明
图1是本申请电子设备一可选的硬件架构的示意图;
图2是本申请电子设备中实时推荐系统一实施例的程序模块示意图;
图3为本申请实时推荐方法一实施例的实施流程示意图;
图4为本申请输出的模型效果曲线图的示例图。
附图标记:
电子设备 2
存储器 21
处理器 22
网络接口 23
实时推荐系统 20
训练模块 201
存储模块 202
分配模块 203
推荐模块 204
流程步骤 S31-S34
本申请目的的实现、功能特点及优点将结合实施例,参照附图做进一步说明。
具体实施方式
为了使本申请的目的、技术方案及优点更加清楚明白,以下结合附图及实施例,对本申请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用以解释本申请,并不用于限定本申请。基于本申请中的实施例,本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例,都属于本申请保护的范围。
需要说明的是,在本申请中涉及“第一”、“第二”等的描述仅用于描述目的,而不能理解为指示或暗示其相对重要性或者隐含指明所指示的技术特征的数量。由此,限定有“第一”、“第二”的特征可以明示或者隐含地包括至少一个该特征。另外,各个实施例之间的技术方案可以相互结合,但是必须是以本领域普通技术人员能够实现为基础,当技术方案的结合出现相互矛盾或无法实现时应当认为这种技术方案的结合不存在,也不在本申请要求的保护范围之内。
进一步需要说明的是,在本文中,术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含,从而使得包括一系列要素的过程、方法、物品或者装置不仅包括那些要素,而且还包括没有明确列出的其他要素,或者是还包括为这种过程、方法、物品或者装置所固有的要素。在没有更多限制的情况下,由语句“包括一个……”限定的要素,并不排除在包括该要素的过程、方法、物品或者装置中还存在另外的相同要素。
首先,本申请提出一种电子设备2。
参阅图1所示,是本申请电子设备2一可选的硬件架构的示意图。本实施例中,所述电子设备2可包括,但不限于,可通过系统总线相互通信连接存储器21、处理器22、网络接口23。需要指出的是,图1仅示出了具有组件21-23的电子设备2,但是应理解的是,并不要求实施所有示出的组件,可以替代的实施更多或者更少的组件。
其中,所述电子设备2可以是机架式服务器、刀片式服务器、塔式服务器或机柜式服务器等计算设备,该电子设备2可以是独立的服务器,也可以是多个服务器所组成的服务器集群。
所述存储器21至少包括一种类型的可读存储介质,所述可读存储介质包括闪存、硬盘、多媒体卡、卡型存储器(例如,SD或DX存储器等)、随机访问存储器(RAM)、静态随机访问存储器(SRAM)、只读存储器(ROM)、电可擦除可编程只读存储器(EEPROM)、可编程只读存储器(PROM)、磁性存储器、磁盘、光盘等。在一些实施例中,所述存储器21可以是所述电子设备2的内部存储单元,例如该电子设备2的硬盘或内存。在另一些实施例中,所述存储器21也可以是所述电子设备2的外部存储设备,例如该电子设备2上配备的插接式硬盘,智能存储卡(Smart Media Card,SMC),安全数字(Secure Digital,SD)卡,闪存卡(Flash Card)等。当然,所述存储器21还可以既包括所述电子设备2的内部存储单元也包括其外部存储设备。本实施例中,所述存储器21通常用于存储安装于所述电子设备2的操作系统和各类应用软件,例如所述实时推荐系统20的程序代码等。此外,所述存储器21还可以用于暂时地存储已经输出或者将要输出的各类数据。
所述处理器22在一些实施例中可以是中央处理器(Central Processing Unit,CPU)、控制器、微控制器、微处理器、或其他数据处理芯片。该处理器22通常用于控制所述电子设备2的总体操作,例如执行与所述电子设备2进行数据交互或者通信相关的控制和处理等。本实施例中,所述处理器22用于运行所述存储器21中存储的程序代码或者处理数据,例如运行所述的实时 推荐系统20等。
所述网络接口23可包括无线网络接口或有线网络接口,该网络接口23通常用于在所述电子设备2与其他电子设备之间建立通信连接。例如,所述网络接口23用于通过网络将所述电子设备2与外部数据平台相连,在所述电子设备2与外部数据平台之间的建立数据传输通道和通信连接。所述网络可以是企业内部网(Intranet)、互联网(Internet)、全球移动通讯系统(Global System of Mobile communication,GSM)、宽带码分多址(Wideband Code Division Multiple Access,WCDMA)、4G网络、5G网络、蓝牙(Bluetooth)、Wi-Fi等无线或有线网络。
至此,己经详细介绍了本申请各个实施例的应用环境和相关设备的硬件结构和功能。下面,将基于上述应用环境和相关设备,提出本申请的各个实施例。
参阅图2所示,是本申请电子设备2中实时推荐系统20一实施例的程序模块图。本实施例中,所述的实时推荐系统20可以被分割成一个或多个程序模块,所述一个或者多个程序模块被存储于所述存储器21中,并由一个或多个处理器(本实施例中为所述处理器22)所执行,以完成本申请。例如,在图2中,所述的实时推荐系统20可以被分割成训练模块201、存储模块202、分配模块203、以及推荐模块204。本申请所称的程序模块是指能够完成特定功能的一系列计算机程序指令段,比程序更适合于描述所述实时推荐系统20在所述电子设备2中的执行过程。以下将就各程序模块201-204的功能进行详细描述。
所述训练模块201,用于通过预定编程语言(如Scala编程语言)创建模型类,利用该模型类在特定的计算引擎(如Spark大数据引擎)上离线训练模型,并将训练得到的模型托管至实时计算后台(如Java后台)。在本实施例中,利用Java语言与Scala语言互通的关系,将Scala编程语音创建的模型类训练 得到的模型用Java实时计算后台进行托管,从而实现实时预测和推荐,降低预测和推荐的时间。
优选地,在本实施例中,所述模型类继承一个统一的父类,所有模型的实现类通过统一的父类接口进行预测推荐,从而方便横向扩展,接入更多的推荐业务。
举例而言,所述模型类的父类可以定义为如下格式:
trait OpmAlgorithmBase extends Serializable{
def predict(input:HashMap[String,String]):Any
}
定义好该父类后,所有的模型实现类都继承这个父类,每个模型实现类都需要实现这个预测方法,实时计算后台(如Java后台)通过调用这个方法进行实时预测。
所述存储模块202,用于通过所述实时计算后台,将用户特征(如用户画像特征,可以是大数据量级)存储至高速缓存区(如Redis数据库)。在本实施例中,由于用户画像特征缓存到Redis数据库,可以大幅缩短用户画像特征查询时间,从而缩短预测时间(单条预测耗时平均2ms)。
所述分配模块203,用于通过所述实时计算后台,将用户标识(用户ID)映射成指定区间内的数字(如哈希成0-100之间的数字),并分配不同数字区间内的用户至不同的模型(即训练得到的模型)。例如,将用户标识哈希成0-100之间的数字后,分成第一数字区间“0-50”和第二数字区间“51-100”,并将第一数字区间内对应的用户分配至第一模型(如A模型),第二数字区间内对应的用户分配至第二模型(如B模型)。
所述推荐模块204,用于通过所述实时计算后台,调用不同的模型从所述 高速缓存区获取不同数字区间内的用户对应的用户特征,根据获取的用户特征进行用户行为预测(如预测用户是否购买产险),并将预测结果实时推荐至用户。
例如,调用第一模型从所述高速缓存区获取第一数字区间内的用户对应的用户特征,根据获取的用户特征针对第一数字区间内的用户进行用户行为预测,并将预测结果实时推荐至第一数字区间内的用户;调用第二模型从所述高速缓存区获取第二数字区间内的用户对应的用户特征,根据获取的用户特征针对第二数字区间内的用户进行用户行为预测,并将预测结果实时推荐至第二数字区间内的用户。
优选地,在其它实施例中,所述实时推荐系统20还用于:
根据不同模型的预测效果,动态调节不同数字区间的大小,以便灵活控制每个模型覆盖的预测用户流量,使得预测效果优的模型分配更多的用户。具体而言,在其它实施例中,若特定模型的预测效果高于预设阈值(如80%),则增大该特定模型所分配的用户数字区间(如增大20%)。例如,如上举例,若第一模型的预测效果为85%,则将第一模型所分配的第一数字区间“0-50”增大20%,得到调节后的第一数字区间“0-60”。相应地,第二模型所分配的第二数字区间“51-100”缩小20%,得到调节后的第二数字区间“61-100”。
更进一步地,在其它实施例中,若特定模型的预测效果高于第一预设阈值(如80%),则将该特定模型所分配的用户数字区间增大第一比例(如增大20%);若特定模型的预测效果高于第二预设阈值(如90%),则将该特定模型所分配的用户数字区间增大第二比例(如增大50%)。
优选地,在其它实施例中,所述实时推荐系统20还用于:
将不同模型的预测效果实时写入所述高速缓存区,并通过特定图表格式(参阅图4所示的曲线图)将不同模型的预测效果显示于设定的前端界面系 统(如应用程序界面系统),从而可以实时地对比不同模型的优劣,方便同时对多个模型进行线上管理(如A/B测试管理等)。
优选地,在其它实施例中,所述实时推荐系统20还用于:
将每个预测项目(或预测业务)的关键数据(如核心用户的信息等)依据时间顺序(如日/周/月/年)统计生成报表,并将生成的报表显示于设定的前端界面系统。
通过上述程序模块201-204,本申请所提出的实时推荐系统20,通过实时计算后台的运作缩短了预测时间,并利用前端界面系统实现了同时对多个模型进行线上管理。本申请完成了从模型训练到上线到效果评估的闭环,通过创建的模型类可以快速更新迭代模型,同时也可以方便地横向扩展,接入更多的推荐业务。
此外,本申请还提出一种实时推荐方法。
参阅图3所示,是本申请实时推荐方法一实施例的实施流程示意图。在本实施例中,根据不同的需求,图3所示的流程图中的步骤的执行顺序可以改变,某些步骤可以省略。
步骤S31,通过预定编程语言(如Scala编程语言)创建模型类,利用该模型类在特定的计算引擎(如Spark大数据引擎)上离线训练模型,并将训练得到的模型托管至实时计算后台(如Java后台)。在本实施例中,利用Java语言与Scala语言互通的关系,将Scala编程语音创建的模型类训练得到的模型用Java实时计算后台进行托管,从而实现实时预测和推荐,降低预测和推荐的时间。
优选地,在本实施例中,所述模型类继承一个统一的父类,所有模型的实现类通过统一的父类接口进行预测推荐,从而方便横向扩展,接入更多的 推荐业务。
举例而言,所述模型类的父类可以定义为如下格式:
trait OpmAlgorithmBase extends Serializable{
def predict(input:HashMap[String,String]):Any
}
定义好该父类后,所有的模型实现类都继承这个父类,每个模型实现类都需要实现这个预测方法,实时计算后台(如Java后台)通过调用这个方法进行实时预测。
步骤S32,通过所述实时计算后台,将用户特征(如用户画像特征,可以是大数据量级)存储至高速缓存区(如Redis数据库)。在本实施例中,由于用户画像特征缓存到Redis数据库,可以大幅缩短用户画像特征查询时间,从而缩短预测时间(单条预测耗时平均2ms)。
步骤S33,通过所述实时计算后台,将用户标识(用户ID)映射成指定区间内的数字(如哈希成0-100之间的数字),并分配不同数字区间内的用户至不同的模型(即训练得到的模型)。例如,将用户标识哈希成0-100之间的数字后,分成第一数字区间“0-50”和第二数字区间“51-100”,并将第一数字区间内对应的用户分配至第一模型(如A模型),第二数字区间内对应的用户分配至第二模型(如B模型)。
步骤S34,于所述实时计算后台,调用不同的模型从所述高速缓存区获取不同数字区间内的用户对应的用户特征,根据获取的用户特征进行用户行为预测(如预测用户是否购买产险),并将预测结果实时推荐至用户。
例如,调用第一模型从所述高速缓存区获取第一数字区间内的用户对应的用户特征,根据获取的用户特征针对第一数字区间内的用户进行用户行为 预测,并将预测结果实时推荐至第一数字区间内的用户;调用第二模型从所述高速缓存区获取第二数字区间内的用户对应的用户特征,根据获取的用户特征针对第二数字区间内的用户进行用户行为预测,并将预测结果实时推荐至第二数字区间内的用户。
优选地,在其它实施例中,所述实时推荐方法还包括如下步骤:
根据不同模型的预测效果,动态调节不同数字区间的大小,以便灵活控制每个模型覆盖的预测用户流量,使得预测效果优的模型分配更多的用户。具体而言,在其它实施例中,若特定模型的预测效果高于预设阈值(如80%),则增大该特定模型所分配的用户数字区间(如增大20%)。例如,如上举例,若第一模型的预测效果为85%,则将第一模型所分配的第一数字区间“0-50”增大20%,得到调节后的第一数字区间“0-60”。相应地,第二模型所分配的第二数字区间“51-100”缩小20%,得到调节后的第二数字区间“61-100”。
更进一步地,在其它实施例中,若特定模型的预测效果高于第一预设阈值(如80%),则将该特定模型所分配的用户数字区间增大第一比例(如增大20%);若特定模型的预测效果高于第二预设阈值(如90%),则将该特定模型所分配的用户数字区间增大第二比例(如增大50%)。
优选地,在其它实施例中,所述实时推荐方法还包括如下步骤:
将不同模型的预测效果实时写入所述高速缓存区,并通过特定图表格式(参阅图4所示的曲线图)将不同模型的预测效果显示于设定的前端界面系统(如应用程序界面系统),从而可以实时地对比不同模型的优劣,方便同时对多个模型进行线上管理(如A/B测试管理等)。
优选地,在其它实施例中,所述实时推荐方法还包括如下步骤:
将每个预测项目(或预测业务)的关键数据(如核心用户的信息等)依据时间顺序(如日/周/月/年)统计生成报表,并将生成的报表显示于设定的前 端界面系统。
通过上述步骤S31-S34及其它相关步骤,本申请所提出的实时推荐方法,通过实时计算后台的运作缩短了预测时间,并利用前端界面系统实现了同时对多个模型进行线上管理。本申请完成了从模型训练到上线到效果评估的闭环,通过创建的模型类可以快速更新迭代模型,同时也可以方便地横向扩展,接入更多的推荐业务。
进一步地,为实现上述目的,本申请还提供一种计算机可读存储介质(如ROM/RAM、磁碟、光盘),所述计算机可读存储介质存储有实时推荐系统20,所述实时推荐系统20可被至少一个处理器22执行,以使所述至少一个处理器22执行如上所述的实时推荐方法的步骤。
通过以上的实施方式的描述,本领域的技术人员可以清楚地了解到上述实施例方法可借助软件加必需的通用硬件平台的方式来实现,当然也可以通过硬件来实现,但很多情况下前者是更佳的实施方式。基于这样的理解,本申请的技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来,该计算机软件产品存储在一个存储介质(如ROM/RAM、磁碟、光盘)中,包括若干指令用以使得一台终端设备(可以是手机,计算机,服务器,空调器,或者网络设备等)执行本申请各个实施例所述的方法。
以上参照附图说明了本申请的优选实施例,并非因此局限本申请的权利范围。上述本申请实施例序号仅仅为了描述,不代表实施例的优劣。另外,虽然在流程图中示出了逻辑顺序,但是在某些情况下,可以以不同于此处的顺序执行所示出或描述的步骤。
本领域技术人员不脱离本申请的范围和实质,可以有多种变型方案实现本申请,比如作为一个实施例的特征可用于另一实施例而得到又一实施例。 凡是利用本申请说明书及附图内容所作的等效结构或等效流程变换,或直接或间接运用在其他相关的技术领域,均同理包括在本申请的专利保护范围内。

Claims (20)

  1. 一种电子设备,其特征在于,所述电子设备包括存储器和处理器,所述存储器上存储有可在所述处理器上运行的实时推荐系统,所述实时推荐系统被所述处理器执行时实现如下步骤:
    通过预定编程语言创建模型类,利用该模型类在特定的计算引擎上离线训练模型,并将训练得到的模型托管至实时计算后台;
    通过所述实时计算后台,将用户特征存储至高速缓存区;
    通过所述实时计算后台,将用户标识映射成指定区间内的数字,并分配不同数字区间内的用户至不同的模型;及
    通过所述实时计算后台,调用不同的模型从所述高速缓存区获取不同数字区间内的用户对应的用户特征,根据获取的用户特征进行用户行为预测,并将预测结果实时推荐至用户。
  2. 如权利要求1所述的电子设备,其特征在于,所述实时推荐系统被所述处理器执行时还用于实现如下步骤:
    根据不同模型的预测效果,动态调节不同数字区间的大小;及
    若特定模型的预测效果高于预设阈值,则增大该特定模型所分配的用户数字区间。
  3. 如权利要求2所述的电子设备,其特征在于,所述实时推荐系统被所述处理器执行时还用于实现如下步骤:
    若特定模型的预测效果高于第一预设阈值,则将该特定模型所分配的用户数字区间增大第一比例;及
    若特定模型的预测效果高于第二预设阈值,则将该特定模型所分配的用户数字区间增大第二比例。
  4. 如权利要求2所述的电子设备,其特征在于,所述模型类继承一个统一的父类,所有模型的实现类通过统一的父类接口进行预测推荐;及
    所述不同数字区间包括第一数字区间和第二数字区间,并将第一数字区间内对应的用户分配至第一模型,第二数字区间内对应的用户分配至第二模型。
  5. 如权利要求3所述的电子设备,其特征在于,所述模型类继承一个统一的父类,所有模型的实现类通过统一的父类接口进行预测推荐;及
    所述不同数字区间包括第一数字区间和第二数字区间,并将第一数字区间内对应的用户分配至第一模型,第二数字区间内对应的用户分配至第二模型。
  6. 如权利要求2所述的电子设备,其特征在于,所述实时推荐系统被所述处理器执行时还用于实现如下步骤:
    将不同模型的预测效果实时写入所述高速缓存区,并通过特定图表格式将不同模型的预测效果显示于设定的前端界面系统;及
    将每个预测项目的关键数据依据时间顺序统计生成报表,并将生成的报表显示于设定的前端界面系统。
  7. 如权利要求3所述的电子设备,其特征在于,所述实时推荐系统被所述处理器执行时还用于实现如下步骤:
    将不同模型的预测效果实时写入所述高速缓存区,并通过特定图表格式将不同模型的预测效果显示于设定的前端界面系统;及
    将每个预测项目的关键数据依据时间顺序统计生成报表,并将生成的报表显示于设定的前端界面系统。
  8. 一种实时推荐方法,应用于电子设备,其特征在于,所述方法包括:
    通过预定编程语言创建模型类,利用该模型类在特定的计算引擎上离线训练模型,并将训练得到的模型托管至实时计算后台;
    通过所述实时计算后台,将用户特征存储至高速缓存区;
    通过所述实时计算后台,将用户标识映射成指定区间内的数字,并分配不同数字区间内的用户至不同的模型;及
    通过所述实时计算后台,调用不同的模型从所述高速缓存区获取不同数字区间内的用户对应的用户特征,根据获取的用户特征进行用户行为预测,并将预测结果实时推荐至用户。
  9. 如权利要求8所述的实时推荐方法,其特征在于,该方法还包括:
    根据不同模型的预测效果,动态调节不同数字区间的大小;及
    若特定模型的预测效果高于预设阈值,则增大该特定模型所分配的用户数字区间。
  10. 如权利要求9所述的实时推荐方法,其特征在于,该方法还包括:
    若特定模型的预测效果高于第一预设阈值,则将该特定模型所分配的用户数字区间增大第一比例;及
    若特定模型的预测效果高于第二预设阈值,则将该特定模型所分配的用户数字区间增大第二比例。
  11. 如权利要求9所述的实时推荐方法,其特征在于,所述模型类继承一个统一的父类,所有模型的实现类通过统一的父类接口进行预测推荐;及
    所述不同数字区间包括第一数字区间和第二数字区间,并将第一数字区间内对应的用户分配至第一模型,第二数字区间内对应的用户分配至第二模型。
  12. 如权利要求10所述的实时推荐方法,其特征在于,所述模型类继承一个统一的父类,所有模型的实现类通过统一的父类接口进行预测推荐;及
    所述不同数字区间包括第一数字区间和第二数字区间,并将第一数字区间内对应的用户分配至第一模型,第二数字区间内对应的用户分配至第二模型。
  13. 如权利要求9所述的实时推荐方法,其特征在于,该方法还包括:
    将不同模型的预测效果实时写入所述高速缓存区,并通过特定图表格式将不同模型的预测效果显示于设定的前端界面系统;及
    将每个预测项目的关键数据依据时间顺序统计生成报表,并将生成的报表显示于设定的前端界面系统。
  14. 如权利要求10所述的实时推荐方法,其特征在于,该方法还包括:
    将不同模型的预测效果实时写入所述高速缓存区,并通过特定图表格式将不同模型的预测效果显示于设定的前端界面系统;及
    将每个预测项目的关键数据依据时间顺序统计生成报表,并将生成的报表显示于设定的前端界面系统。
  15. 一种计算机可读存储介质,所述计算机可读存储介质存储有实时推荐系统,所述实时推荐系统可被至少一个处理器执行,所述实时推荐系统被所述处理器执行时实现如下步骤:
    通过预定编程语言创建模型类,利用该模型类在特定的计算引擎上离线训练模型,并将训练得到的模型托管至实时计算后台;
    通过所述实时计算后台,将用户特征存储至高速缓存区;
    通过所述实时计算后台,将用户标识映射成指定区间内的数字,并分配 不同数字区间内的用户至不同的模型;及
    通过所述实时计算后台,调用不同的模型从所述高速缓存区获取不同数字区间内的用户对应的用户特征,根据获取的用户特征进行用户行为预测,并将预测结果实时推荐至用户。
  16. 如权利要求15所述的计算机可读存储介质,其特征在于,所述实时推荐系统被所述处理器执行时还用于实现如下步骤:
    根据不同模型的预测效果,动态调节不同数字区间的大小;及
    若特定模型的预测效果高于预设阈值,则增大该特定模型所分配的用户数字区间。
  17. 如权利要求16所述的计算机可读存储介质,其特征在于,所述实时推荐系统被所述处理器执行时还用于实现如下步骤:
    若特定模型的预测效果高于第一预设阈值,则将该特定模型所分配的用户数字区间增大第一比例;及
    若特定模型的预测效果高于第二预设阈值,则将该特定模型所分配的用户数字区间增大第二比例。
  18. 如权利要求16所述的计算机可读存储介质,其特征在于,所述模型类继承一个统一的父类,所有模型的实现类通过统一的父类接口进行预测推荐;及
    所述不同数字区间包括第一数字区间和第二数字区间,并将第一数字区间内对应的用户分配至第一模型,第二数字区间内对应的用户分配至第二模型。
  19. 如权利要求17所述的计算机可读存储介质,其特征在于,所述模型 类继承一个统一的父类,所有模型的实现类通过统一的父类接口进行预测推荐;及
    所述不同数字区间包括第一数字区间和第二数字区间,并将第一数字区间内对应的用户分配至第一模型,第二数字区间内对应的用户分配至第二模型。
  20. 如权利要求16或17所述的计算机可读存储介质,其特征在于,所述实时推荐系统被所述处理器执行时还用于实现如下步骤:
    将不同模型的预测效果实时写入所述高速缓存区,并通过特定图表格式将不同模型的预测效果显示于设定的前端界面系统;及
    将每个预测项目的关键数据依据时间顺序统计生成报表,并将生成的报表显示于设定的前端界面系统。
PCT/CN2018/076173 2017-10-13 2018-02-10 实时推荐方法、电子设备及计算机可读存储介质 WO2019071897A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201710953455.9 2017-10-13
CN201710953455.9A CN107807967B (zh) 2017-10-13 2017-10-13 实时推荐方法、电子设备及计算机可读存储介质

Publications (1)

Publication Number Publication Date
WO2019071897A1 true WO2019071897A1 (zh) 2019-04-18

Family

ID=61584929

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/076173 WO2019071897A1 (zh) 2017-10-13 2018-02-10 实时推荐方法、电子设备及计算机可读存储介质

Country Status (2)

Country Link
CN (1) CN107807967B (zh)
WO (1) WO2019071897A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113377018A (zh) * 2020-03-09 2021-09-10 青岛海尔空调器有限总公司 空调器的智能控制方法与空调器的智能控制设备

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108829415B (zh) * 2018-05-29 2023-07-21 努比亚技术有限公司 模型加载方法、服务器及计算机可读存储介质
CN109150983B (zh) * 2018-07-27 2022-02-25 腾讯科技(深圳)有限公司 前端显示的控制方法及装置、数据推荐的控制方法及装置
CN110874737A (zh) * 2018-09-03 2020-03-10 北京京东金融科技控股有限公司 支付方式推荐方法、装置、电子设备、存储介质
CN109829063B (zh) * 2019-01-29 2020-11-27 腾讯科技(深圳)有限公司 一种数据处理方法、装置及存储介质
CN111368314A (zh) * 2020-02-28 2020-07-03 深圳前海微众银行股份有限公司 基于交叉特征的建模、预测方法、装置、设备及存储介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150127419A1 (en) * 2013-11-04 2015-05-07 Oracle International Corporation Item-to-item similarity generation
CN105469263A (zh) * 2014-09-24 2016-04-06 阿里巴巴集团控股有限公司 一种商品推荐方法及装置
CN106202474A (zh) * 2016-07-14 2016-12-07 Tcl集团股份有限公司 一种对象推荐方法和装置
CN106485562A (zh) * 2015-09-01 2017-03-08 苏宁云商集团股份有限公司 一种基于用户历史行为的商品信息推荐方法及系统
CN107122990A (zh) * 2017-03-22 2017-09-01 广州优视网络科技有限公司 应用推荐方法、客户端、服务器及系统

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2012159308A1 (zh) * 2011-06-29 2012-11-29 华为技术有限公司 一种业务交叉时的项目推荐方法及系统
CN105095219B (zh) * 2014-04-23 2019-02-01 华为技术有限公司 微博推荐方法和终端
CN106126641B (zh) * 2016-06-24 2019-02-05 中国科学技术大学 一种基于Spark的实时推荐系统及方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150127419A1 (en) * 2013-11-04 2015-05-07 Oracle International Corporation Item-to-item similarity generation
CN105469263A (zh) * 2014-09-24 2016-04-06 阿里巴巴集团控股有限公司 一种商品推荐方法及装置
CN106485562A (zh) * 2015-09-01 2017-03-08 苏宁云商集团股份有限公司 一种基于用户历史行为的商品信息推荐方法及系统
CN106202474A (zh) * 2016-07-14 2016-12-07 Tcl集团股份有限公司 一种对象推荐方法和装置
CN107122990A (zh) * 2017-03-22 2017-09-01 广州优视网络科技有限公司 应用推荐方法、客户端、服务器及系统

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113377018A (zh) * 2020-03-09 2021-09-10 青岛海尔空调器有限总公司 空调器的智能控制方法与空调器的智能控制设备

Also Published As

Publication number Publication date
CN107807967A (zh) 2018-03-16
CN107807967B (zh) 2021-10-22

Similar Documents

Publication Publication Date Title
WO2019071897A1 (zh) 实时推荐方法、电子设备及计算机可读存储介质
CN108287708B (zh) 一种数据处理方法、装置、服务器及计算机可读存储介质
CN110826071A (zh) 软件漏洞风险预测方法、装置、设备及存储介质
WO2019062189A1 (zh) 电子装置、数据表归档处理的方法、系统及存储介质
WO2019061990A1 (zh) 用户意图预测方法、电子设备及计算机可读存储介质
WO2019024060A1 (zh) 数据存储方法、装置和存储介质
WO2018223941A1 (zh) 多保险产品出单的装置、方法及计算机可读存储介质
CN108449313B (zh) 电子装置、互联网服务系统风险预警方法及存储介质
CN113076104A (zh) 页面生成方法、装置、设备及存储介质
WO2020192141A1 (zh) 接口模拟方法、装置、计算机设备和存储介质
US20170171311A1 (en) System and Method for Preemptive Request Processing
CN112506486A (zh) 搜索系统建立方法、装置、电子设备及可读存储介质
CN115936895A (zh) 基于人工智能的风险评估方法、装置、设备及存储介质
US10331894B2 (en) Static program analysis in an object-relational mapping framework
CN110674426B (zh) 网页行为上报方法和装置
CN111159183B (zh) 报表生成方法、电子装置及计算机可读存储介质
CN116661936A (zh) 页面数据的处理方法、装置、计算机设备及存储介质
CN116450723A (zh) 数据提取方法、装置、计算机设备及存储介质
CN108241732B (zh) 电子装置、信息处理的方法及存储介质
US8868485B2 (en) Data flow cost modeling
CN113722573B (zh) 生成网络安全威胁数据集合的方法、系统和存储介质
CN114281586A (zh) 故障确定方法及装置、电子设备和计算机可读存储介质
CN114510908A (zh) 数据导出方法、装置、计算机设备及存储介质
WO2019080503A1 (zh) 薪资中间指标的计算方法、应用服务器及计算机可存储介质
CN111626438B (zh) 基于模型迁移的用户策略分配方法、装置及电子设备

Legal Events

Date Code Title Description
NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 08/10/2020)

122 Ep: pct application non-entry in european phase

Ref document number: 18866086

Country of ref document: EP

Kind code of ref document: A1