CN111915383A - Window-based article cold start recommendation method and device - Google Patents
Window-based article cold start recommendation method and device Download PDFInfo
- Publication number
- CN111915383A CN111915383A CN201910383441.7A CN201910383441A CN111915383A CN 111915383 A CN111915383 A CN 111915383A CN 201910383441 A CN201910383441 A CN 201910383441A CN 111915383 A CN111915383 A CN 111915383A
- Authority
- CN
- China
- Prior art keywords
- window
- item
- cold start
- recommendation
- behavior data
- Prior art date
- Legal status (The legal status 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 status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0251—Targeted advertisements
- G06Q30/0255—Targeted advertisements based on user history
- G06Q30/0256—User search
Landscapes
- Business, Economics & Management (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Strategic Management (AREA)
- Engineering & Computer Science (AREA)
- Development Economics (AREA)
- Marketing (AREA)
- Economics (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
A cold start recommendation method and device for articles based on windows are provided. The algorithm may include: recommending the newly generated article according to the exploration flow, and acquiring user behavior data corresponding to the newly generated article; and acquiring a new recommendation list by using an article cold start recommendation algorithm based on the user behavior data corresponding to the newly generated article and other articles recommended previously in the window, so as to recommend the articles in the new recommendation list at a utilization flow rate larger than the exploration flow rate.
Description
Technical Field
The present disclosure generally relates to the field of intelligent recommendation technologies, and in particular, to a window-based article cold start recommendation method and apparatus.
Background
At present, the implementation process of the explicit & explicit scheme in terms of cold start of the article is as follows: an exploration (explore) process is performed with a small portion of traffic to find premium content, and then an exploitation (explore) process is performed with respect to the continuously found premium content using the remaining large amount of traffic and a recommendation list is acquired, thereby recommending the premium content in the recommendation list to the user.
However, the existing ex core & ex oil scheme for cold start of articles cannot quickly capture the recently hot articles or quickly eliminate the previously well-behaved and now poorly-behaved articles.
Disclosure of Invention
In order to solve the above problems, the present disclosure provides a method and an apparatus for recommending cold start of an article based on a window.
According to the present disclosure, there is provided a window-based item cold start recommendation method, the algorithm may include: recommending the newly generated article according to the exploration flow, and acquiring user behavior data corresponding to the newly generated article; and acquiring a new recommendation list by using an article cold start recommendation algorithm based on the user behavior data corresponding to the newly generated article and the recommended other articles in the window, so as to recommend the articles in the new recommendation list at a utilization flow rate larger than the exploration flow rate.
According to one embodiment of the present disclosure, the window may comprise a time window or a user behavior data total window.
According to one embodiment of the disclosure, the size of the time window may depend on the recent recommended activity level.
According to one embodiment of the present disclosure, the size of the time window may be inversely proportional to the sum of the number of displays generated per day when recommendations are made for an item and the number of user actions.
According to one embodiment of the present disclosure, the item cold start recommendation algorithm may include an Epsilon-Greedy algorithm, a Thompson sampling algorithm, or a UCB algorithm.
According to one embodiment of the present disclosure, the item may include at least one of textual content, video, pictures, and audio.
According to one embodiment of the disclosure, the user behavior data may relate to at least one of clicking, forwarding, sharing, commenting, collecting, agreeing to, and stepping on a recommended item.
According to one embodiment of the present disclosure, the exploration traffic may be proportional to the number of newly generated items and/or the exploration traffic may not exceed one tenth of the total traffic including the exploration traffic and the utilization traffic.
According to one embodiment of the present disclosure, obtaining a new recommendation list using an item cold start recommendation algorithm may include: based on the user behavior data corresponding to the newly generated item and other recommended items within the window, a score is calculated by an item cold start recommendation algorithm that measures how well each item is accepted to obtain a recommendation list consisting of a certain number of items with scores ranked top.
According to the present disclosure, there is provided a window-based item cold start recommendation device, which may include: the exploration unit is used for recommending the newly generated articles according to exploration flow and acquiring user behavior data corresponding to the newly generated articles; and the utilization unit acquires a new recommendation list by utilizing an article cold start recommendation algorithm based on the user behavior data corresponding to the newly generated article and the recommended other articles in the window, so that the articles in the new recommendation list are recommended at a utilization flow rate larger than the exploration flow rate.
According to one embodiment of the present disclosure, the window may comprise a time window or a user behavior data total window.
According to one embodiment of the disclosure, the size of the time window may depend on the recent recommended activity level.
According to one embodiment of the present disclosure, the size of the time window may be inversely proportional to the sum of the number of displays generated per day when recommendations are made for an item and the number of user actions.
According to one embodiment of the present disclosure, the item cold start recommendation algorithm may include an Epsilon-Greedy algorithm, a Thompson sampling algorithm, or a UCB algorithm.
According to one embodiment of the present disclosure, the item may include at least one of textual content, video, pictures, and audio.
According to one embodiment of the disclosure, the user behavior data may relate to at least one of clicking, forwarding, sharing, commenting, collecting, agreeing to, and stepping on a recommended item.
According to one embodiment of the present disclosure, the exploration traffic may be proportional to the number of newly generated items and/or the exploration traffic may not exceed one tenth of the total traffic including the exploration traffic and the utilization traffic.
According to one embodiment of the present disclosure, obtaining a new recommendation list using an item cold start recommendation algorithm may include: based on the user behavior data corresponding to the newly generated item and other recommended items within the window, a score is calculated by an item cold start recommendation algorithm that measures how well each item is accepted to obtain a recommendation list consisting of a certain number of items with scores ranked top.
According to the present disclosure, there is provided a system comprising at least one computing device and at least one storage device storing instructions, wherein the instructions, when executed by the at least one computing device, cause the at least one computing device to perform the window-based item cold start recommendation method of any of the preceding embodiments.
According to the present disclosure, there is provided a computer-readable storage medium storing instructions that, when executed by at least one computing device, cause the at least one computing device to perform the window-based item cold start recommendation method of any of the preceding embodiments.
By adopting the method and the device, the recently hot articles can be quickly captured, and the articles which are good in performance before and poor in performance now can be quickly eliminated, so that the ecology of a recommendation system is optimized.
Drawings
These and/or other aspects and advantages of the present disclosure will become more apparent and more readily appreciated from the following detailed description of the embodiments of the present disclosure, taken in conjunction with the accompanying drawings of which:
FIG. 1 is a flow chart of a window-based item cold start recommendation method according to an embodiment of the present disclosure;
FIG. 2 is a flow chart of a window-based item cold start recommendation method according to another embodiment of the present disclosure;
FIG. 3 shows a block diagram of a window-based item cold start recommendation device, according to an embodiment of the present disclosure;
FIG. 4 illustrates a block diagram of a system including at least one computing device and at least one storage device storing instructions, according to an embodiment of the disclosure.
Detailed Description
As required, specific embodiments of the present disclosure are disclosed herein; however, it is to be understood that the disclosed embodiments are merely exemplary of the disclosure that may be embodied in various and alternative forms. The figures are not necessarily to scale; some features may be exaggerated or minimized to show details of particular components. Therefore, specific structural and functional details disclosed herein are not to be interpreted as limiting, but merely as a representative basis for teaching one skilled in the art to variously employ the present disclosure.
The recommendation system needs to predict and recommend items that may be of interest to a user in the face of a situation where new items are continuously generated (e.g., a large amount of information, video, songs, etc. are generated every day on the internet), and how to recommend new items to a user who may be of interest to the user is a common cold start problem of items that we often say, where the items are also commonly referred to as materials or content and include, but are not limited to, text content (e.g., blog articles, news information, forum posts), video, audio, pictures, etc.
A window-based item cold start recommendation method and apparatus according to embodiments of the present disclosure is described below with reference to the accompanying drawings.
FIG. 1 is a flow chart of a window-based item cold start recommendation method according to an embodiment of the present disclosure.
As shown in fig. 1, in step S101, newly generated items are recommended by exploring traffic. In one embodiment, the newly generated items may include at least one of textual content (e.g., blog articles, news information, forum posts, etc.), video (e.g., funny videos, fantasy, tv shows, etc.), pictures, and audio (e.g., music, lectures, radio shows, etc.), and the exploration traffic may be proportional to the number of newly generated items. For example, assuming that the total flow rate is 5000 ten thousand, the exploration flow rate may be 100 ten thousand if the number of newly produced items per day is 1 ten thousand, and may be 200 ten thousand if the number of newly produced items per day is 2 ten thousand.
In step S102, user behavior data corresponding to the newly generated item may be acquired. The user behavior data may relate to at least one of clicking, forwarding, sharing, commenting, collecting, likes, and clicks on recommended items. These operations may be considered as user acceptance or rejection, i.e., whether or not they are interested, of the recommended item representation. For example, in one example, the user's action of clicking on a newly generated video may be used to indicate the user's acceptance of the video. In another example, the user's act of forwarding newly released music may be used to indicate the user's acceptance of the music.
In step S103, a new recommendation list may be obtained using an item cold start recommendation algorithm based on user behavior data corresponding to the newly generated item and the recommended other items within the time window.
Here, a time window for applying the article cold start algorithm may be set, and the size of the time window may be previously set based on the article generation speed, and may be then dynamically adjusted based on a change in the article generation speed, as an example; alternatively, here, the size of the time window may be dynamically set based directly on changes in the rate at which the articles are produced. Specifically, the size of the time window may depend on the recent recommendation activity, and by way of example, the size of the time window may be inversely proportional to the sum of the number of impressions generated when recommending items per day and the number of user actions, thereby increasing confidence to more quickly mine new popular items or eliminate old items that are no longer popular. That is, if the sum of the number of displays and the number of user actions generated when a recommendation is made for an item on each day is larger, the size of the time window is smaller. For example, the time window may be sized according to the total number of impressions and clicks generated on a line per day, and if the total number of impressions and clicks generated on a per day basis is relatively small, the time window may be set to a relatively large time window for better confidence of the result, e.g., three days in the case of impressions on the order of tens of millions per day and clicks on the order of millions per day.
For a set time window, user acceptance of items falling therein, i.e., new items recommended by the exploration traffic and other recommended items (e.g., items recommended by the utilization traffic) within the time window may be considered, and accordingly, a new recommendation list may be obtained among the items using an item cold-start recommendation algorithm.
Additionally, the item cold start recommendation algorithm may include, but is not limited to, the Epsilon-Greedy algorithm, the Thompson sampling algorithm, the ucb (upper Confidence bound) algorithm, and the like. And, obtaining a new recommendation list using the item cold start recommendation algorithm may include: based on the user behavior data corresponding to the newly generated item and other items that were previously recommended within the window, a score is calculated using an item cold start recommendation algorithm that measures how well each item is accepted to obtain a recommendation list consisting of a particular number of items with scores ranked top. In one example, a score measuring how well each item is accepted may be calculated using a Thompson sampling algorithm based on the last three days of user behavior data corresponding to newly generated items and other items that were previously recommended (involving at least one of clicking, forwarding, sharing, commenting, favoring, and clicking on recommended items) to obtain a recommendation list consisting of five hundred items with scores ranked top. In another example, a score measuring how well each item was accepted may be calculated using the UCB algorithm based on the last five days of user behavior data corresponding to newly generated items and other items that were previously recommended (involving at least one of clicking, forwarding, sharing, commenting, favoring, praising, and clicking on the recommended item) to obtain a recommendation list consisting of 1000 items with scores ranked top.
Next, in step S104, the items in the new recommendation list are recommended at a usage traffic rate larger than the search traffic rate. The allocation between exploration traffic and utilization traffic needs to be balanced on a case-by-case basis. On one hand, the more the number of newly produced articles is, the more exploration flow is needed to explore the newly produced articles; on the other hand, too much exploration traffic may affect the user experience. In particular embodiments, the exploration traffic is proportional to the number of newly generated items, and/or the exploration traffic does not exceed one tenth of the total traffic including the exploration traffic and the utilization traffic. For example, if the total flow rate is 5000 ten thousand, the probe flow rate does not exceed 500 ten thousand. In addition, the ratio of the search traffic to the utilization traffic may be adjusted as necessary, and the above example is only an example of the present disclosure, and the present disclosure is not limited to the above example. When the recommendation is performed by using the flow, a machine learning model trained in advance can be used for predicting specific users and the articles in the recommendation list so as to individually screen the articles which are more suitable for the recommendation of the specific users.
Compared with the prior art, the method and the device introduce the concept of acquiring the window of the user behavior data based on the recommendation list aiming at the article cold start recommendation algorithm, so that the recently hot articles can be quickly captured, and the articles which are good in performance before and poor in performance now can be quickly eliminated to optimize the ecology of the recommendation system. Although in the above embodiments, the window on which the cold-start recommendation algorithm for an item obtains the recommendation list is a time window, any concept that may define an interval (or window) of user behavior data on which the cold-start recommendation algorithm for an item obtains the recommendation list is included within the scope of the present disclosure. An embodiment in which the window on which the cold start recommendation algorithm for an item obtains the recommendation list is a user behavior data amount window will be described below with reference to fig. 2.
FIG. 2 is a flow chart of a window-based item cold start recommendation method according to another embodiment of the present disclosure.
As shown in fig. 2, in step S201, the newly generated item is recommended by exploring the traffic. In one embodiment, the newly generated items may include at least one of textual content (e.g., blog articles, news information, forum posts, etc.), video (e.g., funny videos, fantasy, tv shows, etc.), pictures, and audio (e.g., music, lectures, radio shows, etc.), and the exploration traffic may be proportional to the number of newly generated items. For example, assuming that the total flow rate is 5000 ten thousand, the exploration flow rate may be 100 ten thousand if the number of newly produced items per day is 1 ten thousand, and may be 200 ten thousand if the number of newly produced items per day is 2 ten thousand.
In step S202, user behavior data corresponding to the newly generated item may be acquired. The user behavior data may relate to at least one of clicking, forwarding, sharing, commenting, collecting, likes, and clicks on recommended items.
In step S203, a new recommendation list may be obtained by using an item cold start recommendation algorithm based on the user behavior data corresponding to the newly generated item and other items recommended previously within the user behavior data total window. Here, the user behavior data total amount window may include a predetermined number of user behavior data that have been recently acquired.
Here, a user behavior data amount window for applying the article cold start algorithm may be set, and the size of the user behavior data amount window may be previously set based on the article generation speed, and may be then dynamically adjusted based on a change in the article generation speed, as an example; alternatively, here, the size of the user behavior data amount window may be dynamically set directly based on a change in the item generation speed. Specifically, the size of the user behavior data total window may depend on the recent recommendation activity, and as an example, the size of the user behavior data total window may be inversely proportional to the sum of the number of displays generated when recommending items on each day and the number of user behaviors, so as to increase the confidence level, so as to more quickly mine new popular items or eliminate old items that are no longer popular. That is, if the sum of the number of displays and the number of user actions generated when a recommendation is made for an item on each day is larger, the size of the user action data total amount window is smaller. For example, the size of the total user behavior data window may be determined based on the total number of impressions and clicks generated on-line per day, and if the total number of impressions and clicks generated per day is relatively small, the total user behavior data window may be set to be a relatively large total user behavior data window in order to make the result more reliable, e.g., in the case of impressions on the order of tens of millions per day and clicks on the order of millions per day, the time window may be set to include the last 3000 tens of thousands of user behavior data.
For the set total user behavior data window, the acceptance of the items falling therein may be considered, that is, the acceptance of the user of the new items recommended by the traffic exploration and other recommended items (for example, the items recommended by the traffic exploration) in the total user behavior data window, and accordingly, a new recommendation list may be obtained from the items by using an item cold-start recommendation algorithm.
Additionally, the item cold start recommendation algorithm may include, but is not limited to, the Epsilon-Greedy algorithm, the Thompson sampling algorithm, the ucb (upper Confidence bound) algorithm, and the like. And, obtaining a new recommendation list using the item cold start recommendation algorithm may include: based on the user behavior data corresponding to the newly generated item and other items that were previously recommended within the window, a score is calculated using an item cold start recommendation algorithm that measures how well each item is accepted to obtain a recommendation list consisting of a particular number of items with scores ranked top. In one example, a score measuring how well each item is accepted may be calculated using a Thompson sampling algorithm based on the last 3000 tens of thousands of user behavior data corresponding to newly generated items and other items that were previously recommended (involving at least one of clicking, forwarding, sharing, commenting, favoring, praising, and clicking on the recommended item) to obtain a recommendation list consisting of 500 items with the top scores. In another example, a score measuring how well each item is accepted may be calculated using the UCB algorithm based on the last 5000 million pieces of user behavior data (relating to at least one of clicking, forwarding, sharing, commenting, favoring, praising, and clicking on the recommended item) corresponding to newly produced items as well as other items that were previously recommended to obtain a recommendation list consisting of 1000 items with the scores ranked top.
Next, in step S204, the items in the new recommendation list are recommended at a usage traffic rate larger than the search traffic rate. The allocation between exploration traffic and utilization traffic needs to be balanced on a case-by-case basis. On one hand, the more the number of newly produced articles is, the more exploration flow is needed to explore the newly produced articles; on the other hand, too much exploration traffic may affect the user experience. In particular embodiments, the exploration traffic is proportional to the number of newly generated items, and/or the exploration traffic does not exceed one tenth of the total traffic including the exploration traffic and the utilization traffic. For example, if the total flow rate is 5000 ten thousand, the probe flow rate does not exceed 500 ten thousand. In addition, the ratio of the search traffic to the utilization traffic may be adjusted as necessary, and the above example is only an example of the present disclosure, and the present disclosure is not limited to the above example. When the recommendation is performed by using the flow, a machine learning model trained in advance can be used for predicting specific users and the articles in the recommendation list so as to individually screen the articles which are more suitable for the recommendation of the specific users.
Fig. 3 shows a block diagram of a window-based item cold start recommendation device 300 according to an embodiment of the present disclosure.
As shown in fig. 3, a window-based item cold start recommendation device 300 according to an embodiment of the present disclosure may include an exploration unit 301 and a utilization unit 302.
The exploration unit 301 may recommend the newly generated item at an exploration traffic and obtain user behavior data corresponding to the newly generated item. The newly generated item may include at least one of textual content, video, pictures, and audio. The user behavior data may relate to at least one of clicking, forwarding, sharing, commenting, collecting, likes, and clicks on recommended items.
The utilization unit 302 may obtain a new recommendation list by using an item cold start recommendation algorithm based on the user behavior data corresponding to the newly generated item and other items that have been recommended previously in the window, so as to recommend the items in the new recommendation list at a utilization flow rate greater than the exploration flow rate. In one example, the window may be a time window (e.g., the last three days) and the size of the time window may depend on the recent recommendation activity, in particular, the size of the time window may be inversely proportional to the sum of the number of displays generated when recommendations are made for items on each day and the number of user activities. In another example, the window may be a total user behavior data window (e.g., the last 3000 ten thousand pieces of user behavior data), and the size of the total user behavior data window may depend on the recent recommendation activity level, and in particular, the size of the total user behavior data window may be inversely proportional to the sum of the number of displays generated per day when a recommendation is made for an item and the number of user behaviors. The item cold start recommendation algorithm may include an Epsilon-Greedy algorithm, a Thompson sampling algorithm, or a UCB algorithm.
Additionally, according to one embodiment of the present disclosure, the exploration traffic may be proportional to the number of newly generated items and/or the exploration traffic may not exceed one tenth of the total traffic including the exploration traffic and the utilization traffic. Obtaining a new recommendation list using an item cold start recommendation algorithm may include: based on the user behavior data corresponding to the newly generated item and other items that were previously recommended within the window, a score is calculated using an item cold start recommendation algorithm that measures how well each item is accepted to obtain a recommendation list consisting of a particular number of items with scores ranked top.
The units included in the window-based item cold start recommendation device according to an exemplary embodiment of the present invention may be respectively configured as software, hardware, firmware or any combination thereof for performing specific functions. These means may correspond, for example, to a dedicated integrated circuit, to pure software code, or to a module combining software and hardware. Further, one or more functions implemented by these apparatuses may also be collectively performed by components in a physical entity device (e.g., a processor, a client, a server, or the like). The specific operations shown above in conjunction with fig. 1 and fig. 2 may be respectively performed by corresponding units in the apparatus shown in fig. 3, and details of the specific operations will not be described herein.
FIG. 4 illustrates a block diagram of a system including at least one computing device and at least one storage device storing instructions, according to an embodiment of the disclosure.
As shown in fig. 4, a system 400 provided according to an embodiment of the present disclosure may include at least one computing device (e.g., a processor) 401 and at least one storage device 402 storing instructions that, when executed by the at least one computing device 401, cause the at least one computing device 401 to perform the window-based item cold-start recommendation method of any of the preceding embodiments.
The computing devices may be deployed in servers or clients, as well as on node devices in a distributed network environment. Further, the computing device may be a PC computer, tablet device, personal digital assistant, smart phone, web application, or other device capable of executing the set of instructions described above. The computing device need not be a single computing device, but can be any device or collection of circuits capable of executing the instructions (or sets of instructions) described above, individually or in combination. The computing device may also be part of an integrated control system or system manager, or may be configured as a portable electronic device that interfaces with local or remote (e.g., via wireless transmission). In the computing device, the processor may include a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), a programmable logic device, a special purpose processor system, a microcontroller, or a microprocessor. By way of example, and not limitation, processors may also include analog processors, digital processors, microprocessors, multi-core processors, processor arrays, network processors, and the like.
Some of the operations described in the window-based item cold start recommendation method according to the exemplary embodiments of the present invention may be implemented by software, some of the operations may be implemented by hardware, and further, the operations may be implemented by a combination of hardware and software. The processor may execute instructions or code stored in one of the memory components, which may also store data. Instructions and data may also be transmitted and received over a network via a network interface device, which may employ any known transmission protocol. The memory component may be integral to the processor, e.g., having RAM or flash memory disposed within an integrated circuit microprocessor or the like. Further, the storage component may comprise a stand-alone device, such as an external disk drive, storage array, or any other storage device usable by a database system. The storage component and the processor may be operatively coupled or may communicate with each other, such as through an I/O port, a network connection, etc., so that the processor can read files stored in the storage component. Further, the computing device may also include a video display (such as a liquid crystal display) and a user interaction interface (such as a keyboard, mouse, touch input device, etc.). All components of the computing device may be connected to each other via a bus and/or a network.
Operations involved in a window-based item cold-start recommendation method according to an exemplary embodiment of the present invention may be described as various interconnected or coupled functional blocks or functional diagrams. However, these functional blocks or functional diagrams may be equally integrated into a single logic device or operated on by non-exact boundaries.
For example, as described above, there is provided a system comprising at least one computing device and at least one storage device storing instructions, wherein the instructions, when executed by the at least one computing device, cause the at least one computing device to perform steps S101 to S104 as described with reference to fig. 1 or steps S201 to S204 as described with reference to fig. 2. That is, the window-based item cold start recommendation method shown in fig. 1 or fig. 2 may be performed by the computing device described above. Since the above-mentioned window-based article cold start recommendation method has been described in detail in fig. 1 and fig. 2, the contents of this portion of the present invention are not repeated.
There is also provided, in accordance with an embodiment of the present disclosure, a computer-readable storage medium storing instructions that, when executed by at least one computing device, cause the at least one computing device to perform the window-based item cold start recommendation method of any of the preceding embodiments.
By adopting the method and the device, the recently hot articles can be quickly captured, and the articles which are good in performance before and poor in performance now can be quickly eliminated, so that the ecology of a recommendation system is optimized.
While exemplary embodiments are described above, it is not intended that these embodiments describe all possible forms of the disclosure. Rather, the words used in the specification are words of description rather than limitation, and it is understood that various changes may be made without departing from the spirit and scope of the disclosure. Furthermore, features of various implementing embodiments may be combined to form further embodiments of the disclosure.
Claims (10)
1. A window-based item cold start recommendation method comprises the following steps:
recommending the newly generated article according to the exploration flow, and acquiring user behavior data corresponding to the newly generated article;
and acquiring a new recommendation list by using an article cold start recommendation algorithm based on the user behavior data corresponding to the newly generated article and the recommended other articles in the window, so as to recommend the articles in the new recommendation list at a utilization flow rate larger than the exploration flow rate.
2. The method of claim 1, wherein the window comprises a time window or a total amount of user behavior data window.
3. The method of claim 2, wherein the size of the time window depends on a recently recommended activity level.
4. The method of claim 3, wherein the size of the time window is inversely proportional to a sum of a number of displays generated per day when the recommendation is made for the item and a number of user actions.
5. The method of claim 1, wherein the item cold start recommendation algorithm comprises an Epsilon-Greedy algorithm, a Thompson sampling algorithm, or a UCB algorithm.
6. The method of claim 1, wherein the item comprises at least one of textual content, video, pictures, and audio.
7. The method of claim 1, wherein the user behavior data relates to at least one of clicking, forwarding, sharing, commenting, favoring, praising, and clicking on a recommended item.
8. A window-based item cold start recommendation device, comprising:
the exploration unit is used for recommending the newly generated articles according to exploration flow and acquiring user behavior data corresponding to the newly generated articles;
and the utilization unit acquires a new recommendation list by utilizing an article cold start recommendation algorithm based on the user behavior data corresponding to the newly generated article and the recommended other articles in the window, so that the articles in the new recommendation list are recommended at a utilization flow rate larger than the exploration flow rate.
9. A system comprising at least one computing device and at least one storage device storing instructions that, when executed by the at least one computing device, cause the at least one computing device to perform the window-based item cold start recommendation method of any of claims 1-7.
10. A computer-readable storage medium storing instructions that, when executed by at least one computing device, cause the at least one computing device to perform the window-based item cold start recommendation method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910383441.7A CN111915383A (en) | 2019-05-09 | 2019-05-09 | Window-based article cold start recommendation method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910383441.7A CN111915383A (en) | 2019-05-09 | 2019-05-09 | Window-based article cold start recommendation method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111915383A true CN111915383A (en) | 2020-11-10 |
Family
ID=73242788
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910383441.7A Pending CN111915383A (en) | 2019-05-09 | 2019-05-09 | Window-based article cold start recommendation method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111915383A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112579901A (en) * | 2020-12-23 | 2021-03-30 | 雄狮汽车科技(南京)有限公司 | Cold start user recommendation method, system and computer readable storage medium |
| CN112579889A (en) * | 2020-12-07 | 2021-03-30 | 北京百度网讯科技有限公司 | Article recommendation method and device, electronic equipment and storage medium |
| CN114490372A (en) * | 2022-01-20 | 2022-05-13 | 北京火山引擎科技有限公司 | Test scheme determination method and device, computer readable medium and electronic equipment |
| CN112579901B (en) * | 2020-12-23 | 2024-06-11 | 雄狮汽车科技(南京)有限公司 | Cold start user recommendation method, system and computer readable storage medium |
-
2019
- 2019-05-09 CN CN201910383441.7A patent/CN111915383A/en active Pending
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112579889A (en) * | 2020-12-07 | 2021-03-30 | 北京百度网讯科技有限公司 | Article recommendation method and device, electronic equipment and storage medium |
| CN112579901A (en) * | 2020-12-23 | 2021-03-30 | 雄狮汽车科技(南京)有限公司 | Cold start user recommendation method, system and computer readable storage medium |
| CN112579901B (en) * | 2020-12-23 | 2024-06-11 | 雄狮汽车科技(南京)有限公司 | Cold start user recommendation method, system and computer readable storage medium |
| CN114490372A (en) * | 2022-01-20 | 2022-05-13 | 北京火山引擎科技有限公司 | Test scheme determination method and device, computer readable medium and electronic equipment |
| CN114490372B (en) * | 2022-01-20 | 2023-12-26 | 北京火山引擎科技有限公司 | Test scheme determining method and device, computer readable medium and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| RU2720899C2 (en) | Method and system for determining user-specific content proportions for recommendation | |
| RU2632100C2 (en) | Method and server of recommended set of elements creation | |
| JP5905638B2 (en) | User interface content optimization system | |
| CN102306171B (en) | A kind of for providing network to access suggestion and the method and apparatus of web search suggestion | |
| US20150169710A1 (en) | Method and apparatus for providing search results | |
| CN109241403B (en) | Project recommendation method and device, machine equipment and computer-readable storage medium | |
| CN104899220B (en) | Application program recommendation method and system | |
| US20150278359A1 (en) | Method and apparatus for generating a recommendation page | |
| CN111708964A (en) | Multimedia resource recommendation method and device, electronic equipment and storage medium | |
| US20120143911A1 (en) | Recommendations based on topic clusters | |
| CN107003877A (en) | The context deep-link of application | |
| EP3519990A1 (en) | Query-time analytics on graph queries spanning subgraphs | |
| EP2567355A2 (en) | Selecting content based on interest tags that are included in an interest cloud | |
| US9836775B2 (en) | System and method for synchronized web scraping | |
| US20120084657A1 (en) | Providing content to a user from multiple sources based on interest tag(s) that are included in an interest cloud | |
| CN102855261A (en) | Method and device for determining video authority value | |
| US20130212105A1 (en) | Information processing apparatus, information processing method, and program | |
| RU2714594C1 (en) | Method and system for determining parameter relevance for content items | |
| CN105069036A (en) | Information recommendation method and apparatus | |
| CN111552884B (en) | Method and apparatus for content recommendation | |
| CN111522940A (en) | Method and device for processing comment information | |
| WO2020151548A1 (en) | Method and device for sorting followed pages | |
| CN111915383A (en) | Window-based article cold start recommendation method and device | |
| CN104572707A (en) | Preferable object information providing method and device | |
| CN111563198A (en) | Material recall method, device, equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |