CN114707075A - A cold start recommended method and device - Google Patents

Abstract

Embodiments of the present application provide a cold start recommendation method and device, which relate to the technical field of terminals and can solve the cold start problem of a recommendation system. The method is applied to a server, and specifically includes: receiving a request for obtaining a recommendation list from a terminal device, where the request for obtaining a recommendation list includes the model of the terminal device used by the new user; determining the access record of the first product and the access to the first section corresponding to the first model record, the first application is a shopping application, and the second application is a community application; the first recommendation list is obtained according to the first product access record; the second recommendation list is obtained according to the first section access record; the similarity of the first model is determined The second product access record corresponding to the model and the similar model, and according to the second product access record, a third recommendation list is obtained, and the product information in the first recommendation list, the second recommendation list and the third recommendation list are deduplicated, Sort to get the final recommendation list; send the final recommendation list to the terminal device.

Description

A cold start recommended method and device

technical field

The present application relates to the field of terminal technologies, and in particular, to a cold start recommendation method and device.

Background technique

Recommendation technology is widely used in today's Internet products, such as search result recommendation on browser websites, topic recommendation on community websites, product recommendation on shopping websites, and friend recommendation on social networking websites. However, most of the existing recommendations are made by the recommendation system based on a large amount of user data. When a new user uses an Internet product for the first time, since there is no user data of the new user, the recommendation system cannot accurately recommend the new user.

The above-mentioned inability to accurately recommend new users is the cold start problem of the recommendation system. In this regard, most Internet products will design relevant guidance interfaces to guide new users to log in using social network accounts, so as to obtain user data according to the user's social network account, so as to make personalized recommendations for users. However, in order to protect privacy, new users may avoid using social network accounts to log in, so that Internet products will not be able to obtain user data, and the recommendation system will not be able to make accurate recommendations for new users.

SUMMARY OF THE INVENTION

In view of this, the present application provides a cold-start recommendation method and device, which can solve the cold-start problem of the recommendation system.

In a first aspect, the present application provides a cold-start recommendation method, which is applied to a server. The method includes: the server receives a request for obtaining a recommendation list from a terminal device, and the request for obtaining a recommendation list includes the model of the terminal device used by the new user, and the The model is the first model; the first product access record and the first section access record corresponding to the first model are determined, the first product access record is filtered from the product access records of the first application, and the first section access record is obtained from Filtered from the section access records of the second application, the first application is a shopping application, and the second application is a community application; the first recommendation list is obtained according to the first product access record; the first recommendation list is obtained according to the first section access record. Second recommendation list; determine the similar model of the first model and the second product access record corresponding to the similar model, and obtain a third recommendation list according to the second product access record, wherein the first recommendation list, the second recommendation list and the third recommendation list are The three recommendation lists all include product information and the corresponding scoring results of the products; the product information in the first recommendation list, the second recommendation list and the third recommendation list are deduplicated and sorted to obtain the final recommendation list, which is used for New users recommend products; the server sends the final recommendation list to the terminal device.

Usually, when a user wants to buy a product, the user can choose to first query the function introduction of the product through a community application, and then determine whether to buy the product according to the function introduction of the product, or select another product with the same function as the product. After determining the product you want to buy, then open the shopping application to complete the product purchase.

Based on the above-mentioned purchasing habits of users, the present application uses the commodity access records of shopping applications and the section access records of community applications to solve the cold start problem of the recommendation system. Due to the large number of commodity access records of shopping applications and section access records of community applications, the server, in the absence of other valid data of the new user, according to the model of the terminal device used by the new user, records the commodity access records and sections. The access records are preliminarily screened, and the access records of the first product and the access records of the first section are obtained. Then, the access records of the first product and the access records of the first section are analyzed respectively to obtain the first recommendation list and the second recommendation list.

Since the first recommendation list is obtained according to the access record of the first product, the first recommendation list is to determine the user's product preference from the aspect of historical shopping. Since the second recommendation list is obtained according to the access records of the first section, the second recommendation list is based on the community section to determine the user's product preference. At the same time, the present application also determines a third recommendation list based on similar models of the first model. Since the user audience of the similar model and the user audience of the first model have overlapping parts, the third recommendation list derived by using the similar model can also express user preference.

Finally, the first recommendation list, the second recommendation list and the third recommendation list are processed to obtain the final recommendation list. It can be considered that the final recommendation list is obtained through various aspects, such as historical shopping, community sections and similar models. , more in line with the preferences of new users. In this way, not only the cold start problem is solved, but because the recommended products are more in line with the preferences of new users, it can further increase user stickiness and improve user satisfaction.

In a possible implementation manner, after the third recommendation list is obtained, the method further includes: based on the posts involved in each section access record in the first section access record, determining the first section corresponding to the post involved in each section access record. Type word, create a mapping relationship between the post and the first type word, the first type word is the word related to the product name in the post; based on the access record of the first section, determine the user's preferred post; according to the mapping relationship, determine the user's preferred post Corresponding first type words; use fuzzy matching algorithm to determine the commodity information corresponding to the first type words, and generate a fourth recommendation list according to the commodity information corresponding to the first type words; Deduplicate and sort the commodity information in the third recommendation list to obtain the final recommendation list, including: deduplicating and sorting commodity information in the first recommendation list, the second recommendation list, the third recommendation list and the fourth recommendation list , to get the final recommendation list.

In this application scheme, the posts in the access records of the first section are scored to determine the posts that the user prefers. Then, based on the mapping relationship between posts and words related to product names (first type words) in the posts, the first type words corresponding to the posts preferred by the user are obtained. The user's preferred post can determine the user's focus, and then further process the user's focus, that is, determine the first type of words corresponding to the user's preferred post, so as to obtain the words related to the product name that the user focuses on. Since words related to product names cannot directly correspond to products, a fuzzy matching algorithm is also needed to match words related to product names to products, and a fourth recommendation list is obtained after matching. In this application, the user's preference is determined through posts, which is equivalent to starting from the user's focus to obtain a fourth recommendation list that is more suitable for the user's needs. Finally, the first recommendation list, the second recommendation list, the third recommendation list and the fourth recommendation list are processed to obtain a final recommendation list. Therefore, the products in the final recommendation list are more in line with user preferences and have a higher accuracy rate.

In a possible implementation manner, after the third recommendation list is obtained, the method further includes: based on the posts involved in each section access record in the first section access record, determining the second corresponding to the post involved in each section access record. Type word, the second type word is the word related to the product characteristics in the post; sort according to the frequency of occurrence of all the second type words, and generate the first sorting result; based on the second type word in the first sorting result, the product selling point is associated , to generate a fifth recommendation list, which includes the commodity information corresponding to the selling points of the commodities; deduplicate and sort the commodity information in the first recommendation list, the second recommendation list and the third recommendation list to obtain the final recommendation list , including: deduplicating and sorting the commodity information in the first recommendation list, the second recommendation list, the third recommendation list and the fifth recommendation list to obtain the final recommendation list. Alternatively, the product information in the first recommendation list, the second recommendation list, the third recommendation list, the fourth recommendation list, and the fifth recommendation list is deduplicated and sorted to obtain a final recommendation list.

In the solution of the present application, words related to product characteristics (words of the second type) are determined from each block access record, and then the words of the second type are sorted to obtain the first sorting result. The first sorting result can indicate the product features that the user is concerned about, and then associate the selling points of the products according to the product features that the user cares about, which is equivalent to associating the products according to the product features that the user cares about, and finally obtain the fifth recommendation list according to the associated products. That is, the present application determines the features of the products that the user pays attention to according to the section access records accessed by the user, and then obtains the products that the user pays attention to. Since the determined commodities are obtained according to the user's section access records, which are the true expressions of the user's intentions, the derived commodities in the fifth recommended list are more in line with the user's needs. Finally, the first recommendation list, the second recommendation list, the third recommendation list, the fourth recommendation list, and the fifth recommendation list are arranged and combined, and after the arrangement and combination, deduplication and sorting are performed to obtain the final recommendation list. Therefore, the products in the final recommendation list are also more in line with the real intention of the user.

In a possible implementation manner, obtaining the first recommendation list according to the first product access record includes: scoring the products involved in each product access record in the first product access record, and obtaining each product access record The scoring results of the products involved; the scoring results of the same product are accumulated to obtain the scoring results of all the products; according to the scoring results of all the products, sort from high to low to obtain the first recommendation list.

In the solution of the present application, the first recommendation list is obtained by scoring the commodities involved in the first commodity access record. In this way, the obtained products in the first recommended list are more objective and accurate.

In a possible implementation manner, scoring the commodities involved in each commodity access record in the first commodity access record, and obtaining a scoring result of the commodities involved in each commodity access record, including: According to the first interaction between the user and the product, the product involved in each product access record is scored, and the first scoring result is obtained; based on the occurrence time of the first interaction in each product access record, each product access record The involved commodities are scored to obtain the second scoring result; the first scoring result and the second scoring result are multiplied to obtain the scoring results of the commodities involved in each commodity access record.

Different first interaction behaviors can indicate different purchase intentions of the user. For example, the purchase intention expressed by the user's adding to the shopping cart behavior is far greater than the purchase intention expressed by the user's browsing behavior. And with the passage of time, the user's purchase intention will change. Therefore, this application scores the product based on the first interactive behavior and the occurrence time of the first interactive behavior, and the obtained scoring result will be more objective and can better express the user's expectations. purchase intent.

In a possible implementation manner, based on the first interaction between the user and the product in each product access record, the products involved in each product access record are scored to obtain a first scoring result, including: based on each product The first interaction behavior between the user and the product in the access record is compared, and the first interaction behavior between the user and the product is compared with the behavior in the behavior scoring rule to obtain the first scoring result. The behavior scoring rule includes the scoring results corresponding to the behavior and the behavior. The solution of the present application discloses that the scoring result of the first interaction behavior is obtained by comparing the first interaction behavior with the behavior in the behavior scoring rule.

In a possible implementation manner, based on the occurrence time of the first interactive behavior in each commodity access record, the commodities involved in each commodity access record are scored, and a second scoring result is obtained, including: based on each commodity access record For the occurrence time of the first interaction behavior in the record, the occurrence time of the first interaction behavior is input into the time decay function to obtain a second scoring result. The solution of the present application discloses using a time decay function to score the occurrence time of the first interaction behavior.

In a possible implementation manner, obtaining the second recommendation list according to the access records of the first panel, including: determining the user preference panel and the user group corresponding to the user preference panel based on the access records of the first panel; The name of the user, the commodities preferred by the user in the user group are determined from the commodity access records, and a second recommendation list is generated based on the commodities preferred by the user.

Since there are many sections involved in community applications, each section is also very different, for example, the mobile section and the game section. Therefore, this application determines user preference sections from multiple sections based on the access records of the first section, which can greatly reduce the difficulty of data processing and accurately locate user preference sections. Then, the user group is determined according to the user preference section, and then the user-preferred products in the user group are determined from the product access records according to the user group. Then choose the product you like according to the performance of the product. In this way, not only is it more in line with the shopping habits of users, but the products obtained are also more in line with the needs of users.

In a possible implementation manner, determining the user group corresponding to the user preference section and the user preference section based on the first section access record, including: based on the user group in each section access record in the first section access record and the section The second interaction behavior and the occurrence time of the second interaction behavior are scored, and the sections involved in each section access record are scored to obtain the user preference section and the user group corresponding to the user preference section.

Different second interaction behaviors can indicate different degrees of attention of users to posts in the forum. For example, the degree of concern expressed by the behavior of users commenting on posts is much greater than that expressed by users' browsing behaviors. And with the passage of time, the user's focus will change. Therefore, this application determines the section that the user pays more attention to based on the second interactive behavior and the occurrence time of the second interactive behavior. In this way, the obtained user preference section will be more It is objective and can better express the concerns of users.

In a possible implementation manner, before determining the similar models of the first model, the method further includes: distinguishing the commodity access records according to the device models, and based on the differentiated commodity access records, obtaining the corresponding access records of each device model. User behavior sequence, the user behavior sequence includes products accessed by the user; determining the similar models of the first model, including: determining the similarity between the user behavior sequence of the first model and the user behavior sequence of the second model, in the user behavior sequence of the first model. When the similarity between the behavior sequence and the user behavior sequence of the second model meets the threshold, the second model is a similar model of the first model, and the second model is all models other than the first model; determine the second product access corresponding to the similar model. record, and obtain a third recommendation list according to the second product access record, including: distinguishing the product access records according to the device model to obtain a second product access record corresponding to a similar model; 2. Score the commodities involved in the commodity access records, and obtain commodity scoring results corresponding to similar device models; and generate a third recommendation list based on the commodity scoring results.

The present application also considers product recommendation for new users from the level of similar models of the first model. The similar models are specifically determined according to the similarity of the user behavior sequence corresponding to each model in the commodity access record. Then, a third recommendation list is obtained according to the second access records of similar models. Since similar models are determined according to the similarity of user behavior sequences, which include the commodities accessed by users, the similarity of user behavior sequences can also be the similarity of commodities. If the similarity of user behavior sequences of two models is If it is higher, it means that the users of this model and the users of the first model have many overlapping demands for commodities, so this model is regarded as a similar model. Therefore, the subsequent third recommendation list obtained by using the second access record of a similar model is also more in line with the user's needs.

In a possible implementation manner, the model of the terminal device used by the new user is determined as the first model through the application program interface API disclosed by the operating system of the terminal device.

In a possible implementation manner, the commodity access records of the first application and the section access records of the second application are obtained through page embedding points, page embedding points include handwritten embedding points and no embedding points, and handwriting embedding points are pre- At the detection position set, if data is detected, the data will be reported, and all data will be reported if there is no buried point. The present application discloses using two methods of handwritten embedment and no embedment to obtain the commodity access record of the first application and the section access record of the second application.

In a possible implementation manner, the first interaction behavior of the user with the product includes clicking to enter the product details page, saving the product, adding to the shopping cart, clicking to buy the product immediately, submitting the product purchase order, successful payment of the product order, canceling the product collection, Product sharing, searching for products, confirming successful product arrival notification, consulting products, commenting products, and clicking to browse the products in the footprint.

In a possible implementation manner, the second interaction behavior of the user with the forum includes the user browsing posts under the forum, the user's comments under the post, the user's reply to the comment under the post, the user sharing the post, the user bookmarking the post, the user's Likes on posts, details of users who clicked to comment on posts, posts posted by users who clicked to comment on posts, user tips on posts.

In a second aspect, the present application further provides another cold-start recommendation method, which is applied to a terminal device. The method includes: the terminal device, in response to an operation of a new user logging in to a first application, sends a request for obtaining a recommendation list to a server, where the recommendation list request includes a new user. The model of the terminal device used by the user, the model of the terminal device is the first model; the terminal device receives the final recommendation list sent from the server; the terminal device recommends commodities for the new user according to the final recommendation list. In the solution of the present application, when the user triggers product recommendation, the terminal device obtains the final recommendation list through the server, and then recommends the product for the new user according to the final recommendation list.

In a third aspect, a server is provided, and the server has the function of implementing the method of the first aspect above. This function can be implemented by hardware or by executing corresponding software by hardware. The hardware or software includes one or more modules corresponding to the above functions.

In a fourth aspect, a terminal device is provided, and the terminal device has a function of implementing the method of the second aspect. This function can be implemented by hardware or by executing corresponding software by hardware. The hardware or software includes one or more modules corresponding to the above functions.

In a fifth aspect, a server is provided, the server comprising: a wireless communication module, a memory and one or more processors; the wireless communication module, the memory and the processor are coupled;

Wherein, the memory is used to store computer program code, and the computer program code includes computer instructions; when the computer instructions are executed by the processor, the server is caused to execute the method according to any one of the above first aspects.

In a sixth aspect, a terminal device is provided, the terminal device includes: a wireless communication module, a memory, and one or more processors; the wireless communication module, the memory and the processor are coupled;

Wherein, the memory is used to store computer program codes, and the computer program codes include computer instructions; when the computer instructions are executed by the processor, the electronic device is made to execute the method of the second aspect.

In a seventh aspect, a computer-readable storage medium is provided, and instructions are stored in the computer-readable storage medium, and when the computer-readable storage medium is run on a computer, the computer can execute the cold-start recommendation method according to any one of the above-mentioned first aspects. and the cold start recommended method of any one of the above second aspects.

In an eighth aspect, there is provided a computer program product comprising instructions, which, when run on a computer, enables the computer to execute the cold start recommendation method of any one of the above-mentioned first aspect and the method of any one of the above-mentioned second aspect. Cold start recommended method.

In a ninth aspect, an apparatus (for example, the apparatus may be a chip system) is provided, the apparatus includes a processor for supporting a server to implement the functions involved in the first aspect, and supporting a terminal device to implement the second aspect functions involved. In a possible design, the apparatus further includes a memory for storing necessary program instructions and data of the server and necessary program instructions and data of the terminal device. When the device is a system-on-chip, it may be composed of chips, or may include chips and other discrete devices.

Wherein, for the technical effect brought by any one of the design methods in the third aspect to the ninth aspect, reference may be made to the technical effects brought by different design methods in the first aspect and the second aspect, which will not be repeated here.

Description of drawings

FIG. 1 is a schematic diagram of an interface display for logging in to a first application provided by an embodiment of the present application;

FIG. 2 is a schematic structural diagram of a system architecture provided by an embodiment of the present application;

FIG. 3 is a schematic diagram of a hardware structure of a terminal device according to an embodiment of the present application;

FIG. 4 is a schematic diagram of a software structure of a terminal device according to an embodiment of the present application;

FIG. 5 is a schematic diagram of a hardware structure of a server according to an embodiment of the present application;

FIG. 6 is one of the schematic flowcharts of a cold start recommendation method provided by an embodiment of the present application;

FIG. 7 is the second schematic flowchart of a cold start recommendation method provided by an embodiment of the present application;

FIG. 8 is a third schematic flowchart of a cold start recommendation method provided by an embodiment of the present application;

FIG. 9 is a fourth schematic flowchart of a cold start recommendation method provided by an embodiment of the present application;

FIG. 10 is a fifth schematic flowchart of a cold start recommendation method provided by an embodiment of the present application;

FIG. 11 is a sixth schematic flowchart of a cold start recommendation method provided by an embodiment of the present application;

12 is an application schematic diagram of a cold start recommendation method provided by an embodiment of the present application;

FIG. 13 is a schematic structural diagram of a chip system provided by an embodiment of the present application.

Detailed ways

The technical solutions in the embodiments of the present application will be described below with reference to the accompanying drawings in the embodiments of the present application. Wherein, in the description of the present application, unless otherwise specified, "at least one" refers to one or more, and "a plurality" refers to two or more than two. In addition, in order to clearly describe the technical solutions of the embodiments of the present application, in the embodiments of the present application, words such as "first" and "second" are used to distinguish the same or similar items with basically the same function and effect. Those skilled in the art can understand that the words "first", "second" and the like do not limit the quantity and execution order, and the words "first", "second" and the like are not necessarily different.

With the diversification of Internet products, the number of information items, the number of commodities and the types of commodities are growing rapidly. Usually, users need to spend a lot of time to find the information or commodities they need. And with the browsing of a large amount of irrelevant information and products, the user's desire to browse or purchase will be gradually reduced, which will lead to the loss of users.

In order to solve the above problems, the recommendation system came into being. The recommendation system can be understood as: simulating customer service personnel to help users, on the Internet platform, to quickly find the information or products that users need. The personalized recommendation in the recommendation system is to recommend the information and products that the user is interested in according to the user's interest characteristic data and purchase behavior data, so as to improve the user's stickiness with the Internet platform and avoid customer loss.

However, this kind of recommendation system can only accurately recommend information or products to users when it obtains relevant user data such as user interest characteristic data and purchase behavior data. If a new user logs in to the Internet platform for the first time, since the Internet platform does not have the user data of the new user, it is impossible to predict the interest of the new user based on the historical behavior of the new user, so that personalized recommendations cannot be made for the new user. This is also the user coldness in the recommendation system. Startup problem.

In practical applications, in response to the user's cold start problem, most Internet platforms will guide new users to log in using social accounts through carefully set interactive interfaces when they log in for the first time, so as to obtain new users' social account information, such as: follow Relationships, Friendships, Posting, etc. Then use the social account information of the new user to generate an interest model for the new user, and then make accurate recommendations for the new user. But to protect privacy, new users often refuse to log in with social accounts.

Referring to (a) in FIG. 1 , the mobile phone can receive a user's click operation on the icon 102 of the mall APP in the main interface (ie, the desktop) 101 of the mobile phone. If the mall APP is not running, the mobile phone can start the mall APP in response to the click operation. After the mall APP is started, if the user is a new user of the mall APP, the mall APP displays an interface 103 as shown in (b) in FIG. 1 . The interface 103 may include a WeChat™ account shortcut login button, a QQ™ account shortcut login button, and a change login method button 104 (eg, login through a mobile phone number). In response to the user's operation on the change login method button 104 on the interface 103, the user logs in to the mall APP through other login methods. In this case, the mall APP will not be able to obtain the user's social account information, and thus cannot obtain more user information.

Alternatively, when a new user logs in for the first time, the application platform provides users with coarse-grained interest options to capture user interests. For example, a music application platform will allow users to choose the type of music they like to listen to (rock, classical, folk, etc.), and a fitness application platform will allow users to choose the type of sports they are interested in (yoga, skipping, boxing, etc.) . But this kind of interest selection interface requires careful interaction design, otherwise it will affect the user's experience. Moreover, the interests of new users during registration and real visits may change, and the captured interests of users may also not be used for recommendation.

To this end, the present application provides a cold-start recommendation method and device, which captures other user behavior data corresponding to the model of the terminal device used by the new user. And based on other user behavior data corresponding to the model, make accurate recommendations for new users.

As shown in FIG. 2 , a schematic diagram of a system architecture provided by an embodiment of the present application is shown. The system may include terminal devices and servers. Among them, the terminal device may be installed with an application (application, APP), and the application may be a shopping application (eg, Taobao APP, Glory Mall APP, JD APP) and community applications (eg, Glory Club APP, Douban APP), etc. . The server can be an application management server or a cloud server.

In the embodiment of the present application, the terminal sends a request for obtaining a recommendation list to the server in response to the operation of the new user logging in to the first application, where the recommendation list request includes the model (first model) of the terminal device used by the new user. In response to the request for obtaining the recommendation list, the server determines the first product access record and the first section access record corresponding to the first model. Then, the first recommendation list is obtained according to the access record of the first product, and the second recommendation list is obtained according to the access record of the first section. Next, the server determines a similar model of the first model and a second commodity access record corresponding to the similar model, and obtains a third recommendation list according to the second access record. Finally, the first recommendation list, the second recommendation list and the third recommendation list are deduplicated and sorted to obtain the final recommendation list, and the final recommendation list is sent to the terminal device, and the terminal device recommends the user according to the commodities in the final recommendation list.

In a possible design, the terminal device first responds to the operation of the new user logging into the first application, and sends a data acquisition request to the server, and the data acquisition request is used to request to acquire the commodity access record of the first application and the section access of the second application Record. In response to the data acquisition request, the server sends the commodity access record of the first application and the section access record of the second application to the terminal device. After receiving the commodity access record of the first application and the section access record of the second application, the terminal device distinguishes the commodity access record according to the model (the first model) of the terminal device used by the new user, and obtains the corresponding product of the first model. The access record of the first product and the access record of the first section. Then, the first recommendation list is obtained according to the access record of the first product, and the second recommendation list is obtained according to the access record of the first section. Next, a similar model of the first model and a second commodity access record corresponding to the similar model are determined, and a third recommendation list is obtained according to the second access record. Finally, the first recommendation list, the second recommendation list and the third recommendation list are deduplicated and sorted to obtain a final recommendation list, and the terminal device recommends commodities for the user according to the final recommendation list.

Exemplarily, the terminal device in this embodiment of the present application may be a mobile phone, a tablet computer, a desktop computer (desktop computer), a handheld computer, a notebook computer (laptop computer), and an ultra-mobile personal computer (ultra-mobile personal computer, UMPC). ), netbook, personal digital assistant (PDA), augmented reality (AR) \ virtual reality (virtual reality, VR) device and other devices that can install the above target application, the embodiment of the present application is for the first device The specific form is not particularly limited.

The operating system of the terminal device may be Android (Android), Hongmeng system, IOS system or other operating systems, and the embodiment of the present application does not limit the operating system type of the first device.

In the embodiment of the present application, the terminal device shown in FIG. 3 is an electronic device 300 (such as a mobile phone) as an example to illustrate the structure of the terminal device provided by the embodiment of the present application. As shown in FIG. 3 , the electronic device 300 may include a processor 310 , an external memory interface 320 , an internal memory 321 , a universal serial bus (USB) interface 330 , a charge management module 340 , a power management module 341 , and a battery 342 , Antenna 1, Antenna 2, Mobile Communication Module 350, Wireless Communication Module 360, Audio Module 370, Speaker 370A, Receiver 370B, Microphone 370C, Headphone Interface 370D, Sensor Module 380, Key 390, Motor 391, Indicator 392, Camera 393 , display screen 394, and subscriber identification module (subscriber identification module, SIM) card interface 395, pupil detector 396 and iris detector 397 and so on.

The sensor module 380 may include a pressure sensor, a gyro sensor, an air pressure sensor, a magnetic sensor, an acceleration sensor, a distance sensor, a proximity light sensor, a fingerprint sensor, a temperature sensor, a touch sensor, an ambient light sensor, a bone conduction sensor, and the like.

It can be understood that the structure illustrated in this embodiment does not constitute a specific limitation on the electronic device 300 . In other embodiments, the electronic device 300 may include more or fewer components than shown, or some components may be combined, or some components may be split, or a different arrangement of components. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.

The processor 310 may include one or more processing units, for example, the processor 310 may include an application processor (application processor, AP), a modem processor, a graphics processor (graphics processing unit, GPU), an image signal processor ( image signal processor, ISP), controller, memory, video codec, digital signal processor (DSP), baseband processor, and/or neural-network processing unit (NPU), etc. . Wherein, different processing units may be independent devices, or may be integrated in one or more processors.

The controller may be the nerve center and command center of the electronic device 300 . The controller can generate an operation control signal according to the instruction operation code and timing signal, and complete the control of fetching and executing instructions.

A memory may also be provided in the processor 310 for storing instructions and data. In some embodiments, the memory in processor 310 is cache memory. This memory may hold instructions or data that have just been used or recycled by the processor 310 . If the processor 310 needs to use the instruction or data again, it can be called directly from memory. Repeated accesses are avoided, and the waiting time of the processor 310 is reduced, thereby increasing the efficiency of the system.

In some embodiments, processor 310 may include one or more interfaces. The interface may include an integrated circuit (inter-integrated circuit, I3C) interface, an integrated circuit built-in audio (inter-integrated circuitsound, I3S) interface, a pulse code modulation (pulse code modulation, PCM) interface, a universal asynchronous receiver (universal asynchronous receiver) interface /transmitter, UART) interface, mobile industry processor interface (MIPI), general-purpose input/output (GPIO) interface, subscriber identity module (SIM) interface, and/or Universal serial bus (universal serial bus, USB) interface, etc.

It can be understood that the interface connection relationship between the modules illustrated in this embodiment is only a schematic illustration, and does not constitute a structural limitation of the electronic device 300 . In other embodiments, the electronic device 300 may also adopt different interface connection manners in the foregoing embodiments, or a combination of multiple interface connection manners.

The charging management module 340 is used to receive charging input from the charger. While the charging management module 340 charges the battery 342 , the electronic device 300 can also be powered by the power management module 341 .

The power management module 341 is used to connect the battery 342 , the charging management module 340 and the processor 310 . The power management module 341 receives input from the battery 342 and/or the charging management module 340, and supplies power to the processor 310, the internal memory 321, the external memory, the display screen 394, the camera 393, and the wireless communication module 360. In some other embodiments, the power management module 341 may also be provided in the processor 310 . In other embodiments, the power management module 341 and the charging management module 340 may also be provided in the same device.

The wireless communication function of the electronic device 300 can be implemented by the antenna 1, the antenna 2, the mobile communication module 350, the wireless communication module 360, the modulation and demodulation processor, the baseband processor, and the like.

Antenna 1 and Antenna 2 are used to transmit and receive electromagnetic wave signals. Each antenna in electronic device 300 may be used to cover a single or multiple communication frequency bands. Different antennas can also be reused to improve antenna utilization. For example, the antenna 1 can be multiplexed as a diversity antenna of the wireless local area network.

The mobile communication module 350 may provide a wireless communication solution including 2G/3G/4G/5G, etc. applied on the electronic device 300 . The mobile communication module 350 may include at least one filter, switch, power amplifier, low noise amplifier (low noise amplifier, LNA) and the like. The mobile communication module 350 can receive electromagnetic waves from the antenna 1, filter and amplify the received electromagnetic waves, and transmit them to the modulation and demodulation processor for demodulation. The mobile communication module 350 can also amplify the signal modulated by the modulation and demodulation processor, and then convert it into electromagnetic waves for radiation through the antenna 1 .

The modem processor may include a modulator and a demodulator. Wherein, the modulator is used to modulate the low frequency baseband signal to be sent into a medium and high frequency signal. The demodulator is used to demodulate the received electromagnetic wave signal into a low frequency baseband signal. Then the demodulator transmits the demodulated low-frequency baseband signal to the baseband processor for processing. The low frequency baseband signal is processed by the baseband processor and passed to the application processor. The application processor outputs sound signals through audio devices (not limited to the speaker 370A, the receiver 370B, etc.), or displays images or videos through the display screen 394 .

The wireless communication module 360 can provide wireless local area networks (WLAN) (such as wireless fidelity (Wi-Fi) networks), bluetooth (BT), and global navigation satellite systems applied on the electronic device 300 . (global navigation satellite system, GNSS), frequency modulation (frequency modulation, FM), near field communication technology (near field communication, NFC), infrared technology (infrared, IR) and other wireless communication solutions. The wireless communication module 360 may be one or more devices integrating at least one communication processing module. The wireless communication module 360 receives electromagnetic waves via the antenna 2 , frequency modulates and filters the electromagnetic wave signals, and sends the processed signals to the processor 310 . The wireless communication module 360 can also receive the signal to be sent from the processor 310 , perform frequency modulation on it, amplify it, and convert it into electromagnetic waves for radiation through the antenna 2 .

In some embodiments, the antenna 1 of the electronic device 300 is coupled with the mobile communication module 350, and the antenna 2 is coupled with the wireless communication module 360, so that the electronic device 300 can communicate with the network and other devices through wireless communication technology. Wireless communication technologies may include global system for mobile communications (GSM), general packet radio service (GPRS), code division multiple access (CDMA), broadband code division multiple access wideband code division multiple access (WCDMA), time-division code division multiple access (TD-SCDMA), long term evolution (LTE), BT, GNSS, WLAN, NFC, FM, and / or IR technology, etc. GNSS may include global positioning system (GPS), global navigation satellite system (GLONASS), Beidou satellite navigation system (BDS), quasi-zenith satellite system , QZSS) and/or satellite based augmentation systems (SBAS).

The electronic device 300 implements a display function through a GPU, a display screen 394, an application processor, and the like. The GPU is a microprocessor for image processing, and is connected to the display screen 394 and the application processor. The GPU is used to perform mathematical and geometric calculations for graphics rendering. Processor 310 may include one or more GPUs that execute program instructions to generate or alter display information.

Display screen 394 is used to display images, videos, and the like. The display screen 394 includes a display panel. The display panel can use a liquid crystal display (LCD), a light-emitting diode (LED), an organic light-emitting diode (OLED), an active matrix organic light emitting diode or an active matrix organic Active-matrix organic light emitting diode (AMOLED), flexible light-emitting diode (FLED), Miniled, MicroLed, Micro-oLed, quantum dot light emitting diodes (QLED) Wait.

Wherein, if the display screen 394 in the embodiment of the present application integrates a touch sensor, the display screen 394 may be called a touch screen. The touch sensor may also be referred to as a "touch panel". That is, the display screen 394 may include a display panel and a touch panel. A touch sensor is used to detect touch operations on or near it. After the touch sensor detects the touch operation, it can trigger the driver of the core layer (such as the TP driver) to periodically scan the touch parameters generated by the touch operation. Then, the driver of the kernel layer transmits the touch parameters to the relevant modules of the upper layer, so that the relevant modules can determine the touch events corresponding to the touch parameters.

Additionally, display screen 394 may provide visual output related to touch operations. In other embodiments, the touch sensor may also be disposed on the surface of the electronic device 300 instead of being integrated in the display screen 394 . At this time, the positions of the touch sensor and the display screen 394 may be different. In the embodiment of the present application, the specific process of the display method of the playback interface is described by taking the screen as an example of a screen integrated with a touch sensor.

The electronic device 300 can realize the shooting function through the ISP, the camera 393, the video codec, the GPU, the display screen 394, and the application processor. The ISP is used to process the data fed back by the camera 393 . Camera 393 is used to capture still images or video. A digital signal processor is used to process digital signals, in addition to processing digital image signals, it can also process other digital signals. Video codecs are used to compress or decompress digital video. The electronic device 300 may support one or more video codecs. In this way, the electronic device 300 can play or record videos in various encoding formats, such as: Moving Picture Experts Group (moving picture experts group, MPEG) 3, MPEG3, MPEG3, MPEG4, and so on.

NPU is a neural-network (NN) computing processor. By borrowing the structure of biological neural network, such as the transmission mode between neurons in the human brain, it can quickly process the input information, and can also continuously learn by itself. Applications such as intelligent cognition of the electronic device 300 can be implemented through the NPU, such as image recognition, face recognition, speech recognition, text understanding, and the like.

The external memory interface 320 can be used to connect an external memory card, such as a Micro SD card, to expand the storage capacity of the electronic device 300 . The external memory card communicates with the processor 310 through the external memory interface 320 to realize the data storage function. For example to save files like music, video etc in external memory card. Internal memory 321 may be used to store computer executable program code, which includes instructions. The processor 310 executes various functional applications and data processing of the electronic device 300 by executing the instructions stored in the internal memory 321 . For example, in this embodiment of the present application, the processor 310 may execute instructions stored in the internal memory 321, and the internal memory 321 may include a program storage area and a storage data area. The storage program area can store an operating system, an application program required for at least one function (such as a sound playback function, an image playback function, etc.), and the like. The storage data area may store data (such as audio data, phone book, etc.) created during the use of the electronic device 300 and the like. In addition, the internal memory 321 may include high-speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, universal flash storage (UFS), and the like.

The electronic device 300 may implement audio functions through an audio module 370, a speaker 370A, a receiver 370B, a microphone 370C, an earphone interface 370D, and an application processor. Such as music playback, recording, etc.

The audio module 370 is used for converting digital audio information into analog audio signal output, and also for converting analog audio input into digital audio signal. Audio module 370 may also be used to encode and decode audio signals. Speaker 370A, also referred to as "horn", is used to convert audio electrical signals into sound signals. The receiver 370B, also referred to as "earpiece", is used to convert audio electrical signals into sound signals. The microphone 370C, also called "microphone" or "microphone", is used to convert sound signals into electrical signals. The headphone jack 370D is used to connect wired headphones.

The keys 390 include a power-on key, a volume key, and the like. Keys 390 may be mechanical keys. It can also be a touch key. The electronic device 300 may receive key inputs and generate key signal inputs related to user settings and function control of the electronic device 300 . Motor 391 can generate vibrating cues. The motor 391 can be used for incoming call vibration alerts, and can also be used for touch vibration feedback. The indicator 392 can be an indicator light, which can be used to indicate the charging status, the change of power, and can also be used to indicate messages, missed calls, notifications, and the like. The SIM card interface 395 is used to connect a SIM card. The SIM card can be contacted and separated from the electronic device 300 by inserting into the SIM card interface 395 or pulling out from the SIM card interface 395 . The electronic device 300 may support 3 or N SIM card interfaces, where N is a positive integer greater than 3. The SIM card interface 395 can support Nano SIM card, Micro SIM card, SIM card and so on.

The software system of the above mobile phone may adopt a layered architecture, an event-driven architecture, a microkernel architecture, a microservice architecture, or a cloud architecture. The embodiments of the present invention take an Android® system with a layered architecture as an example to illustrate the software structure of a mobile phone.

The layered architecture divides the software into several layers, and each layer has a clear role and division of labor. Layers communicate with each other through interfaces. In some embodiments, the Android® system is divided into four layers, from top to bottom, the application layer, the application framework layer, the Android runtime and system libraries, and the kernel layer.

The application layer can include a series of application packages.

As shown in Figure 4, the application package may include applications such as camera, mail, calendar, call, map, navigation, WLAN, Bluetooth, music, mall, community club and so on.

The application framework layer provides an application programming interface (application programming interface, API) and a programming framework for the applications of the application layer. The application framework layer includes some predefined functions.

The application framework layer may include a window manager, a content provider, an activity manager, a resource manager, a notification manager, a view system, etc., which is not limited in this embodiment of the present application.

Activity Manager (Activity Manager): used to manage the life cycle of each application. Applications usually run in the operating system in the form of Activity. For each Activity, there will be a corresponding application record (ActivityRecord) in the activity manager, and this ActivityRecord records the state of the application's Activity. The activity manager can use this ActivityRecord as an identifier to schedule the Activity process of the application.

Window Manager (WindowManagerService): used to manage the graphical user interface (GUI) resources used on the screen, which can be used for: obtaining the screen size, creating and destroying windows, displaying and hiding windows, and layout of windows , focus management and input method and wallpaper management.

The system library and kernel layer below the application framework layer can be called the underlying system. The underlying system includes the underlying display system used to provide display services. For example, the underlying display system includes the display driver in the kernel layer and the surface in the system library. manager, etc.

The Android Runtime includes core libraries and a virtual machine. The Android runtime is responsible for the scheduling and management of the Android system. The core library consists of two parts: one is the function functions that the java language needs to call, and the other is the core library of Android. The application layer and the application framework layer run in virtual machines. The virtual machine executes the java files of the application layer and the application framework layer as binary files. The virtual machine is used to perform functions such as object lifecycle management, stack management, thread management, safety and exception management, and garbage collection.

A system library can include multiple functional modules. For example: surface manager (surface manager), media library (Media Libraries), 3D graphics processing library (eg: OpenGL ES), 2D graphics engine (eg: SGL), etc.

The Surface Manager is used to manage the display subsystem and provides a fusion of 2D and 3D layers for multiple applications.

The media library supports playback and recording of a variety of commonly used audio and video formats, as well as still image files. The media library can support a variety of audio and video encoding formats, such as: MPEG4, H.264, MP3, AAC, AMR, JPG, PNG, etc.

OpenGL ES is used to implement 3D graphics drawing, image rendering, compositing, and layer processing.

SGL is a drawing engine for 2D drawing.

The kernel layer is the layer between hardware and software. The kernel layer contains at least display drivers, camera drivers, audio drivers, and sensor drivers.

In the embodiment of the present application, the server 500 shown in FIG. 5 is used as an example to illustrate the structure of the server provided by the embodiment of the present application. As shown in FIG. 5 , a schematic diagram of a hardware structure of a server provided by an embodiment of the present application is shown. The server 500 includes at least one processor 501 , communication lines 502 , memory 503 and at least one communication interface 504 .

The processor 501 may be a general-purpose central processing unit (CPU), a microprocessor, an application-specific integrated circuit (ASIC), or one or more processors for controlling the execution of the programs of the present application. integrated circuit.

Communication line 502 may include a path to communicate information between the components described above.

The communication interface 504, using any device such as a transceiver, is used to communicate with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), and the like.

The memory 503 may be read-only memory (ROM) or other types of static storage devices that can store static information and instructions, random access memory (RAM) or other types of information and instructions that can be stored It can also be an electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM), or other optical disk storage, optical disk storage ( including compact discs, laser discs, compact discs, digital versatile discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or capable of carrying or storing desired program code in the form of instructions or data structures and capable of being stored by a computer any other medium taken, but not limited to this. The memory may exist independently and be connected to the processor through communication line 502 . The memory can also be integrated with the processor.

The memory 503 is used for storing computer-executed instructions for executing the solution of the present application, and the execution is controlled by the processor 501 . The processor 501 is configured to execute the computer-executed instructions stored in the memory 503, so as to implement the communication method of the multicast/broadcast service provided by the following embodiments of the present application.

Optionally, the computer-executed instructions in the embodiment of the present application may also be referred to as application code, which is not specifically limited in the embodiment of the present application.

In a specific implementation, as an embodiment, the processor 501 may include one or more CPUs, such as CPU0 and CPU1 in FIG. 5 .

In a specific implementation, as an embodiment, the server 500 may include multiple processors, such as the processor 501 and the processor 507 in FIG. 5 . Each of these processors can be a single-CPU processor or a multi-CPU processor. A processor herein may refer to one or more devices, circuits, and/or processing cores for processing data (eg, computer program instructions).

Optionally, the server 500 may further include an output device 505 and an input device 506 . The output device 505 is in communication with the processor 501 and can display information in a variety of ways. For example, the output device 505 may be a liquid crystal display (LCD), a light emitting diode (LED) display device, a cathode ray tube (CRT) display device, or a projector (projector), etc. . Input device 506 is in communication with processor 501 and can receive user input in a variety of ways. For example, the input device 506 may be a mouse, a keyboard, a touch screen device, a sensor device, or the like.

The above-mentioned server 500 may be a general-purpose device or a dedicated device. In a specific implementation, the server 500 may be a desktop computer, a portable computer, a network server, a personal digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, an embedded device, or a device with a similar structure in FIG. 5 . . This embodiment of the present application does not limit the type of the server 500 .

The following uses FIG. 6 as an example to describe a cold start recommendation method provided by the embodiment of the present application, which specifically includes:

601. Determine the model of the terminal device used by the new user to access the first application.

The model of the terminal device used by the new user is the first model.

After a new user enters the first application, the first application guides the new user to log in to the first application through the interactive interface, and after logging in to the first application, the first application displays a recommendation interface for the new user. Before the first application displays the recommendation interface for the new user, the first application needs to determine the model of the terminal device used by the new user.

Optionally, the first application may be an application program in a mobile phone, or a website platform in a computer. The recommendation interface displayed for the user can be used for recommending books, recommending social friends, recommending videos, recommending music, recommending restaurants, recommending products, and so on.

The first application determines the model of the terminal device used by the new user. Specifically, the terminal device information of the new user may be determined through the operating system of the terminal device used by the new user. At present, the operating systems of most terminal devices have a public application program interface (Application Program Interface, API) for a third-party application (Aapplication, App) to obtain some basic information of the terminal device. After being authorized by the user, the third-party App can gain access to the API exposed by the operating system of the terminal device, and then obtain the required information through the API.

In the Android operating system, the third-party App (ie, the first application) can obtain the terminal device model and all the installed APP information (installation package name). In the iOS® system, the third-party App (ie, the first application) cannot directly obtain the list of apps installed on the user terminal device, but can obtain this information indirectly by querying whether an App is installed on the terminal device. This method requires a third-party app (ie, the first app) to have a built-in id list of commonly used iOS apps, and the id list of commonly used iOS apps can be captured from the web version of the App Store.

602. Acquire a commodity access record of the first application.

The commodity access record of the first application may be acquired by a terminal device or acquired by a server. If it is acquired by the terminal device, the specific acquisition method may be that the terminal device sends an instruction to obtain the commodity access record to the cloud server of the first application. Commodity access records. It may also be that the cloud server of the first application periodically actively sends the commodity access records of the first application to all terminal devices that install the first application. Or after the cloud server of the first application detects the new commodity access record of the first application, it automatically sends the new commodity access record to all terminal devices that install the first application. Alternatively, when the terminal device downloads the installation package of the first application, it is automatically carried in the installation package of the first application. As of the current moment, all commodity access records on the first application are recorded. In the case of server acquisition, the specific acquisition method is that the server sends an instruction to acquire the commodity access record to the cloud server of the first application, and the cloud server of the first application sends the commodity access record of the first application to the server in response to the instruction to acquire the commodity access record.

The first application may be a shopping application, for example, a mall APP, a shopping APP, and the like.

Commodity access records can be obtained through page embedding technology. The page embedding technology can obtain data in two ways: handwritten embedding and no embedding. Handwriting embedment means: after determining the data that needs to be monitored (such as the user's click behavior, browsing behavior, etc.), then write code in the corresponding position. When the third-party App detects the preset monitoring data, it directly reports the detected data. This method of handwriting buried points is very flexible, but the workload is also large, and corresponding codes must be inserted in every place that needs to be monitored. No burying point means: no need for developers to manually bury points, and obtain data by counting all events and reporting them regularly. Although the method without buried points is not as cumbersome as handwritten buried points, because all events are counted, it is necessary to filter out the required data later.

Usually, page embedding monitors data such as page views (Page View, PV), the number of unique visitors (UniqueVisitor, UV), dwell time, and user interaction behavior. Among them, PV refers to the page views or clicks of the website. UV refers to the number of visitors distinguished by IP address. If an IP address is repeatedly accessed within a period of time, it is also a UV. The stay duration specifically refers to the time a user stays on a certain page or a visit (session) of a third-party App. User interaction behaviors refer to all behaviors triggered by the user on the first application, such as browsing, clicking, bookmarking, sharing, and the like.

The commodity access record refers to the record of the user's visit to the commodity in the shopping application stored by the first application within a preset period of time. The commodity access records may include user behavior records and user equipment records.

The user behavior record refers to the record of the first interaction behavior between the user and the product. For example, the user behavior record may include the name of the user, the name of the commodity accessed by the user, the first interaction behavior of the user with the commodity, and the occurrence time of the first interaction behavior. The first interaction between the user and the product can include clicking to enter the product details page, saving the product, adding to the shopping cart, clicking to buy the product immediately, submitting the product purchase order, successful payment of the product order, canceling the product collection, product sharing, searching for products, and product arrival. Product notification confirms success, consults products, reviews products, and clicks to browse products in the footprint. The user behavior record can directly reflect the real needs of the user, and by analyzing the first interaction behavior between the user and the product, the product that the user is interested in can be determined. Personalized recommendations for users based on the products they are interested in can better meet the real needs of users.

The user equipment record may be a record of the terminal equipment used by the user when the user accesses the first application. For example, the user equipment record may include the user's name, the terminal device model used by the user, and the time when the device runs the first application.

If the user accessing the first application is a new user, the first application cannot provide data for reference, so this application uses the terminal device model used by the new user to query the commodity access records of other users of the same terminal device model , to provide recommendation data for new users based on other users' product access records.

Exemplarily, the commodity access records are shown in Table 1 below.

Table 1

603. In the commodity visit record, determine the first commodity visit record corresponding to the first model.

There may be multiple commodity access records of the first application, and as the online time of the first application is longer, the number of commodity access records also increases. From a plurality of commodity access records, a commodity access record using a terminal device of the first model is selected to obtain a first commodity access record.

Exemplarily, in combination with Table 1, a first commodity access record is obtained, and the first commodity access record is shown in Table 2 below.

Table 2

604. Obtain a first recommendation list according to the access record of the first product.

Based on each commodity access record in the first commodity access record, the commodities involved in each commodity access record are scored, and a scoring result of the commodities involved in each commodity access record is obtained. Then, based on the product name, the scoring results of the same product are accumulated to obtain the final scoring results of all products. Then, according to the scoring results of all the commodities, the commodities are sorted, and the first recommendation list is generated in combination with the sorting results. Finally, products may be recommended for new users according to the product information in the first recommendation list.

The scoring of the commodities involved in each commodity access record is carried out from two aspects. On the one hand, it is scored according to the first interaction between the user and the commodity in each commodity access record, and on the other hand, it is based on each commodity. The time of occurrence of the first interaction in the access record is scored. Finally, multiply the scoring results of the above two aspects to obtain the final scoring result of the commodities involved in each commodity access record.

Scoring according to the first interaction behavior between the user and the product in each product access record, specifically, comparing the first interaction behavior between the user and the product in each product access record with the behavior in the behavior scoring rules. When the first interaction between the user and the product in the product access record is consistent with one of the behaviors in the behavior scoring rules, the scoring result of the behavior specified in the behavior scoring rules will be used as the current product access record. The scoring result of the first interaction behavior. If the first interaction behavior between the user and the product in the product access record has no corresponding behavior in the behavior scoring rule, the scoring result of the first interaction behavior between the user and the product is marked as 0. Exemplary, behavior scoring rules are shown in Table 3:

table 3

Based on Table 2 and Table 3, the first interaction behavior between the user and the product is scored, and the scoring results obtained are shown in Table 4 below:

Table 4

Scoring according to the occurrence time of the first interaction in each commodity access record may be a scoring result of determining the occurrence time of the first interaction in the commodity access record based on a time decay function. Exemplarily, the time decay function is as follows:

Among them, t0 is the initial temperature, t1 is the cooling coefficient, t0 and t1 are fixed coefficient values, and the interval time refers to the interval between the occurrence time of the first interaction behavior and the time to be recommended. Exemplarily, t0=1, t1=0.2.

Substitute the occurrence time of the first interaction in each commodity access record into the above time decay function, calculate F(t), and then normalize the result of F(t) to obtain a [0, 1] value in between.

As we all know, the purchasing needs of users will change over time. The closer to the recommendation time, the more informative the first interaction between the user and the product. If you recommend products for new users, the time is October 19th. See Table 2, Xiao Wang entered the details page of product A on March 10, and Xiao Zhao collected product C on October 18. Comparatively speaking, if October 18th is closer to the recommended time of October 19th, the score on October 18th is higher than that on March 10th.

After determining the scoring result of the first interaction behavior between the user and the product and the scoring result of the occurrence time of the first interaction behavior, the scoring result of the product involved in each product access record satisfies the following formula:

A=B*C

Among them, A is the scoring result of the product involved in each product access record, B is the scoring result of the first interaction behavior in each product access record, and C is the time when the first interaction behavior in each product access record occurred. Scoring results.

Finally, multiply the scoring result of the first interactive behavior in Table 4 and the scoring result of the occurrence time of the first interactive behavior calculated based on the time decay function to obtain the scoring value of the corresponding product in each product access record. The scoring value of the corresponding product in each product access record is shown in Table 5 below:

table 5

After obtaining the scoring result of the commodity corresponding to each commodity access record, the scoring results of the same commodity are accumulated to obtain the summation and scoring result value corresponding to each commodity. Combining with Table 6, the product corresponding to the first product access record is product A, the product corresponding to the third product access record is also product A, and the product corresponding to the Nth product access record is product C, and then all the products of product A are The scoring results are accumulated, that is, the scoring result of 0.025 obtained from the first product access record and the scoring result of 0.04 obtained from the third product access record are accumulated, and the corresponding scoring result of product A is 0.065. Then, the scoring results of other commodities involved are calculated in turn.

Finally, a recommendation list is generated according to the scoring results of all commodities, and the recommendation list includes commodity information and the scoring results corresponding to the commodities, wherein the commodity information may be the commodity name and the recommended serial number. Exemplarily, the recommended list is shown in Table 6:

Table 6

After the recommendation list is obtained, the first N products in the recommendation list can be selected according to the recommendation requirements to form the first recommendation list. Exemplarily, the recommendation requirement may be to select the top 10 commodities in the recommendation list as the recommended commodities. Alternatively, products whose scoring results exceed the threshold are regarded as recommended products. The specific recommended requirements can be adjusted according to actual needs, which is not limited in this application. After the first recommendation list is obtained, the product information in the first recommendation list and the scoring result corresponding to the product may be recommended for new users, or the first recommendation list and other recommendation lists may be combined for new user recommendation.

Specifically, 603-604 may be executed by a terminal device, or may be executed by a server. If 603-604 are executed by the server, before 603, the method further includes: the server receives a request for obtaining a recommendation list from the terminal device, where the request for obtaining a recommendation list includes the model of the terminal device used by the new user, the model of the terminal device for the first model. After obtaining the first recommendation list through 604, the server sends the first recommendation list to the terminal device, and the terminal device recommends commodities for the new user based on commodity information in the first recommendation list.

Another cold start recommendation method provided by the embodiment of the present application is described below by taking FIG. 7 as an example, and the method further includes:

701. Determine the model of the terminal device used by the new user to access the first application.

See 601 for details.

702. Acquire a section access record of the second application.

The section access record of the second application may be obtained by the terminal device or obtained by the server. If it is acquired by a terminal device, the specific acquisition method may be that the terminal device sends an instruction to acquire the section access record to the cloud server of the second application. The block access record; it can also be that the cloud server of the second application periodically actively sends the block access record of the second application to all terminal devices where the second application is installed; or the cloud server of the second application detects a new update of the second application. After the section access record of the second application is saved, the new section access record is automatically sent to all terminal devices that install the second application; or when the terminal device downloads the installation package of the second application, it is automatically carried in the installation package of the second application. At the moment, access records of all sections on the second application. In the case of server acquisition, the specific acquisition method is that the server sends an instruction to acquire the block access record to the cloud server of the second application, and the cloud server of the second application sends the block access record of the second application to the server in response to the instruction to acquire the block access record.

The second application may be a community application, such as a community club APP.

The Community Club APP can provide multiple interest sections for users to communicate and interact. Multiple interest sections may include: mobile phone section, smart life section, user interface design (UI) section, cloud service section, interest block section, game section and offline club section. Among them, the posts under the mobile phone section are the function introductions of different mobile phone models; the posts under the smart life section are the function introductions of smart devices, which can be notebooks, tablets, earphone speakers, smart screens, routers, and smart wearable devices, etc.; UI The posts under the section are introductions of UI-related functions; the posts under the cloud service section are introductions of functions related to cloud services; the posts under the interest block section are introductions of different hobbies, such as love of food, love of travel, and love of sports , love digital, love themes, love games, love photography and slow life, etc. The post under the game section is the introduction of the game; the post under the offline club section is the introduction of offline activities.

Section access records can also be obtained through page embedding technology. For details on page embedding, please refer to the description at 402, and details will not be repeated here.

The section access record refers to the record of the user's access to each section in the community application stored by the second application within a preset period of time. Board access records can include board behavior records and board device records.

The forum behavior record refers to the record of the user's second interaction behavior with the forum. For example, the section behavior record may include the user's name, the section name browsed by the user, the post name under the section name browsed by the user, the user's second interaction with the section, and a record of the time when the second interaction occurred. The second interaction behavior of the user with the forum may include the user browsing the posts under the forum, the user's comment below the post, the user's reply to the comment below the post, the user sharing the post, the user bookmarking the post, the user liking the post, the user clicking The details of the user who commented on the post, the post published by the user who clicked on the commented post, the user's tip for the post.

The board device record may be a record of the terminal device used by the user when the user accesses the second application. For example, the board device record may include the user's name, the terminal device model used by the user, and the time when the device runs the second application. For the specific content of the block access record, reference may be made to the content in Table 1, which will not be described in detail in this application.

In addition, the section access record may only include the section access record of the logged-in user of the second application, and may also include the section access record of the non-login user of the second application. The login user refers to a user who has registered an account when accessing the second application, and has logged into the account for access. A non-login user refers to a user who accesses the second application without a login account when accessing the second application. When the user accesses the second application, the second application reminds the user to log in to the second application through a pop-up window. If the user refuses to log in to the second application, it will be marked and numbered with the non-login user, and then all the second interactions of the non-login user will be recorded. behavior until the non-login user exits the second application.

703. In the block access record, determine the first block access record corresponding to the first model.

There may be multiple access records of the second application, and as the online time of the second application is longer, the number of access records of the second application will increase. From the plurality of section access records, the section access records using the first model of equipment are selected to obtain the first section access records. For the specific content of the access record of the first block, reference may be made to the content in Table 2, which will not be repeated in this application.

704. Determine the user preference section and the user group corresponding to the user preference section based on the access record of the first section.

Based on the second interaction behavior and the occurrence time of the second interaction between the user and the block in each block access record in the first block access record, the blocks involved in each block access record are scored, so as to obtain the user's preferred block.

Exemplarily, referring to 604, the scoring of the forum is divided into two aspects, on the one hand, the score is based on the second interaction behavior between the user and the forum in each forum access record, and on the other hand, the score is based on the second interaction behavior in each forum access record. The time-to-occurrence of the interaction is scored. Finally, multiply the scoring results of the above two aspects to obtain the final scoring results of the sections involved in each section access record.

In the second interaction behavior, the user's behavior of browsing the posts in the forum, the corresponding score is 0.25; the user's behavior of commenting under the post in the forum, the corresponding score is 0.3; the user's behavior of sharing the posts in the forum , the corresponding scoring result is 0.4; the behavior of the user's favorite post in the forum, the corresponding scoring result is 0.4, and the behavior of the user liking the post in the forum, the corresponding scoring result is 0.6. The scoring results of the second interaction behavior between other users and the forum can be set in combination with actual needs, which is not limited in this application.

The score of the occurrence time of the second interaction behavior may be specifically calculated according to the time decay function in 604 above.

After determining the scoring result of the second interaction behavior between the user and the forum and the scoring result of the occurrence time of the second interaction behavior, the scoring result of the forum involved in each forum access record satisfies the following formula:

A1=B1*C1

Among them, A1 is the scoring result of the section involved in each section access record, B1 is the scoring result of the second interaction behavior in each section access record, and C1 is the occurrence time of the second interaction behavior in each section access record. Scoring results.

The scoring results of the sections involved in the access records of all sections are calculated in turn, and then based on the section name, the scoring results of the same section are accumulated to obtain the final scores of all involved sections. Then, according to the scoring results of all the sections, the sections are sorted, and according to the sorting results, the top 3 sections are used as the user preference section of the same model.

Then, from the access records of the first panel, based on the name of the user preference panel, the access records of the second panel that only contain the user preference panel are filtered out. Combining with 702, it can be known that each section access record includes the user's name. Then record the name of the user in the access record of each section in the second section visit, and then delete the duplicate name to obtain the user group corresponding to the user preference section.

First, perform the first screening through the same model, and filter out the access records of the device with the same model as the terminal device used by the new user. Then, the sections in the screened section access records are scored, and then the user preference sections are screened out. Finally, based on the user's preference section, making personalized recommendations for users will better meet the needs of users.

705. Based on the name of each user in the user group, determine the commodity preferred by each user in the user group from the commodity access record, and generate a second recommendation list based on the commodity preferred by the user.

According to the name of each user in the user group, the commodity access records of each user in the user group are indexed from the commodity access records, and then the commodities involved in each commodity access record are scored according to the commodity access records of each user. For the specific scoring process, please refer to 604, and then sort based on the scoring results to obtain the products preferred by the users in the user group. Finally, a second recommendation list is generated according to the product information of the product preferred by the user and the corresponding scoring result of the product. The obtained second recommendation list can be directly used to recommend new users, or the second recommendation list can be combined with other recommendation lists to form a new recommendation list. User recommendation.

Among them, 703-705 may be executed by the terminal device, and may also be executed by the server. If 703-705 are performed by the server, before 703, the method further includes: the server receives a request for obtaining a recommendation list from the terminal device, where the request for obtaining a recommendation list includes the model of the terminal device used by the new user, the model of the terminal device for the first model. After obtaining the second recommendation list through 705, the server sends the second recommendation list to the terminal device, and the terminal device may recommend commodities for the new user based on commodity information in the second recommendation list.

Another cold start recommendation method provided by the embodiment of the present application is described below by taking FIG. 8 as an example, and the method further includes:

801. Determine the model of the terminal device used by the new user to access the first application.

See 601 for details.

802. Obtain a section access record of the second application.

See 702 for details.

803. In the block access record, determine the first block access record corresponding to the first model.

See 703 for details.

804. Based on the posts involved in each section access record in the first section access record, determine a first type word corresponding to the post involved in each section access record, and create a mapping relationship between the post and the first type word.

Among them, the first type of words are words related to product names involved in the post.

The first section access record is the section access record using the terminal device of the first model. View the access record of the first forum, you can learn the posts involved in each access record of the forum in the access record of the first forum. Then, by looking up the posts, words related to the product name involved in each post can be obtained, and the words related to the product name are the first type words. And establish a mapping relationship between the post and the first type word corresponding to the post. Exemplarily, if the post involved in the access record of the first section is a post in the mobile section, the word related to the product name (ie, the first type word) in the post may be the mobile phone name.

805. Determine, based on the access record of the first section, a post preferred by the user.

Combining with 702, it can be seen that the first section access record includes the second interaction between the user and the section and the occurrence time of the second interaction, wherein the second interaction between the user and the section is usually the interaction between the user and the posts under the section. , that is, the occurrence time of the second interaction behavior and the second interaction behavior between the user and the forum, which can be considered as the occurrence time of the second interaction behavior and the second interaction behavior between the user and the post. Then, score the second interaction behavior of the user and the post and the occurrence time of the second interaction behavior in each section access record in the first section access record, so as to determine the scoring result of the post involved in each section access record , and finally accumulate the scoring results of the same post to obtain the final scoring value of each post. Wherein, for the scoring of the second interaction behavior between the user and the post and the occurrence time of the second interaction behavior, reference may be made to 704, which will not be repeated here.

Then, according to the scoring results of the posts, the posts are sorted, and according to the sorting results, the user preference posts of the same model are filtered out. Exemplarily, the top 10 posts in the sorting result may be selected as user preference posts.

806. According to the mapping relationship, determine the first type word corresponding to the post preferred by the user.

Combining with 804, after the user's preferred post is determined, according to the mapping relationship, the first type word corresponding to the user's preferred post is obtained.

807. Use a fuzzy matching algorithm to determine commodity information corresponding to the first type word, and generate a fourth recommendation list according to the commodity information corresponding to the first type word.

Exemplarily, fuzzy matching can be implemented through FuzzyWuzzy. FuzzyWuzzy is a simple and easy to use fuzzy string matching toolkit. It calculates the difference between two sequences based on the edit distance (Levenshtein Distance) algorithm. Edit distance is the minimum number of edit operations required to convert two strings from one to the other. Editing operations can include replacing one character with another, inserting a character, and deleting a character. In general, the smaller the edit distance, the greater the similarity between the two sequences.

FuzzyWuzzy can perform fuzzy matching by any one of the four functions: simple matching (Ratio), partial matching (Partial Ratio), ignoring order matching (Token Sort Ratio), and deduplication subset matching (Token Set Ratio), thereby according to The first type word matches the item.

For example, the user's preferred posts are the following post 1, post 2 and post 3. The title of post 1 is: [Photography Tutorial] How to take a good-looking INS green-orange night scene, commodity name A. The title of post 2 is: What is the flagship machine, and the product name A tells you the answer. The title of post 3 is: The night view street shot of the product name B in May.

The first type word corresponding to post 1 is "product name A", the first type word corresponding to post 2 is "product name A", and the first type word corresponding to post 3 is "product name B". Then, the "commodity name A" corresponding to post 1, the "commodity name A" corresponding to post 2, and the "commodity name B" corresponding to post 3 are related to the product catalog tree of the vertical domain through FuzzyWuzzy in turn. Or after the first type words of all the posts are determined, the first type words of all the posts are related to the product directory tree name of the vertical domain together through FuzzyWuzzy. For example, if "commodity name A" is "A Yao Magic4 Pro" and commodity B is "V30", then according to FuzzyWuzzy, "commodity name A" can be matched to the product " Yaoyao Magic4 Pro", and "commodity name B" can be matched to "A Yaoyao V30".

Among them, vertical domain refers to a small domain that is vertically subdivided under a large domain. Vertical refers to vertical extension, not horizontal expansion. Segmentation refers to the in-depth development of major businesses in the vertical industry sector. Taking mobile phones as an example, mobile phones are a vertical field, which can be subdivided into certain mobile phones, certain Yao mobile phones, certain Mi mobile phones, and certain fruit mobile phones.

The product catalog tree is used to represent the hierarchical relationship of all products under a brand, so that the development of all products under the brand can be clearly described, and the pros and cons of the products can be further shown. The hierarchical relationship of products can be determined according to the performance of product components.

After the product information is matched by FuzzyWuzzy, the matched product information is sorted according to the scoring result of the posts preferred by the user, and then a fourth recommendation list is generated. Subsequent recommendations may be made for new users according to the product information in the fourth recommendation list and the scoring results corresponding to the products; or the fourth recommendation list may be combined with other recommendation lists to be recommended for new users.

Among them, 803-807 may be executed by the terminal device, and may also be executed by the server. If 803-807 are performed by the server, before 803, the method further includes: the server receives a request for obtaining a recommendation list from the terminal device, where the request for obtaining a recommendation list includes the model of the terminal device used by the new user, the model of the terminal device for the first model. After obtaining the fourth recommendation list through 807, the server sends the fourth recommendation list to the terminal device, and the terminal device recommends products for the new user based on the product information in the fourth recommendation list.

Another cold start recommendation method provided by the embodiment of the present application is described below by taking FIG. 9 as an example, and the method further includes:

901. Determine the model of the terminal device used by the new user to access the first application.

See 601 for details.

902. Obtain a section access record of the second application.

See 702 for details.

903. In the block access record, determine the first block access record corresponding to the first model.

See 703 for details.

904. Based on the posts involved in each of the forum access records in the first forum access record, determine a second type word corresponding to the posts involved in each of the forum access records.

Among them, the second type of words are words related to product characteristics involved in the post.

The first section access record is the section access record using the terminal device of the first model. View the access record of the first forum, you can learn the posts involved in each access record of the forum in the access record of the first forum. Then, by looking up the posts, it is possible to obtain the words related to the product characteristics involved in each post, and the words related to the product characteristics are the second type words. Exemplarily, if the post involved in the access record of the first section is a post under the mobile section, the words related to product characteristics (ie, the second type word) in the post may be chips, cameras, memory, and batteries.

905. Sort all the second type words according to the frequency of occurrence, generate a first sorting result, and associate the selling points of the commodities based on the second type words in the first sorting result to generate a fifth recommendation list.

The fifth recommendation list includes commodity information corresponding to the commodity selling point.

According to the frequency of occurrence of the second type words, all the second type words are sorted to obtain the first sorting result. By counting the frequency of the second type of words in the post, the first sorting result is obtained, and the first sorting result may be: camera>chip>battery.

Wherein, the second type of words may be determined according to the product dictionary, or may be determined according to the trained model. Usually, the product dictionary includes products and product characteristics. After determining the products introduced by the post, the second type of words can be obtained. Alternatively, the post is fed into the trained model, and the model outputs the second type word corresponding to the post. The training process of the model is as follows: obtaining multiple posts and their corresponding second-type words, and the second-type words of each post are manually annotated. Then multiple posts are input into the initial model to obtain the initial second-type words corresponding to each post, and finally the initial model is corrected by the loss function, so that the deviation between the output second-type words and the artificially-labeled second-type words is consistent with Threshold to get the trained model.

A commodity selling point database may be stored in the terminal device, and the commodity selling point database includes all commodities in the mall and commodity selling points of each commodity. Fuzzy matching is performed between the commodity selling points in the commodity selling point database and each second type word in the first sorting result, so as to obtain the commodity selling point corresponding to each second type word. A second type word may match the selling point of multiple products. Then, according to the performance of the commodity selling point of each commodity in the plurality of commodities, the commodities are sorted to obtain the initial commodity ranking corresponding to the second type word.

Combining 904, there are multiple second-type words in the first sorting result, each second-type word is matched with a commodity selling point, and one second-type word can be matched with multiple commodity selling points. Then, the matched commodity selling points are sorted according to the performance of the commodity selling points from high to low. Since the selling points of commodities can be directly corresponded to the commodities, the sorting result of the selling points of the commodities can be converted into the sorting results of the commodities. Therefore, a second type word can correspond to a product sorting result.

Since there are multiple second-type words in the first sorting result, multiple commodity sorting results can be obtained. After obtaining multiple commodity sorting results, if the multiple commodity sorting results are inconsistent, the sorting results of the second type words in the first sorting result are used as the main, and the multiple commodity sorting results are adjusted to obtain the final commodity sorting result. , the final product sorting result is the fifth recommendation list.

Exemplarily, the product selling point of product 1 is: quad-curved screen, 50 million main camera + 50 million super wide-angle lens + 64 million pixel periscope telephoto lens, 4800mAh battery, 66W super fast charge and a new generation of Snapdragon 8 mobile platform chip. The selling points of Commodity 2 are: front and rear symmetrical hyperbolic screen, 100-megapixel camera + 50-megapixel ultra-wide-angle lens, 4800mAh battery, 66W super fast charge, multi-mirror video camera replacement, Qualcomm Snapdragon™ 778G chip, etc. The selling points of the third product are: hyperboloid screen, 50 million main camera + 8 million ultra-wide-angle lens, 4000mAh battery, 66W super fast charge, Dimensity 1000+ chip.

The aforementioned first sorting result is camera>chip>battery. The second type word in the first sorting result is associated with the commodity selling points in the above-mentioned commodity 1, commodity 2, and commodity 3. For the camera in the first sorting result, the selling point of the related product 1 is: 50 million main camera + 50 million super wide-angle lens + 64 million pixel periscope telephoto lens; the selling point of the related product 2 is: : 100 million pixel camera + 50 million super wide-angle lens; the selling point of the related product 3 is: 50 million main camera + 8 million super wide-angle lens. Combining the product's selling point performance, the order of the products corresponding to the camera is: product two > product one > product three.

For the chips in the first sorting result, the selling point of the product associated with the first product is: the new generation of Snapdragon 8 mobile platform chip; the selling point of the product associated with the second product is: Qualcomm Snapdragon™ 778G chip; the product associated with the third product The selling point of the product is: Dimensity 1000+ chip. Combining the commodity selling point performance of the commodity, the order of commodities corresponding to the chip is: commodity 1 > commodity 2 > commodity 3.

For the batteries in the first sorting result, the selling point of the product associated with the first item is: 4800mAh battery + 66W super fast charge; the selling point of the product associated with the second item is: 4800mAh battery + 66W super fast charge; the product associated with the third item The selling point of the product is: 4000mAh battery + 66W super fast charge. Combining the product selling point performance of the product, the order of the products corresponding to the battery is: product one = product two > product three.

Combining the above, it can be seen that the order of commodities associated with cameras, chips and batteries is different. In this case, the sorting of each second-type word in the first sorting result is the main, and the commodity sorting associated with each second-type word is supplementary, and the multiple commodity sorting results are adjusted into one commodity sorting result, Thereby, a fifth recommendation list is obtained, and the order of commodities in the fifth recommendation list is: commodity 2 > commodity 1 > commodity 3.

A plurality of second type words are obtained by processing the posts involved in each section access record in the first section access record. Then, it is sorted according to the frequency of the second type of words to obtain the first sorting result. Next, fuzzy matching is performed between each second type word in the first sorting result and the commodity selling point in the commodity selling point database, and each second type word is associated with the commodity selling point of the commodity. performance to determine the ordering of items corresponding to each second type word. Finally, the final fifth recommendation list is generated based on the sorting of the second type words in the first sorting result, supplemented by the sorting of the commodities corresponding to each second type word. Subsequently, products may be recommended for new users according to the product information in the fifth recommendation list, or the fifth recommendation list may be combined with other recommendation lists to be recommended for new users.

Among them, 903-905 may be executed by the terminal device, and may also be executed by the server. If 903-905 are performed by the server, before 903, the method further includes: the server receives a request for obtaining a recommendation list from the terminal device, where the request for obtaining a recommendation list includes the model of the terminal device used by the new user, the model of the terminal device for the first model. After obtaining the fifth recommendation list through 905, the server sends the fifth recommendation list to the terminal device, and the terminal device may recommend commodities for the new user based on commodity information in the fifth recommendation list.

Another cold start recommendation method provided by the embodiment of the present application is described below by taking FIG. 10 as an example, and the method further includes:

1001. Determine the model of the terminal device used by the new user to access the first application.

See 601 for details.

1002. Acquire a commodity access record of the first application.

See 602 for details.

1003. Distinguish the commodity access records of the first application based on the model of the terminal device, and obtain a user behavior sequence corresponding to each model based on the differentiated commodity access records of the first application.

The commodity access records of the first application are distinguished according to the device models in the user device records. After distinguishing, for each device model, establish a user behavior sequence P, P=[ p1, p2, p3, …]. The user behavior sequence P may be determined according to the commodity access record of the first application corresponding to the device model within a preset time period. Among them, p1, p2, p3 represent the product particle size. Product granularity can be all product pools, as well as various levels of products within each product pool. For example, in the mobile phone product pool, p1, p2, p3 can be any one of a foldable mobile phone, a vertical mobile phone, a slide-type mobile phone, a rotary mobile phone, and the like.

1004. Determine a similar device model of the first model.

The similar device models of any device model can be determined by the similarity between the user behavior sequence P of the device and the user behavior sequences P of other devices. Specifically, it can be determined by calculating the cosine similarity between the user behavior sequence P of the device and the user behavior sequence P of another device.

Exemplarily, if the product of the similar device model to be determined is a mobile phone, the similarity calculation may be performed on the mobile phone of other models in the mall and the mobile phone of this model to obtain the similarity calculation result. After calculation, multiple similarity results can be obtained. Usually, the top 3 device models in the similarity results can be selected as similar device models.

Combining with 601, it can be seen that the device model used by the new user logging in the first application is the first model. Through similarity calculation, the similar device models of the first model can be obtained. The similar device models of the first model can include multiple, specifically. It can be selected according to the actual situation, which is not limited in this application.

1005. Determine a second product access record corresponding to the similar device model, and obtain a third recommendation list according to the second product access record.

The third recommendation list includes product information and a scoring result corresponding to the product.

Distinguish the commodity access records of the first application according to the device model to obtain a second commodity access record corresponding to a similar device model. The commodities involved in the second commodity access record are scored, and the scoring results of commodities corresponding to similar device models are obtained. For the specific scoring process, please refer to 604. Then, sort the items according to their ratings. Combining with 1004, it can be known that if the application selects 3 similar equipment models, the application can obtain 3 sorting results.

Then, based on the similarity calculation result between each similar device model and the first model, the three sorting results obtained above are sorted according to the size of the similarity calculation result, from high to low as a whole, to obtain a third recommendation list. Subsequently, products may be recommended for new users according to the product information in the third recommendation list and the scoring results corresponding to the products, or the third recommendation list may be combined with other recommendation lists to be recommended for new users.

Among them, 1003-1005 may be executed by the terminal device, and may also be executed by the server. If 1003-1005 are executed by the server, before 1003, the method further includes: the server receives a request for obtaining a recommendation list from the terminal device, where the request for obtaining a recommendation list includes the model of the terminal device used by the new user, the model of the terminal device for the first model. After obtaining the third recommendation list through 1005, the server sends the third recommendation list to the terminal device, and the terminal device recommends products for the new user based on the product information in the third recommendation list.

By mining the products preferred by users of similar device models, it is possible to break through the predicament of no valid data from a new aspect, and then recommend products of interest to users, thereby solving the cold start problem.

Another cold start recommendation method provided by the embodiment of the present application is described below by taking FIG. 11 as an example. The method includes four parts, which are the first part: the first recommendation list generation method; the second part: the second recommendation list, the first recommendation list The generation method of the fourth recommendation list and the fifth recommendation list; the third part: the generation method of the third recommendation list and the fourth part: the fusion process.

The first part: a method for generating a first recommendation list, specifically: when a new user logs in to the first application, the commodity access record of the first application can be obtained through the cloud server of the first application. Then, according to the model of the terminal device used by the new user, the product access records are distinguished, and the first product access record corresponding to the model (first model) of the terminal device used by the new user is obtained. Finally, according to the first product access record, Get the first recommendation list. For the detailed generation process of the first recommendation list, refer to 601-604, which will not be described in detail here.

The second part: the generation method of the second recommendation list, the fourth recommendation list, and the fifth recommendation list, specifically including the generation method of the second recommendation list, the generation method of the fourth recommendation list and the generation method of the fifth recommendation list.

The method for generating the second recommendation list is as follows: obtaining the block access records of the second application through the cloud server of the second application, and then distinguishing the block access records according to the model of the terminal device used by the new user to obtain the block access records used by the new user. The access record of the first section corresponding to the model of the terminal device (the first model), and then according to the access record of the first section, the user preference section and the user group corresponding to the user preference section are obtained. For example, the user preference sections are Section A, Section B, and Section C. Section A corresponds to user group a, section B corresponds to user group b, and section C corresponds to user group c. Then, according to the names of the users in the user group and the commodity access records, the commodities preferred by the users in the user group are determined, and finally a second recommendation list is generated according to the commodities preferred by the users. For the detailed process of generating the second recommendation list, refer to 701-705, which will not be described in detail here.

The method for generating the fourth recommendation list is as follows: first, obtain the section access records of the second application through the cloud server of the second application, and secondly, according to the model of the terminal device used by the new user, distinguish the section access records to obtain the section access records used by the new user. The first section access record corresponding to the model (first model) of the terminal device of words, and create a mapping between posts and first-type words. Then, according to the access records of the first section, the posts preferred by the user are determined. Then according to the mapping relationship, determine the first type word corresponding to the post preferred by the user, and finally use the fuzzy matching algorithm to determine the commodity information corresponding to the first type word, and generate a fourth recommendation list according to the commodity information corresponding to the first type word. For the detailed generation process of the fourth recommendation list, refer to 801-807, which will not be described in detail here.

The method for generating the fifth recommendation list is as follows: obtaining the block access records of the second application through the cloud server of the second application, and then distinguishing the block access records according to the model of the terminal device used by the new user to obtain the block access records used by the new user. The first block access record corresponding to the model of the terminal device (the first model), then according to the first block access record, the second type word is determined, and finally the selling point is associated according to the second type word, and a fifth recommendation list is generated according to the selling point. For the detailed generation process of the fifth recommendation list, refer to 901-905, which will not be described in detail here.

The third part: a method for generating a third recommendation list, specifically: acquiring the commodity access record of the first application through the cloud server of the first application. Then, based on the model of the terminal device, the product access records of the first application are distinguished, and based on the differentiated product access records of the first application, the user behavior sequence corresponding to each model is obtained, and then the similar device models of the first model are determined. , and then determine the second product access record corresponding to the similar device model, and obtain a third recommendation list according to the second product access record. For the detailed generation process of the third recommendation list, refer to 1001-1005, which will not be described in detail here.

The fourth part: the fusion process. Specifically: after obtaining the first recommendation list, the second recommendation list, the third recommendation list, the fourth recommendation list and the fifth recommendation list, make the first recommendation list, the second recommendation list, the third recommendation list, the fourth recommendation list and the fourth recommendation list. The product information in the list and the fifth recommendation list are fused to obtain the final recommendation list.

Exemplarily, the fusion process may be to select commodity information corresponding to some commodities from the first recommendation list, the second recommendation list, the third recommendation list, the fourth recommendation list and the fifth recommendation list according to a set ratio. Then, the selected commodities are sorted in the order that the first recommendation list comes first, followed by the second recommendation list, the third recommendation list, the fourth recommendation list, and the fifth recommendation list. The set ratio can be 3:2:1:1:1.

After sorting, the duplicate products will be deleted, and after deletion, the products will be filled in order from the corresponding recommendation list. Exemplarily, it is found that the product corresponding to the product information in the third recommendation list is consistent with the product corresponding to the product information in the first recommendation list, the product information in the third recommendation list is deleted, and then the deleted product information in the third recommendation list is deleted. The next product information of the lost product information is filled into the vacant position to generate the final recommendation list. The number of products in the final recommendation list is related to the parameter value passed in the front end. The front-end will pass the number of products to be recommended and displayed to the back-end in the form of recNum. After getting the final recommendation list, the back-end will intercept the recNum products in order and return them to the user.

Or, after obtaining the first recommendation list, the second recommendation list, the third recommendation list, the fourth recommendation list, and the fifth recommendation list, the fusion process may also be performed from the first recommendation list, the second recommendation list, and the third recommendation list. , the fourth recommendation list and the fifth recommendation list, two or more lists are arbitrarily selected for sorting and deduplication of commodity information to generate a final recommendation list. For example, the product information in the first recommendation list, the second recommendation list, the third recommendation list and the fourth recommendation list can be deduplicated and sorted to generate the final recommendation list, and the first recommendation list and the second recommendation list , the commodity information in the third recommendation list and the fifth recommendation list are deduplicated and sorted to generate the final recommendation list, or, as shown in FIG. The product information is deduplicated and sorted to generate the final recommendation list.

In order to realize the above-mentioned functions, the above-mentioned electronic device includes corresponding hardware structures and/or software modules for executing each function. Those skilled in the art should easily realize that, in conjunction with the units and algorithm steps of each example described in the embodiments disclosed herein, the embodiments of the present application can be implemented in hardware or a combination of hardware and computer software. Whether a function is performed by hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Skilled artisans may implement the described functionality using different methods for each particular application, but such implementations should not be considered beyond the scope of this application.

In this embodiment of the present application, the electronic device may be divided into functional modules according to the foregoing method examples. For example, each functional module may be divided corresponding to each function, or two or more functions may be integrated into one processing module. The above-mentioned integrated modules can be implemented in the form of hardware, and can also be implemented in the form of software function modules. It should be noted that, the division of modules in the embodiments of the present application is schematic, and is only a logical function division, and there may be other division manners in actual implementation.

Other embodiments of the present application provide an electronic device, and the electronic device may include: a communication module, a memory, and one or more processors. The communication module and memory are coupled to the processor. The memory is used to store computer program code comprising computer instructions.

Another embodiment of the present application provides a chip system. As shown in FIG. 13 , the chip system includes at least one processor 1301 and at least one interface circuit 1302 . The processor 1301 and the interface circuit 1302 may be interconnected by wires. For example, interface circuitry 1302 may be used to receive signals from other devices. As another example, the interface circuit 1302 may be used to send signals to other devices (eg, the processor 1301).

For example, the interface circuit 1302 may read instructions stored in memory in the device and send the instructions to the processor 1301 . When executed by the processor 1301, the instructions can cause the electronic device to perform each of the above-described embodiments. Certainly, the chip system may also include other discrete devices, which are not specifically limited in this embodiment of the present application.

Embodiments of the present application further provide a computer-readable storage medium, where the computer-readable storage medium includes computer instructions, when the computer instructions are executed on the electronic device, the electronic device is made to execute the electronic device (for example, a mobile phone) in the foregoing method embodiments. The various functions performed or.

Embodiments of the present application further provide a computer program product, which, when the computer program product runs on a computer, enables the computer to perform various functions or functions performed by an electronic device (eg, a mobile phone) in the foregoing method embodiments.

From the description of the above embodiments, those skilled in the art can clearly understand that, for the convenience and brevity of the description, only the division of the above functional modules is used as an example for illustration. In practical applications, the above functions can be allocated by Different functional modules are completed, that is, the internal structure of the device is divided into different functional modules, so as to complete all or part of the functions described above.

In the several embodiments provided in this application, it should be understood that the disclosed apparatus and method may be implemented in other manners. For example, the apparatus embodiments described above are only illustrative. For example, the division of modules or units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components may be combined or May be integrated into another device, or some features may be omitted, or not implemented. On the other hand, the shown or discussed mutual coupling or direct coupling or communication connection may be through some interfaces, indirect coupling or communication connection of devices or units, and may be in electrical, mechanical or other forms.

Units described as separate components may or may not be physically separated, and components shown as units may be one physical unit or multiple physical units, that is, may be located in one place, or may be distributed in multiple different places. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution in this embodiment.

In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit. The above-mentioned integrated units may be implemented in the form of hardware, or may be implemented in the form of software functional units.

The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a readable storage medium. Based on such understanding, the technical solutions of the embodiments of the present application can be embodied in the form of software products in essence, or the parts that contribute to the prior art, or all or part of the technical solutions, which are stored in a storage medium , including several instructions to cause a device (which may be a single chip, a chip, etc.) or a processor (processor) to execute all or part of the methods of the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a removable hard disk, a read only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk and other media that can store program codes.

The above contents are only specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any changes or substitutions within the technical scope disclosed in the present application should be covered within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims (18)

1. A cold start recommendation method is applied to a server and is characterized by comprising the following steps:

receiving a recommendation list acquisition request from a terminal device, wherein the recommendation list acquisition request comprises the model of the terminal device used by a new user, and the model of the terminal device is a first model;

determining a first commodity access record and a first version access record corresponding to a first model, wherein the first commodity access record is obtained by screening from a commodity access record of a first application, the first version access record is obtained by screening from a version access record of a second application, the first application is a shopping application, and the second application is a community application;

obtaining a first recommendation list according to the first commodity access record;

obtaining a second recommendation list according to the first version access record;

determining a similar model of the first model and a second commodity access record corresponding to the similar model, and obtaining a third recommendation list according to the second commodity access record, wherein the first recommendation list, the second recommendation list and the third recommendation list all comprise commodity information and a scoring result corresponding to commodities;

removing duplication and sorting commodity information in the first recommendation list, the second recommendation list and the third recommendation list to obtain a final recommendation list, wherein the final recommendation list is used for recommending commodities for a new user;

and sending the final recommendation list to the terminal equipment.

2. The method of claim 1, wherein after the deriving the third recommendation list, the method further comprises:

determining a first type word corresponding to the post related to each block access record in the first block access record based on the post related to each block access record in the first block access record, and creating a mapping relation between the post and the first type word, wherein the first type word is a word related to a product name in the post;

determining a post preferred by a user based on the first version access record;

determining a first type word corresponding to the post preferred by the user according to the mapping relation;

determining commodity information corresponding to the first type of words by using a fuzzy matching algorithm, and generating a fourth recommendation list according to the commodity information corresponding to the first type of words;

the removing duplication and sorting of the commodity information in the first recommendation list, the second recommendation list and the third recommendation list to obtain a final recommendation list includes:

and performing duplicate removal and sorting on the commodity information in the first recommendation list, the second recommendation list, the third recommendation list and the fourth recommendation list to obtain the final recommendation list.

3. The method of claim 1, wherein after the deriving the third recommendation list, the method further comprises:

determining a second type word corresponding to the post related to each block access record in the first block access record based on the post related to each block access record in the first block access record, wherein the second type word is a word related to product characteristics in the post;

sequencing according to the occurrence frequency of all the second type words to generate a first sequencing result;

generating a fifth recommendation list based on the commodity selling points associated with the second type words in the first sequencing result, wherein the fifth recommendation list comprises commodity information corresponding to the commodity selling points;

the removing duplication and sorting of the commodity information in the first recommendation list, the second recommendation list and the third recommendation list to obtain a final recommendation list includes:

removing duplication and sorting commodity information in the first recommendation list, the second recommendation list, the third recommendation list and the fifth recommendation list to obtain a final recommendation list;

or,

and performing duplicate removal and sorting on the commodity information in the first recommendation list, the second recommendation list, the third recommendation list, the fourth recommendation list and the fifth recommendation list to obtain the final recommendation list.

4. The method of claim 1, wherein deriving a first recommendation list based on the first item access record comprises:

scoring the commodities related to each commodity access record in the first commodity access record to obtain a scoring result of the commodities related to each commodity access record;

accumulating the scoring results of the same commodity to obtain the scoring results of all commodities;

and sorting the commodities from high to low according to the scoring results of all commodities to obtain a first recommendation list.

5. The method of claim 4, wherein the scoring the items involved in each item access record in the first item access record to obtain a scoring result of the items involved in each item access record comprises:

based on the first interaction behavior of the user and the commodities in each commodity access record, scoring the commodities related to each commodity access record to obtain a first scoring result;

based on the occurrence time of the first interaction behavior in each commodity access record, scoring the commodities related to each commodity access record to obtain a second scoring result;

and multiplying the first scoring result and the second scoring result to obtain the scoring result of the commodity related to each commodity access record.

6. The method according to claim 5, wherein the scoring the commodities related to each commodity access record based on the first interaction behavior of the user with the commodities in each commodity access record to obtain a first scoring result comprises:

and comparing the first interaction behaviors of the user and the commodities in each commodity access record with behaviors in a behavior scoring rule to obtain a first scoring result, wherein the behavior scoring rule comprises the behaviors and scoring results corresponding to the behaviors.

7. The method of claim 5, wherein the scoring the item involved in each item access record based on the occurrence time of the first interaction behavior in each item access record to obtain a second scoring result comprises:

and inputting the occurrence time of the first interactive behavior into a time attenuation function based on the occurrence time of the first interactive behavior in each commodity access record to obtain a second scoring result.

8. The method of claim 1, wherein deriving a second recommendation list based on the first version of the access record comprises:

determining a user preference block and a user group corresponding to the user preference block based on the first block access record;

and determining the commodities preferred by the users in the user group from the commodity access records based on the names of the users in the user group, and generating a second recommendation list based on the commodities preferred by the users.

9. The method of claim 8, wherein determining the user preference block and the user group corresponding to the user preference block based on the first version access record comprises:

and based on the second interaction behavior between the user and the layout in each layout access record in the first layout access record and the occurrence time of the second interaction behavior, scoring the layout related to each layout access record to obtain the user group corresponding to the user preference layout and the user preference layout.

10. The method of claim 1, wherein prior to said determining a similar model number of said first model, said method further comprises:

distinguishing the commodity access records according to equipment models, and obtaining a user behavior sequence corresponding to each equipment model based on the distinguished commodity access records, wherein the user behavior sequence comprises products accessed by users;

the determining a similar model of the first model includes:

determining the similarity between the user behavior sequence of the first model and the user behavior sequence of the second model, wherein when the similarity between the user behavior sequence of the first model and the user behavior sequence of the second model meets a threshold value, the second model is the similar model of the first model, and the second models are all models except the first model;

determining a second commodity access record corresponding to the similar model, and obtaining a third recommendation list according to the second commodity access record, wherein the third recommendation list comprises:

distinguishing the commodity access records according to equipment models to obtain second commodity access records corresponding to the similar models;

based on the second commodity access record, scoring the commodities related to the second commodity access record to obtain a commodity scoring result corresponding to the similar model;

and generating a third recommendation list based on the commodity scoring result.

11. The method according to claim 1 or 2, characterized in that the model of the terminal device used by the new user is determined to be the first model through an application program interface, API, exposed by the operating system of the terminal device.

12. The method according to claim 1 or 2, wherein the commodity access record of the first application and the block access record of the second application are obtained through page burial points, the page burial points comprise handwriting burial points and non-burial points, the handwriting burial points are located at preset detection positions, if data are detected, the data are reported, and the non-burial points are all data reported.

13. The method of claim 5 or 6, wherein the first interaction behavior of the user with the commodity comprises clicking into a commodity detail page, collecting the commodity, joining a shopping cart, clicking to purchase the commodity immediately, submitting a commodity purchase order, paying for the commodity order successfully, canceling the commodity collection, sharing the commodity, searching for the commodity, confirming that the commodity arrival notification is successful, consulting the commodity, commenting on the commodity, and clicking on the commodity in the browsing footprint.

14. The method of claim 9, wherein the second interaction behavior of the user with the layout comprises browsing of posts by the user under the layout, comments by the user under the posts, replies to comments by the user under the posts, sharing of posts by the user, favorite posts by the user, endorsement of posts by the user, detailed information of the user that clicks on the comment posts by the user, posts posted by the user that clicks on the comment posts, and appreciation of posts by the user.

15. A cold start recommendation method is applied to terminal equipment and is characterized by comprising the following steps:

responding to an operation of a new user for logging in a first application, and sending a request for obtaining a recommendation list to a server, wherein the recommendation list request comprises the model of terminal equipment used by the new user, and the model of the terminal equipment is a first model;

receiving a final recommendation list sent by the server;

and recommending commodities for the new user according to the final recommendation list.

16. A server, characterized in that the server comprises: a wireless communication module, memory, and one or more processors; the wireless communication module, the memory and the processor are coupled;

wherein the memory is to store computer program code comprising computer instructions; the computer instructions, when executed by the processor, cause the server to perform the method of any of claims 1-14.

17. A terminal device, characterized in that the terminal device comprises: a wireless communication module, memory, and one or more processors; the wireless communication module, the memory and the processor are coupled;

wherein the memory is to store computer program code comprising computer instructions; the computer instructions, when executed by the processor, cause the terminal device to perform the method of claim 15.

18. A computer-readable storage medium comprising computer instructions;

the computer instructions, when executed on a terminal device, cause the server to perform the method of any of claims 1-14, cause the terminal device to perform the method of claim 15.

Images