Project recommendation method and apparatus, computer device, and computer-readable storage medium

Abstract

The application discloses a project recommendation method and device, computer equipment and a computer readable storage medium, and relates to the technical field of artificial intelligence and digital medical treatment. The method is based on tags, uses a server to calculate the matching degree between each selectable project and a user, recommends a project with high adaptation, and enables the user to select the project by himself / herself without relying on service personnel, avoids deviation of manual recommendation, and realizes accurate information recommendation. The method comprises the following steps: in response to a project recommendation instruction, determining a user to be recommended indicated by the project recommendation instruction; matching a plurality of user tags of the user to be recommended according to historical operation records and basic user information of the user to be recommended; obtaining a plurality of preset recommendation projects, calculating the feature similarity between the plurality of user tags and each preset recommendation project in the plurality of preset recommendation projects, and extracting a target recommendation project from the plurality of preset recommendation projects according to the feature similarity; and based on the target recommendation project, recommending a project to the user to be recommended.

Description

Technical Field

[0001] This application relates to the fields of artificial intelligence and digital healthcare technology, and in particular to project recommendation methods, apparatus, computer equipment and computer-readable storage media. Background Technology

[0002] In recent years, with people paying increasing attention to their health, annual health checkups have become essential. A health checkup is a medical examination conducted to assess an individual's health status, identify early signs of disease and potential health risks. Early health checkups can help detect diseases in time, avoiding the serious consequences of delayed diagnosis.

[0003] When users have a need for a physical examination, because each user's physical condition, lifestyle, and medical history are different, they need to go to the hospital or medical examination institution in advance to choose the examination items that suit them. The applicant recognizes that currently, the physical examination items offered by hospitals or medical examination institutions are presented to users in the form of packages. Each package includes several pre-set fixed examination items, but users actually lack a deep understanding of each item, making it difficult to choose a package that matches their own situation from a large number of options. This necessitates hospitals or medical examination institutions to deploy more staff to explain and recommend items to users. However, in reality, users' cognitive levels vary, and they cannot accurately describe their own physical condition, which can easily lead to inaccurate and inaccurate recommendations, making it difficult to effectively recommend items that truly reflect their needs. Summary of the Invention

[0004] In view of this, this application provides a project recommendation method, apparatus, computer equipment, and computer-readable storage medium, the main purpose of which is to solve the problem that users' cognitive levels vary, they cannot accurately describe their physical condition, which easily leads to deviations in the recommended physical examination packages, low accuracy, and difficulty in effectively reflecting the project recommendations.

[0005] According to the first aspect of this application, a project recommendation method is provided, the method comprising:

[0006] In response to a project recommendation instruction, determine the user to be recommended as indicated by the project recommendation instruction;

[0007] Based on the historical operation records and basic user information of the user to be recommended, match multiple user tags of the user to be recommended;

[0008] Obtain multiple preset recommendation items, calculate the feature similarity between the multiple user tags and each preset recommendation item in the multiple preset recommendation items, and extract the target recommendation item from the multiple preset recommendation items according to the feature similarity;

[0009] Based on the target recommended items, recommended items are made to the user to be recommended.

[0010] Optionally, the step of matching multiple user tags of the user to be recommended based on the user's historical operation records and basic user information includes:

[0011] Match the historical operation records of the user to be recommended, and extract at least one attribute tag from the historical operation records. The at least one attribute tag is obtained by extracting attributes of the content that interacts with the historical operation behavior of the user to be recommended, as indicated by the historical operation records.

[0012] Query the basic user information of the user to be recommended, and extract at least one status tag from the basic user information. The at least one status tag is a user attribute indicated by the basic user information and / or an original tag bound to the basic user information.

[0013] The at least one attribute label and the at least one status label are used as the plurality of user labels.

[0014] Optionally, extracting at least one attribute tag from the historical operation record includes:

[0015] The historical browsing records are queried from the user's browsing history, and the object attributes of the browsing history objects are extracted as at least one attribute tag. The browsing history objects are one or more of the following: symptoms, departments, and medications; and / or,

[0016] In the historical operation record, query multiple historical operation objects related to the historical operation behavior of the user to be recommended, and use the object attributes of the multiple historical operation objects as the at least one attribute tag. The historical operation behavior includes one or more of the following: commenting, collecting, liking, watching, browsing, clicking, adding to cart, and placing an order.

[0017] Optionally, after querying the basic user information of the user to be recommended and extracting at least one status tag from the basic user information, the method further includes:

[0018] A plurality of preset sample labels are obtained, and the following processing is performed on each of the at least one attribute label and the at least one state label: the plurality of preset sample labels and the label are vectorized to obtain a plurality of sample label vectors of the plurality of preset sample labels and a specified label vector of the label; the cosine similarity between each sample label vector of the plurality of sample label vectors and the specified label vector is calculated respectively; and at least one candidate preset sample label with a cosine similarity higher than a first similarity threshold is extracted from the plurality of preset sample labels.

[0019] Obtain at least one candidate preset sample label for each of the at least one attribute label and the at least one status label to obtain multiple candidate preset sample labels;

[0020] Push the multiple candidate preset sample tags to the user to be recommended, and determine the target preset sample tag triggered by the user to be recommended from the multiple candidate preset sample tags;

[0021] The at least one attribute label, the at least one status label, and the target preset sample label are used as the multiple user labels.

[0022] Optionally, after querying the basic user information of the user to be recommended and extracting at least one status tag from the basic user information, the method further includes:

[0023] Obtain a preset display strategy, and determine the first push label indicated by the preset display strategy and at least one associated label of the first push label from the at least one attribute label and the at least one status label;

[0024] Push the first push tag and the at least one associated tag to the user to be recommended, and obtain the specified associated tag triggered by the user to be recommended in the at least one associated tag;

[0025] Obtain at least one secondary associated tag from the remaining tags that is associated with the specified associated tag as indicated by the preset display strategy, wherein the remaining tags are the tags other than the first push tag and the at least one associated tag among the at least one attribute tag and the at least one status tag;

[0026] Push the at least one secondary association tag to the user to be recommended, and continue to determine the secondary association tag triggered by the user to be recommended from the at least one secondary association tag, until all tags in the at least one attribute tag and the at least one status tag are traversed;

[0027] The tags triggered by the user to be pushed to are determined as the multiple user tags.

[0028] Optionally, the step of obtaining multiple preset recommendation items, calculating the feature similarity between the multiple user tags and each preset recommendation item in the multiple preset recommendation items, and extracting the target recommendation item from the multiple preset recommendation items includes:

[0029] Obtain the tag features corresponding to each user tag among the multiple user tags to obtain multiple user tag features, and perform vector transformation on the multiple user tag features to obtain user feature vectors;

[0030] Query the project features of each of the multiple preset recommendation projects, and perform vector transformation on the project features of each preset recommendation project to obtain multiple project feature vectors of the multiple preset recommendation projects;

[0031] Calculate the vector distance between the user feature vector and each of the multiple project feature vectors, and query the feature similarity of the vector distance mapping to obtain the feature similarity between the user feature vector and each project feature vector.

[0032] The feature vectors of the multiple items are sorted in descending order of feature similarity to obtain the sorting result;

[0033] Extract the feature vectors of a specified number of items that rank first in the sorting results, and determine the specified number of preset recommended items indicated by the specified number of item feature vectors.

[0034] Query the project category corresponding to each of the specified number of preset recommended projects to obtain the specified number of project categories;

[0035] Calculate the frequency of occurrence of each item category in the specified number of item categories, and determine the target item category with the highest frequency of occurrence in the specified number of item categories;

[0036] Extract the preset recommended items belonging to the target item category from the specified number of preset recommended items and use them as the target recommended items.

[0037] Optionally, the step of recommending items to the user to be recommended based on the target recommended items includes:

[0038] Generate a first recommendation page including the target recommended items, and push the first recommendation page to the user to be recommended; or,

[0039] The similarity between the target recommended item and each of the multiple preset recommended items is queried. Multiple candidate recommended items with a similarity greater than a second similarity threshold are extracted from the multiple preset recommended items. A second recommended page including the target recommended item and the multiple candidate recommended items is generated, and the second recommended page is pushed to the user to be recommended.

[0040] According to a second aspect of this application, a project recommendation device is provided, the device comprising:

[0041] A determination module is used to determine the user to be recommended as indicated by the project recommendation instruction in response to the project recommendation instruction;

[0042] The matching module is used to match multiple user tags of the user to be recommended based on the user's historical operation records and basic user information.

[0043] The extraction module is used to acquire multiple preset recommendation items, calculate the feature similarity between the multiple user tags and each preset recommendation item in the multiple preset recommendation items, and extract the target recommendation item from the multiple preset recommendation items according to the feature similarity.

[0044] The recommendation module is used to recommend projects to the user to be recommended based on the target recommendation projects.

[0045] Optionally, the matching module is configured to match the historical operation records of the user to be recommended, extract at least one attribute tag from the historical operation records, wherein the at least one attribute tag is obtained by extracting attributes of content that interacts with the historical operation behavior of the user to be recommended as indicated by the historical operation records; query the basic user information of the user to be recommended, extract at least one status tag from the basic user information, wherein the at least one status tag is a user attribute indicated by the basic user information and / or an original tag bound to the basic user information; and use the at least one attribute tag and the at least one status tag as the plurality of user tags.

[0046] Optionally, the matching module is configured to query multiple historical browsing objects browsed by the user to be recommended in the historical operation record, and extract the object attributes of the multiple historical browsing objects as the at least one attribute tag, wherein the historical browsing objects are one or more of symptoms, departments, and medicines; and / or, query multiple historical operation objects related to the historical operation behavior of the user to be recommended in the historical operation record, and use the object attributes of the multiple historical operation objects as the at least one attribute tag, wherein the historical operation behavior includes one or more of commenting, collecting, liking, watching, browsing, clicking, adding to cart, and placing an order.

[0047] Optionally, the matching module is further configured to acquire multiple preset sample tags, and perform the following processing on each tag in the at least one attribute tag and the at least one status tag: perform vector transformation on the multiple preset sample tags and the tag to obtain multiple sample tag vectors of the multiple preset sample tags and a specified tag vector of the tag; calculate the cosine similarity between each sample tag vector in the multiple sample tag vectors and the specified tag vector; and extract at least one candidate preset sample tag from the multiple preset sample tags whose cosine similarity is higher than a first similarity threshold; acquire at least one candidate preset sample tag for each tag in the at least one attribute tag and the at least one status tag to obtain multiple candidate preset sample tags; push the multiple candidate preset sample tags to the user to be recommended, and determine the target preset sample tag triggered by the user to be recommended from the multiple candidate preset sample tags; and use the at least one attribute tag, the at least one status tag, and the target preset sample tag as the multiple user tags.

[0048] Optionally, the matching module is further configured to: obtain a preset display strategy; determine, from the at least one attribute tag and the at least one status tag, the first push tag indicated by the preset display strategy and at least one associated tag of the first push tag; push the first push tag and the at least one associated tag to the user to be recommended; and obtain a specified associated tag triggered by the user to be recommended in the at least one associated tag; obtain, from the remaining tags, at least one secondary associated tag that is associated with the specified associated tag indicated by the preset display strategy, wherein the remaining tags are other tags in the at least one attribute tag and the at least one status tag besides the first push tag and the at least one associated tag; push the at least one secondary associated tag to the user to be recommended; and continue to determine, from the at least one secondary associated tag, the secondary associated tag triggered by the user to be recommended, until all tags in the at least one attribute tag and the at least one status tag are traversed; and determine the tag triggered by the user to be recommended as the plurality of user tags.

[0049] Optionally, the extraction module is configured to: acquire tag features corresponding to each user tag among the plurality of user tags, obtain plurality of user tag features, and perform vector transformation on the plurality of user tag features to obtain user feature vectors; query the item features of each preset recommendation item among the plurality of preset recommendation items, and perform vector transformation on the item features of each preset recommendation item to obtain plurality of item feature vectors of the plurality of preset recommendation items; calculate the vector distance between the user feature vector and each item feature vector among the plurality of item feature vectors, and query the feature similarity of the vector distance mapping to obtain the feature similarity between the user feature vector and each item feature vector; and sort the feature similarity by... The feature vectors of the multiple projects are sorted from high to low to obtain a sorting result; a specified number of project feature vectors at the top of the sorting result are extracted to determine a specified number of preset recommended projects indicated by the specified number of project feature vectors; the project categories corresponding to each preset recommended project in the specified number of preset recommended projects are queried to obtain a specified number of project categories; the frequency of occurrence of each project category in the specified number of project categories is counted, and the target project category with the highest frequency of occurrence is determined in the specified number of project categories; preset recommended projects belonging to the target project category are extracted from the specified number of preset recommended projects as the target recommended projects.

[0050] Optionally, the recommendation module is configured to generate a first recommendation page including the target recommendation item and push the first recommendation page to the user to be recommended; or, query the item similarity between the target recommendation item and each of the plurality of preset recommendation items, extract a plurality of candidate recommendation items whose item similarity is greater than a second similarity threshold from the plurality of preset recommendation items, generate a second recommendation page including the target recommendation item and the plurality of candidate recommendation items, and push the second recommendation page to the user to be recommended.

[0051] According to a third aspect of this application, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described in any of the first aspects above.

[0052] According to a fourth aspect of this application, a computer-readable storage medium is provided that stores a computer program thereon, which, when executed by a processor, implements the steps of the method described in any one of the first aspects above.

[0053] By employing the above technical solutions, this application provides a project recommendation method, apparatus, computer equipment, and computer-readable storage medium. Responding to a project recommendation instruction, this application determines the user to be recommended as indicated by the instruction, matches multiple user tags of the user to be recommended based on the user's historical operation records and basic user information, obtains multiple preset recommendation items, calculates the feature similarity between the multiple user tags and each preset recommendation item, extracts a target recommendation item from the multiple preset recommendation items according to the feature similarity, and recommends projects to the user based on the target recommendation item. Using user tags as a basis, the server's computing power is used to calculate the matching degree between each available item and the user, recommending items with high compatibility to the user for reference. This allows users to choose the items they want to purchase based on the content pushed by the platform, without relying on service personnel, avoiding biases caused by manual recommendations, and achieving accurate and effective information recommendation.

[0054] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0055] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0056] Figure 1 This illustration shows a schematic flowchart of a project recommendation method provided in an embodiment of this application;

[0057] Figure 2A This paper illustrates a schematic diagram of another project recommendation method provided in an embodiment of this application.

[0058] Figure 2B This illustration shows a schematic flowchart of a project recommendation method provided in an embodiment of this application;

[0059] Figure 3 This paper shows a schematic diagram of the structure of a project recommendation device provided in an embodiment of this application;

[0060] Figure 4 A schematic diagram of the device structure of a computer device provided in an embodiment of this application is shown. Detailed Implementation

[0061] Exemplary embodiments of the present application will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this application will be thorough and complete, and will fully convey the scope of the present application to those skilled in the art.

[0062] This application provides a project recommendation method, such as... Figure 1 As shown, the method includes:

[0063] 101. In response to the project recommendation instruction, determine the user to be recommended as indicated by the project recommendation instruction.

[0064] This application can be applied to service platforms that provide project purchase services, such as physical examination platforms, medical institution platforms, etc.; the projects recommended to users in this application can be combined projects such as physical examination packages or examination packages, or individual projects such as teeth cleaning projects or gastroscopy projects. This application does not make specific limitations on this.

[0065] Service platforms typically feature front-end applications for users, allowing them to view and purchase available services such as health check packages and beauty packages. Since these platforms contain a vast array of services, and users generally prefer to quickly find and purchase services that suit their needs, this embodiment of the application, upon receiving a project recommendation instruction from the front-end application, identifies the user to whom the instruction is directed. The service platform then recommends projects tailored to the user's specific circumstances, enabling targeted viewing. In practice, the front-end application can define specific projects associated with recommendation functionality, such as health check packages or beauty packages. When the front-end application detects a user searching for these specific projects, accessing their relevant pages, or requesting guidance from the front-end application on a navigation page, it sends a project recommendation instruction to the service platform, prompting the platform to execute the recommendation process. Specifically, the front-end application can include unique user information such as user ID, account number, or username in the recommendation instruction, allowing the service platform to directly identify the user to be recommended.

[0066] 102. Based on the historical operation records and basic user information of the users to be recommended, match multiple user tags of the users to be recommended.

[0067] To ensure that the recommended projects are of interest to the user and match their own situation, the service platform will match multiple user tags for the user based on the user's historical operation records and basic user information.

[0068] Historical operation records are used to determine which content the user to be recommended is interested in. For example, for a service platform providing physical examination services, records such as the user's consultations, browsing, clicks, purchases, and favorites can all be used as historical operation records. Basic user information is used to determine the content that matches the user to be recommended. For example, information such as age, gender, and region provided by the user when registering on the service platform, or the user's electronic medical records, can all be used as the user's basic user information. Since the historical operation records and basic user information contain a large amount of information, analyzing them one by one would involve a significant amount of work. Therefore, in this embodiment, feature extraction is performed on the information, and several simple user tags are used to indicate the characteristics of the historical operation records and basic user information. These user tags are then used to push items suitable for the user's own situation to the user to be recommended.

[0069] 103. Obtain multiple preset recommendation items, calculate the feature similarity between multiple user tags and each preset recommendation item in the multiple preset recommendation items, and extract the target recommendation item from the multiple preset recommendation items according to the feature similarity.

[0070] The service platform prepares a number of pre-selected items for users to choose from. After determining the user tags for the user to be recommended to, the service platform obtains the pre-selected items and determines which of these items is most suitable for the user based on the user tags. The user tags essentially describe the user's actual situation, and each pre-selected item also has its own characteristics. Items with higher similarity to the user's actual situation are more suitable and more likely to be selected by the user. Therefore, when evaluating which pre-selected item to recommend to the user, the service platform calculates the feature similarity between the user tags and each pre-selected item, and extracts the target item from the pre-selected items based on feature similarity. The platform then makes recommendations to the user based on the target item.

[0071] 104. Based on the target recommendation items, recommend items to the users to be recommended.

[0072] When recommending projects to potential users based on a target project, the platform can directly present the target project to the user, allowing them to view its details and decide whether to purchase it. Alternatively, the platform can select similar projects, additional projects purchased by other users of the target project, and recommend these alongside the target project to the user, providing more options.

[0073] The method provided in this application, in response to a project recommendation instruction, determines the user to be recommended as indicated by the instruction, matches multiple user tags of the user to be recommended based on the user's historical operation records and basic user information, obtains multiple preset recommendation items, calculates the feature similarity between the multiple user tags and each preset recommendation item in the multiple preset recommendation items, extracts a target recommendation item from the multiple preset recommendation items according to the feature similarity, and recommends items to the user based on the target recommendation item. Based on the user tags, the method uses the server's computing power to calculate the matching degree between each available item and the user, recommending items with high compatibility with the user for reference. This allows users to choose the items they want to buy based on the content pushed by the platform, without relying on service personnel, avoiding the bias of manual recommendations, and achieving accurate and effective information recommendation.

[0074] Furthermore, as a refinement and extension of the specific implementation methods of the above embodiments, and in order to fully illustrate the specific implementation process of this embodiment, this application provides another project recommendation method, such as... Figure 2A As shown, the method includes:

[0075] 201. In response to the project recommendation instruction, determine the user to be recommended as indicated by the project recommendation instruction.

[0076] The applicant recognizes that many service platforms currently offer standardized packages, such as health check-up packages and beauty packages, with all contents pre-configured by the platform. However, users on these platforms have varying levels of understanding of each package, making it difficult to find the most suitable option from a large pool of choices. Therefore, this application proposes a package recommendation method whereby the service platform intelligently selects suitable packages for each user based on their individual circumstances, ensuring that the recommended packages are appropriate for the user and easily accepted and appreciated.

[0077] The service platform provides a front-end application for users, allowing them to browse various services offered on the platform. The front-end application can be configured with recommended services, such as health checkups or beauty treatments. When the front-end application detects a user searching for these services, entering their relevant pages, or being prompted to select a service on a navigation page, it sends a service platform recommendation request. This prompts the service platform to execute the recommendation process. Specifically, the front-end application can include unique user information such as user ID, account number, or username in the recommendation request, enabling the service platform to directly identify the user to be recommended to.

[0078] 202. Match the historical operation records of the user to be recommended, and extract at least one attribute tag from the historical operation records.

[0079] In this embodiment, to ensure that the recommended items for the user are of interest to the user and match the user's own situation, the service platform matches multiple user tags for the user based on the user's historical operation records and basic user information. User tags may include attribute tags, which are obtained by extracting attributes from the content that interacts with the user's historical operation behavior as indicated by the historical operation records. The content that interacts with the user may include content viewed, favorited, or commented on by the user. The process of extracting attribute tags is described below:

[0080] On one hand, we can examine what content the user to be recommended has previously viewed and use the attributes of this content as attribute tags. This allows us to recommend content similar to the user's historical activity records. Specifically, we can query the user's historical browsing history for multiple browsing objects and extract their attributes as at least one attribute tag. These browsing objects can be one or more of the following: symptoms, departments, or medications. For example, if the user to be recommended has previously viewed gastroscopy, and the attribute of gastroscopy is gastroenterology, then gastroenterology can be used as an attribute tag for the user to be recommended.

[0081] Alternatively, one could examine which objects the user to be recommended has interacted with based on actions such as commenting and saving, and use the attributes of these objects as attribute tags. That is, by querying the historical operation records for multiple historical operation objects related to the user's historical behavior, and using the object attributes of these historical operation objects as at least one attribute tag, where historical operation behavior includes one or more of the following: commenting, saving, liking, watching, browsing, clicking, adding to cart, and placing an order. For example, if the user to be recommended has saved an article about meningitis, and the article's attribute is neurology, then neurology could be used as an attribute tag for the user to be recommended.

[0082] It should be noted that in practical applications, both of the above methods can be used simultaneously to extract attribute tags for users to be recommended, or only one method can be used to extract attribute tags, or the method used to extract attribute tags can be chosen based on the content of the user's historical operation records. This application does not impose any specific limitations on this.

[0083] 203. Query the basic user information of the user to be recommended, and extract at least one status tag from the basic user information.

[0084] In this embodiment, after obtaining at least one attribute tag based on the past behavior of the user to be recommended, the service platform also considers the basic information of the user to be recommended, that is, querying the basic user information of the user to be recommended and extracting at least one status tag from the basic user information. The at least one status tag is a user attribute indicated by the basic user information and / or an original tag bound to the basic user information, such as user age, user gender, and region. Furthermore, since some medical institutions' service platforms are linked to users' electronic medical records, which clearly indicate the user's basic physical condition, for service platforms providing physical examinations, tags extracted from electronic medical records are more consistent with the user's actual physical condition. Therefore, in this embodiment, the original tags bound to diseases involved in the electronic medical records can also be used as status tags. For example, if a user's electronic medical record indicates that they have fatty liver, then the original tag "gastroenterology" set by the service platform for fatty liver can be used as a status tag for that user. This application does not limit the specific method of obtaining status tags.

[0085] 204. Use at least one attribute label and at least one status label as multiple user labels.

[0086] In this embodiment of the application, after obtaining at least one attribute tag and at least one status tag, the service platform will use these tags as multiple user tags for the user to be recommended, and subsequently rely on these user tags to complete the project recommendation operation.

[0087] In an alternative implementation, since the service platform also sets up a series of preset sample tags, these preset sample tags are comprehensive and there are similar relationships between the tags. For example, the gastroenterology department and the gastroenterology department are actually related, making the similarity of the tags corresponding to the two departments very high. Therefore, after determining multiple user tags, considering that the users to be recommended also have a certain awareness of their current state, and the accuracy of their self-awareness is quite high, the service platform can also determine some candidate preset sample tags similar to the user tags. These candidate preset sample tags are first pushed to the user for reference and selection, so that the user's awareness of their current state is reflected in the tags they choose, for the service platform to combine and refer to, thereby ensuring that the recommended items are more in line with the user's current actual situation. The specific process of pushing candidate preset sample tags to the user is as follows: First, the service platform will obtain multiple preset sample tags and perform the following processing on each of at least one attribute tag and at least one state tag: Perform vector transformation on the multiple preset sample tags and tags to obtain multiple sample tag vectors of multiple preset sample tags and a specified tag vector of the tag. Calculate the cosine similarity between each sample tag vector in the multiple sample tag vectors and the specified tag vector, and extract at least one candidate preset sample tag from the multiple preset sample tags whose cosine similarity is higher than the first similarity threshold. After processing each of the at least one attribute tag and at least one status tag as described above, the service platform obtains multiple candidate preset sample tags and pushes these tags to the users to be recommended. The front-end application displays these candidate preset sample tags to the users, who then select tags that interest them or that match their actual situation based on their own judgment. The front-end application then feeds this information back to the service platform. In this way, the service platform can determine the target preset sample tag triggered by the user from among the multiple candidate preset sample tags, using at least one attribute tag, at least one status tag, and the target preset sample tag as multiple user tags. This allows the platform to subsequently consider the tags selected by the user in the project recommendation process.

[0088] In another optional implementation, the number of user tags obtained after performing steps 202 and 203 may be large. Considering that some tags may not match the current situation of the user to be recommended, the service platform will also filter the tags as needed, retaining only some accurate tags to ensure the accuracy of subsequent recommendations. The tag filtering can be done in a guided manner, that is, the service platform sets a preset display strategy, selects a priority tag and some other tags associated with that tag for the user to choose from according to the preset display strategy. After the user makes a selection, the secondary tags associated with the selected tag are displayed, and so on, until all currently determined tags are displayed. The tags selected by the user are used as reference tags for subsequent recommendations, and other unselected tags, i.e., tags that the user is not currently interested in or feels do not match their own situation, can be ignored. The specific process of tag filtering is described below: First, the service platform obtains the preset display strategy and determines the first push tag indicated by the preset display strategy and at least one associated tag of the first push tag from at least one attribute tag and at least one status tag. Subsequently, the service platform pushes the first push tag and at least one associated tag to the user to be recommended, and obtains the specified associated tags triggered by the user in at least one associated tag. Next, the service platform obtains at least one secondary associated tag from the remaining tags, which are related to the specified associated tag according to the preset display strategy. The remaining tags are the tags other than the first push tag and at least one associated tag from at least one attribute tag and at least one status tag. The platform then pushes at least one secondary associated tag to the user to be recommended, and continues to determine the secondary associated tags triggered by the user in at least one secondary associated tag, until all tags in at least one attribute tag and at least one status tag have been traversed. Finally, the service platform determines the tags triggered by the user to be recommended as multiple user tags. Through this triage method, the service platform guides the user to be recommended to describe their actual situation and what they want to know using tags. This not only transforms the user's thoughts into tags but also achieves the goal of simplifying tags, filtering out inaccurate tags to avoid affecting subsequent recommendation results, and further improving the accuracy of the recommended items.

[0089] 205. Obtain the tag features corresponding to each user tag from multiple user tags to obtain multiple user tag features, and perform vector transformation on the multiple user tag features to obtain user feature vectors. Query the project features of each preset recommended project from multiple preset recommended projects, and perform vector transformation on the project features of each preset recommended project to obtain multiple project feature vectors of multiple preset recommended projects. Calculate the vector distance between the user feature vector and each project feature vector among the multiple project feature vectors, and query the feature similarity of the vector distance mapping to obtain the feature similarity between the user feature vector and each project feature vector.

[0090] In this embodiment of the application, after obtaining multiple user tags, the system begins to use these user tags to evaluate multiple preset recommendation items currently connected to the platform, and determines which preset recommendation item is more similar to the actual situation of the user to be recommended and is more suitable for the user to be recommended.

[0091] When evaluating multiple preset recommended items, since user characteristics are represented by explicit tags, each tag is also associated with a set of features. Therefore, the service platform obtains the tag features corresponding to each user tag from multiple user tags, resulting in multiple user tag features. Because the similarity between two vectors can be expressed using the cosine value, meaning similarity can actually be calculated using vectorization, the service platform performs vectorization on the multiple user tag features to obtain user feature vectors. Similarly, for multiple preset recommended items, the service platform uses the same method to query the item features of each preset recommended item within the multiple preset recommended items, and performs vectorization on the item features of each preset recommended item, resulting in multiple item feature vectors for the multiple preset recommended items.

[0092] After the vector transformation is completed, the service platform begins to calculate the similarity. Specifically, the service platform calculates the vector distance between the user feature vector and each of the multiple item feature vectors, and queries the feature similarity of the vector distance mapping to obtain the feature similarity between the user feature vector and each item feature vector. The vector distance can be either Euclidean distance or Manhattan distance between vectors; this application does not impose specific limitations on this.

[0093] 206. Extract the target recommendation item from multiple preset recommendation items based on feature similarity.

[0094] In this embodiment, after determining the similarity between each preset recommended item and multiple user tags, the service platform extracts the target recommended item from the multiple preset recommended items according to feature similarity. In an optional implementation, the process of extracting the target recommended item is as follows:

[0095] First, the service platform sorts the feature vectors of multiple projects in descending order of feature similarity, obtaining a sorting result. Then, it extracts a specified number of feature vectors from the top-ranked project in the sorting result, determining the specified number of preset recommended projects indicated by these specified feature vectors. The specified number can be preset by the platform staff, for example, it can be set to 5, 10, 15, etc. This application does not limit the specific value of the specified number.

[0096] Next, the service platform queries the project categories corresponding to each of the specified number of preset recommended projects, obtaining the specified number of project categories. It then calculates the frequency of each project category within the specified number of project categories and identifies the target project category with the highest frequency among these categories. For example, assuming the specified number is 10, and the obtained project categories are neurology, gastroenterology, and gastroenterology, with neurology appearing 6 times (60% probability), gastroenterology appearing 3 times (30% probability), and gastroenterology appearing 1 time (10% probability), the target project category with the highest frequency is neurology.

[0097] Finally, the service platform will extract the target recommended items from a specified number of preset recommended items that belong to the target item category. For example, if the target item category is neurology, and the specified number of preset recommended items belonging to neurology are

Class A Imaging Examination Package

Class B Imaging Examination Package

Class C Lumbar Puncture Package

[0098] It should be noted that the calculation process in steps 205 to 206 above can actually be implemented based on the KNN (K-Nearest Neighbor) algorithm. KNN makes decisions based on the dominant category among k objects, rather than a single object category. In KNN, the distance between vectors is used as an indicator of dissimilarity between vectors, avoiding the matching problem between vectors. In fact, given the training set data and labels, KNN inputs test data, compares the features of the test data with the corresponding features in the training set, and finds the K most similar data in the training set. The category corresponding to the test data is then the category that appears most frequently among the K data. Therefore, through the KNN algorithm, the item category that best matches multiple user labels can be found, and items under that item category can be recommended to the user, thus ensuring the accuracy of the decision.

[0099] 207. Based on the target recommendation items, recommend items to the users to be recommended.

[0100] In this embodiment, after determining the target recommendation item, the service platform recommends the item to the user to be recommended. There are two methods for recommending items to the user: one is to directly recommend the target item to the user, i.e., generate a first recommendation page including the target item and push the first recommendation page to the user; the other method is to simultaneously recommend similar but not selected items to the user, i.e., query the similarity between the target item and each of multiple preset recommendation items, extract multiple candidate recommendation items with a similarity greater than a second similarity threshold from the multiple preset recommendation items, generate a second recommendation page including the target item and multiple candidate items, and push the second recommendation page to the user. This ensures that some unselected items that are very similar to the target item can also be seen by the user, providing them with more options.

[0101] The process described above uses the KNN algorithm to recommend items. In practical applications, to improve the accuracy of recommendations, a recommendation algorithm model can be built based on item attributes, user basic information, and user historical activity records. Item attributes can include textual descriptions of the item, disease tags, electronic medical records, etc. Electronic medical records can include information about diseases, departments, and medications previously treated by the patient. Historical activity records can include user comments, favorites, likes, views, browsing, clicks, add-to-cart actions, and purchases. By building this model, the content recommended to the user can be largely determined by the user's own attributes. For example, if a user has previously viewed articles or videos about colon cancer, consulted relevant doctors, or purchased related medications, when the user enters the health checkup page, the system can use a more specific tag-based filtering method to recommend a suitable package [A: Gastrointestinal Examination], or suggest that the user add some gastrointestinal-related items when ordering package [B].

[0102] The overall technical solution of this application is described below through an example: (e.g.) Figure 2BAs shown, the user to be recommended is User A. User A's historical activity records indicate that User A has browsed the digestive system section, consulted a gastroenterologist, and purchased gastrointestinal medications. User A's electronic medical record in their basic user information indicates that they have undergone a gastroscopy. Thus, the user tags obtained from the historical activity records and basic user information are

Gastrointestinal Diseases

Digestive System Disorders

Comprehensive Workplace Health Checkup Package

Digestive System Checkup Package

Neurology Specialist Checkup Package

Digestive System Checkup Package

[0103] The method provided in this application embodiment is based on user tags and utilizes the server's computing power to calculate the matching degree between each available item and the user. Items with a high degree of compatibility with the user are recommended to the user for reference, enabling the user to choose the items they want to purchase based on the content pushed by the platform, without relying on service personnel, avoiding the bias caused by manual recommendations, and achieving accurate and effective information recommendation.

[0104] Furthermore, as Figure 1 To specifically implement the method, this application provides a project recommendation device, such as... Figure 3 As shown, the device includes: a determination module 301, a matching module 302, an extraction module 303, and a recommendation module 304.

[0105] The determining module 301 is used to determine the user to be recommended indicated by the project recommendation instruction in response to the project recommendation instruction;

[0106] The matching module 302 is used to match multiple user tags of the user to be recommended based on the user's historical operation records and basic user information.

[0107] The extraction module 303 is used to acquire multiple preset recommendation items, calculate the feature similarity between the multiple user tags and each preset recommendation item in the multiple preset recommendation items, and extract the target recommendation item from the multiple preset recommendation items according to the feature similarity.

[0108] The recommendation module 304 is used to recommend projects to the user to be recommended based on the target recommendation project.

[0109] In a specific application scenario, the matching module 302 is used to match the historical operation records of the user to be recommended, extract at least one attribute tag from the historical operation records, wherein the at least one attribute tag is obtained by extracting attributes of the content that interacts with the historical operation behavior of the user to be recommended as indicated by the historical operation records; query the basic user information of the user to be recommended, extract at least one status tag from the basic user information, wherein the at least one status tag is the user attribute indicated by the basic user information and / or the original tag bound to the basic user information; and use the at least one attribute tag and the at least one status tag as the plurality of user tags.

[0110] In a specific application scenario, the matching module 302 is used to query multiple historical browsing objects browsed by the user to be recommended in the historical operation record, and extract the object attributes of the multiple historical browsing objects as the at least one attribute tag, wherein the historical browsing objects are one or more of symptoms, departments, and medicines; and / or, query multiple historical operation objects related to the historical operation behavior of the user to be recommended in the historical operation record, and use the object attributes of the multiple historical operation objects as the at least one attribute tag, wherein the historical operation behavior includes one or more of commenting, collecting, liking, watching, browsing, clicking, adding to cart, and placing an order.

[0111] In a specific application scenario, the matching module 302 is used for matching and also for acquiring multiple preset sample tags. For each tag in the at least one attribute tag and the at least one status tag, the following processing is performed: vector transformation is performed on the multiple preset sample tags and the tag to obtain multiple sample tag vectors of the multiple preset sample tags and a specified tag vector of the tag; the cosine similarity between each sample tag vector in the multiple sample tag vectors and the specified tag vector is calculated; at least one candidate preset sample tag with a cosine similarity higher than a first similarity threshold is extracted from the multiple preset sample tags; at least one candidate preset sample tag is acquired for each tag in the at least one attribute tag and the at least one status tag to obtain multiple candidate preset sample tags; the multiple candidate preset sample tags are pushed to the user to be recommended; and a target preset sample tag triggered by the user to be recommended is determined from the multiple candidate preset sample tags; the at least one attribute tag, the at least one status tag, and the target preset sample tag are used as the multiple user tags.

[0112] In a specific application scenario, the matching module 302 is used for matching modules and also for obtaining a preset display strategy; determining the first push tag indicated by the preset display strategy and at least one associated tag of the first push tag from the at least one attribute tag and the at least one status tag; pushing the first push tag and the at least one associated tag to the user to be recommended, and obtaining a specified associated tag triggered by the user to be recommended in the at least one associated tag; obtaining at least one secondary associated tag that is associated with the specified associated tag indicated by the preset display strategy from the remaining tags, wherein the remaining tags are other tags in the at least one attribute tag and the at least one status tag besides the first push tag and the at least one associated tag; pushing the at least one secondary associated tag to the user to be recommended, and continuing to determine the secondary associated tags triggered by the user to be recommended from the at least one secondary associated tags, until all tags in the at least one attribute tag and the at least one status tag are traversed; and determining the tag triggered by the user to be recommended as the plurality of user tags.

[0113] In a specific application scenario, the extraction module 303 is used to obtain the tag features corresponding to each user tag among the multiple user tags, to obtain multiple user tag features, and to perform vector transformation on the multiple user tag features to obtain user feature vectors; to query the project features of each preset recommendation project among the multiple preset recommendation projects, and to perform vector transformation on the project features of each preset recommendation project to obtain multiple project feature vectors of the multiple preset recommendation projects; to calculate the vector distance between the user feature vector and each project feature vector among the multiple project feature vectors, and to query the feature similarity of the vector distance mapping to obtain the feature similarity between the user feature vector and each project feature vector; and to perform feature extraction according to the features. The feature vectors of the multiple items are sorted in descending order of similarity to obtain a sorting result; a specified number of item feature vectors at the top of the sorting result are extracted to determine a specified number of preset recommended items indicated by the specified number of item feature vectors; the item category corresponding to each preset recommended item in the specified number of preset recommended items is queried to obtain a specified number of item categories; the frequency of occurrence of each item category in the specified number of item categories is counted, and the target item category with the highest frequency of occurrence is determined in the specified number of item categories; preset recommended items belonging to the target item category are extracted from the specified number of preset recommended items as the target recommended items.

[0114] In a specific application scenario, the recommendation module 304 is used to generate a first recommendation page including the target recommendation item and push the first recommendation page to the user to be recommended; or, query the item similarity between the target recommendation item and each of the plurality of preset recommendation items, extract a plurality of candidate recommendation items whose item similarity is greater than a second similarity threshold from the plurality of preset recommendation items, generate a second recommendation page including the target recommendation item and the plurality of candidate recommendation items, and push the second recommendation page to the user to be recommended.

[0115] The apparatus provided in this application embodiment, in response to a project recommendation instruction, determines the user to be recommended as indicated by the instruction, matches multiple user tags of the user to be recommended based on the user's historical operation records and basic user information, obtains multiple preset recommendation items, calculates the feature similarity between the multiple user tags and each preset recommendation item in the multiple preset recommendation items, extracts a target recommendation item from the multiple preset recommendation items according to the feature similarity, and recommends items to the user based on the target recommendation item. Based on the user tags, the apparatus uses the server's computing power to calculate the matching degree between each available item and the user, recommending items with high compatibility with the user for reference. This allows users to choose the items they want to purchase based on the content pushed by the platform, without relying on service personnel, avoiding the bias caused by manual recommendations, and achieving accurate and effective information recommendation.

[0116] It should be noted that other corresponding descriptions of the functional units involved in the project recommendation device provided in this application embodiment can be found in the following references. Figure 1 and Figures 2A to 2B The corresponding description in [the document] will not be repeated here.

[0117] In an exemplary embodiment, see Figure 4 The invention also provides a computer device including a bus, a processor, a memory, and a communication interface. It may also include an input / output interface and a display device, wherein the various functional units can communicate with each other via the bus. The memory stores a computer program, and the processor executes the program stored in the memory, performing the recommended method described in the above embodiments.

[0118] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the project recommendation method.

[0119] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented in hardware or by using software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) and includes several instructions to cause a computer device (such as a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0120] Those skilled in the art will understand that the accompanying drawings are merely schematic diagrams of a preferred embodiment, and the modules or processes shown in the drawings are not necessarily essential for implementing this application.

[0121] Those skilled in the art will understand that the modules in the apparatus of the implementation scenario can be distributed within the apparatus of the implementation scenario as described, or they can be located in one or more apparatuses different from this implementation scenario, with corresponding changes. The modules of the above-described implementation scenario can be combined into one module, or they can be further divided into multiple sub-modules.

[0122] The serial numbers in this application are for descriptive purposes only and do not represent the superiority or inferiority of the implementation scenario.

[0123] The above disclosures are only a few specific implementation scenarios of this application. However, this application is not limited to these. Any variations that can be conceived by those skilled in the art should fall within the protection scope of this application.

Claims

1. A project recommendation method, characterized in that, include: In response to a project recommendation instruction, the user to be recommended as indicated by the project recommendation instruction is determined. The project recommendation instruction is fed back to the service platform by the front-end application when it detects that the user is searching for a specified project, or enters the relevant page of the specified project, or when the user requests the front-end application to guide him to select a project on the guidance page. Based on the historical operation records and basic user information of the user to be recommended, multiple user tags of the user to be recommended are matched; wherein, the multiple user tags include at least one attribute tag and at least one status tag, or the multiple user tags include at least one attribute tag, at least one status tag, and a target preset sample tag; wherein, determining the target preset sample tag includes: obtaining multiple preset sample tags, and performing the following processing on each tag among the at least one attribute tag and the at least one status tag: performing vector transformation on each tag among the multiple preset sample tags, the at least one attribute tag, and the at least one status tag to obtain multiple sample tag vectors of the multiple preset sample tags, and the at least one... The algorithm calculates the cosine similarity between each of the multiple sample label vectors and the designated label vectors of each of the at least one attribute label and at least one state label, and extracts at least one candidate preset sample label from the multiple preset sample labels whose cosine similarity is higher than a first similarity threshold; obtains at least one candidate preset sample label for each of the at least one attribute label and at least one state label, resulting in multiple candidate preset sample labels; pushes the multiple candidate preset sample labels to the user to be recommended, and determines the target preset sample label triggered by the user to be recommended from the multiple candidate preset sample labels; The process involves acquiring multiple preset recommendation items, calculating the feature similarity between the multiple user tags and each preset recommendation item, and extracting target recommendation items from the multiple preset recommendation items based on feature similarity. Determining the target recommendation item includes: generating a user feature vector based on the multiple user tags; determining multiple item feature vectors based on the multiple preset recommendation items; calculating the vector distance between the user feature vector and each item feature vector; querying the feature similarity of the vector distance mapping to obtain the feature similarity between the user feature vector and each item feature vector; and then sorting the items according to feature similarity from high to low. The process involves sorting the feature vectors of the multiple projects to obtain a sorting result; extracting a specified number of project feature vectors that rank first from the sorting result to determine a specified number of preset recommended projects indicated by the specified number of project feature vectors; querying the project category corresponding to each preset recommended project in the specified number of preset recommended projects to obtain a specified number of project categories; calculating the frequency of occurrence of each project category in the specified number of project categories and determining the target project category with the highest frequency in the specified number of project categories; and extracting preset recommended projects belonging to the target project category from the specified number of preset recommended projects as the target recommended projects. Based on the target recommended items, recommended items are made to the user to be recommended.

2. The method according to claim 1, characterized in that, The step of matching multiple user tags of the user to be recommended based on the user's historical operation records and basic user information includes: Match the historical operation records of the user to be recommended, and extract at least one attribute tag from the historical operation records. The at least one attribute tag is obtained by extracting attributes of the content that interacts with the historical operation behavior of the user to be recommended, as indicated by the historical operation records. Query the basic user information of the user to be recommended, and extract at least one status tag from the basic user information. The at least one status tag is a user attribute indicated by the basic user information and / or an original tag bound to the basic user information. The at least one attribute label and the at least one status label are used as the plurality of user labels.

3. The method according to claim 2, characterized in that, Extracting at least one attribute tag from the historical operation record includes: The historical browsing records are queried from the user's browsing history, and the object attributes of the historical browsing objects are extracted as at least one attribute tag. The historical browsing objects are one or more of the following: symptoms, departments, and medications; and / or, In the historical operation record, query multiple historical operation objects related to the historical operation behavior of the user to be recommended, and use the object attributes of the multiple historical operation objects as the at least one attribute tag. The historical operation behavior includes one or more of the following: commenting, collecting, liking, watching, browsing, clicking, adding to cart, and placing an order.

4. The method according to claim 2, characterized in that, After querying the basic user information of the user to be recommended and extracting at least one status tag from the basic user information, the method further includes: Obtain a preset display strategy, and determine the first push label indicated by the preset display strategy and at least one associated label of the first push label from the at least one attribute label and the at least one status label; Push the first push tag and the at least one associated tag to the user to be recommended, and obtain the specified associated tag triggered by the user to be recommended in the at least one associated tag; Obtain at least one secondary associated tag from the remaining tags that is associated with the specified associated tag as indicated by the preset display strategy, wherein the remaining tags are the tags other than the first push tag and the at least one associated tag among the at least one attribute tag and the at least one status tag; Push the at least one secondary association tag to the user to be recommended, and continue to determine the secondary association tag triggered by the user to be recommended from the at least one secondary association tag, until all tags in the at least one attribute tag and the at least one status tag are traversed; The tags triggered by the users to be recommended are identified as the multiple user tags.

5. The method according to claim 1, characterized in that, The process of generating user feature vectors based on the multiple user tags and determining multiple item feature vectors based on the multiple preset recommendation items includes: Obtain the tag features corresponding to each user tag among the multiple user tags to obtain multiple user tag features, and perform vector transformation on the multiple user tag features to obtain user feature vectors; The project features of each of the multiple preset recommended projects are queried, and the project features of each preset recommended project are vectorized to obtain multiple project feature vectors of the multiple preset recommended projects.

6. The method according to claim 1, characterized in that, The step of recommending projects to the user to be recommended based on the target recommended projects includes: Generate a first recommendation page including the target recommended items, and push the first recommendation page to the user to be recommended; or, The similarity between the target recommended item and each of the multiple preset recommended items is queried. Multiple candidate recommended items with a similarity greater than a second similarity threshold are extracted from the multiple preset recommended items. A second recommended page including the target recommended item and the multiple candidate recommended items is generated, and the second recommended page is pushed to the user to be recommended.

7. A project recommendation device, characterized in that, include: The determination module is used to determine the user to be recommended in response to the project recommendation instruction. The project recommendation instruction is fed back to the service platform by the front-end application when it detects that the user is searching for a specified project, or enters the relevant page of the specified project, or when the user requests the front-end application to guide him to select a project on the guidance page. The matching module is used to match multiple user tags of the user to be recommended based on the user's historical operation records and basic user information; wherein, the multiple user tags include at least one attribute tag and at least one status tag, or the multiple user tags include at least one attribute tag, at least one status tag, and a target preset sample tag; wherein, determining the target preset sample tag includes: obtaining multiple preset sample tags, and performing the following processing on each tag among the at least one attribute tag and the at least one status tag: performing vector transformation on each tag among the multiple preset sample tags, the at least one attribute tag, and the at least one status tag to obtain multiple sample tag vectors of the multiple preset sample tags, and the target preset sample tag vectors. The system uses at least one attribute label and at least one status label to determine the specified label vector for each label. It calculates the cosine similarity between each sample label vector and the specified label vector of each label in the plurality of sample label vectors, and extracts at least one candidate preset sample label from the plurality of preset sample labels whose cosine similarity is higher than a first similarity threshold. It then obtains at least one candidate preset sample label for each label in the at least one attribute label and at least one status label, resulting in a plurality of candidate preset sample labels. Finally, it pushes the plurality of candidate preset sample labels to the user to be recommended, and determines the target preset sample label triggered by the user to be recommended from among the plurality of candidate preset sample labels. An extraction module is used to acquire multiple preset recommendation items, calculate the feature similarity between the multiple user tags and each preset recommendation item in the multiple preset recommendation items, and extract target recommendation items from the multiple preset recommendation items according to the feature similarity; wherein, the extraction module is used to generate user feature vectors based on the multiple user tags, determine multiple item feature vectors based on the multiple preset recommendation items, calculate the vector distance between the user feature vector and each item feature vector in the multiple item feature vectors respectively, and query the feature similarity of the vector distance mapping to obtain the feature similarity between the user feature vector and each item feature vector; and sort the target recommendation items according to the feature similarity from high to low. The multiple project feature vectors are sorted sequentially to obtain a sorting result; a specified number of project feature vectors at the top of the sorting result are extracted to determine a specified number of preset recommended projects indicated by the specified number of project feature vectors; the project categories corresponding to each preset recommended project in the specified number of preset recommended projects are queried to obtain a specified number of project categories; the frequency of occurrence of each project category in the specified number of project categories is counted, and the target project category with the highest frequency of occurrence is determined in the specified number of project categories; preset recommended projects belonging to the target project category are extracted from the specified number of preset recommended projects as the target recommended projects; The recommendation module is used to recommend projects to the user to be recommended based on the target recommendation projects.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images