Information recommendation method and device, electronic equipment and readable storage medium
By combining information characteristics, object characteristics and behavioral characteristics, and using deep neural networks and click-through rate prediction models to calculate the recommendation probability, the accuracy challenge of the information recommendation system in the case of information overload is solved, and the efficiency and quality of information matching are improved.
Patent Information
- Application Number
- CN202310531113.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-05-11
- Publication Date
- 2024-11-12
AI Technical Summary
Existing information recommendation systems have challenges in improving the accuracy of information recommendation, especially in dealing with information overload, which makes it difficult to effectively match user needs and product information.
By obtaining the information characteristics of the recommended object, the object characteristics and object behavior characteristics of the recommended object, the deep neural network and click-through rate prediction model calculate the target information vector, the target object vector and the predicted click-through rate, and then calculate the information to be recommended for the recommended object Recommendation probability, determine whether to recommend information.
Improve the accuracy of information recommendation, alleviate cold start and sparse problems, and ensure the quality and efficiency of information recommendation.
Smart Images

Figure CN118939865A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the field of data technology, in particular to the field of artificial intelligence technology such as deep learning and natural language processing. Provided are an information recommendation method, device, electronic device and readable storage medium. Background Art
[0002] With the development of the Internet, information of various types and platforms is growing explosively, and the problem of information overload is becoming increasingly serious. Recommendation systems have emerged as the times require. They are essentially information filtering systems that can improve the efficiency of users in obtaining effective information, thereby promoting transactions on platforms / merchants. How to improve the accuracy of information recommendations is a technical problem that needs to be solved urgently. Summary of the invention
[0003] According to a first aspect of the present disclosure, there is provided an information recommendation method, comprising: obtaining information features of information to be recommended, and object features and object behavior features of a recommended object; obtaining a target information vector and a target object vector based on the object features and the object behavior features; obtaining a predicted click-through rate based on the information features and the object features; and obtaining a recommendation probability of the information to be recommended corresponding to the recommended object based on the predicted click-through rate, the target information vector and the target object vector.
[0004] According to a second aspect of the present disclosure, an information recommendation device is provided, including: an acquisition unit, used to acquire information characteristics of the information to be recommended, and object characteristics and object behavior characteristics of the recommended object; a first processing unit, used to obtain a target information vector and a target object vector according to the object characteristics and the object behavior characteristics; a second processing unit, used to obtain a predicted click-through rate according to the information characteristics and the object characteristics; and a third processing unit, used to obtain a recommendation probability of the information to be recommended corresponding to the recommended object according to the predicted click-through rate, the target information vector and the target object vector.
[0005] According to a third aspect of the present disclosure, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method as described above.
[0006] According to a fourth aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause the computer to execute the method as described above.
[0007] According to a fifth aspect of the present disclosure, a computer program product is provided, comprising a computer program, wherein the computer program implements the method as described above when executed by a processor.
[0008] It can be seen from the above technical solutions that the present invention obtains the recommendation probability of the information to be recommended corresponding to the recommendation object by combining the information characteristics of the information to be recommended itself, the object characteristics of the recommendation object itself, and the object behavior characteristics of the recommendation object, and then determines whether to recommend the information to be recommended to the recommendation object based on the recommendation probability, which can improve the accuracy of the obtained recommendation probability, thereby achieving the purpose of more accurately recommending information to the recommendation object.
[0009] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS
[0010] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.
[0011] Figure 1 is a schematic diagram according to a first embodiment of the present disclosure;
[0012] Figure 2 is a schematic diagram according to a second embodiment of the present disclosure;
[0013] Figure 3 is a schematic diagram according to a third embodiment of the present disclosure;
[0014] Figure 4 is a schematic diagram according to a fourth embodiment of the present disclosure;
[0015] Figure 5 The block diagram is a block diagram of an electronic device used to implement the information recommendation method of the embodiment of the present disclosure. DETAILED DESCRIPTION
[0016] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, the description of well-known functions and mechanisms is omitted in the following description.
[0017] Figure 1 Schematic diagram of the first embodiment of the present disclosure. Figure 1 As shown, the information recommendation method of this embodiment specifically includes the following steps:
[0018] S101, obtaining information characteristics of the information to be recommended, and object characteristics and object behavior characteristics of the recommended object;
[0019] S102, obtaining a target information vector and a target object vector according to the object characteristics and the object behavior characteristics;
[0020] S103, obtaining a predicted click rate according to the information features and the object features;
[0021] S104: Obtain a recommendation probability of the information to be recommended corresponding to the recommended object according to the predicted click rate, the target information vector and the target object vector.
[0022] The information recommendation method of this embodiment combines the information characteristics of the information to be recommended itself, the object characteristics of the recommendation object itself, and the object behavior characteristics of the recommendation object, and obtains the recommendation probability of the recommendation object corresponding to the information to be recommended based on the obtained target information vector, target object vector and predicted click-through rate, and then determines whether to recommend the information to be recommended to the recommendation object based on the recommendation probability, which can improve the accuracy of the obtained recommendation probability, thereby achieving the purpose of more accurately recommending information to the recommendation object.
[0023] The information to be recommended in this embodiment is product demand information, which includes the product required by the user and the user's contact information. The product required by the user can be a physical product or a virtual product. The user's contact information can be a phone number, an email address, a social application account, etc. In this embodiment, the user's permission has been obtained for the use of the contact information. The recommended object in this embodiment is the enterprise or company that provides the product. For example, the recommended object in this embodiment can be enterprise A that provides physical product A, or company B that provides virtual product B.
[0024] The information recommendation method of this embodiment is used to determine the recommendation probability of the product demand information generated by the user (i.e., the information to be recommended) for the enterprise or company that provides the product, and then determine whether to recommend the product demand information to the recommendation object based on the recommendation probability. After the product demand information is recommended to the recommendation object, if the recommendation object purchases the product demand information, the specific content of the product demand information is further displayed to the recommendation object, so that the recommendation object can communicate with the corresponding user based on the contact information in the product demand information, and thus provide corresponding services to users with product needs more quickly and accurately.
[0025] For example, the information to be recommended in this embodiment may be "I want to buy an excavator, my contact information is xxx", and this information can be recommended to the excavator company so that the excavator company can communicate with the user based on the contact information in the information, thereby being more conducive to the user's purchase and the excavator company's service provision.
[0026] The information features of the information to be recommended obtained by executing S101 in this embodiment may include features such as product brand, industry to which the product belongs, and product description of the product required by the user in the information to be recommended.
[0027] When executing S101 to obtain information features of the information to be recommended, this embodiment may further obtain user portrait features such as age, gender, and occupation according to the contact information of the user in the information to be recommended as visitor features of the information to be recommended.
[0028] The object features of the recommended object obtained by executing S101 in this embodiment may include features such as the industry to which the recommended object belongs and the keywords of the promotion information placed on the recommended object.
[0029] The object behavior characteristics of the recommended object obtained by executing S101 in this embodiment include the purchase characteristics and search characteristics of the recommended object; wherein the purchase characteristics include the identification information of the product demand information purchased by the recommended object within a certain time range (for example, within a week, within a month, etc.), the product brand of the product required by the user in the purchased product demand information, the industry to which the product belongs, and other characteristics; the search characteristics include the search keywords entered by the recommended object when searching within a certain time range, the industry to which the search keywords belong, the geographical location at the time of search, and other characteristics.
[0030] After executing S101 to obtain information features, object features, object behavior features, and visitor features, this embodiment can also process the features into vectors according to a preset processing method. For example, text features are segmented, and the word vector of each word is obtained and then the average value is obtained, and the average value is normalized, so as to use the normalized result as the vector of text features; for discrete features such as gender, one-hot encoding can be used to vectorize the features; for continuous features such as age, the features can be vectorized by directly normalizing them.
[0031] It can be understood that this embodiment does not limit the number of recommended information and recommended objects, and can determine the recommendation probability that one piece of recommended information corresponds to one recommended object, can determine the recommendation probability that multiple pieces of recommended information correspond to one recommended object, can determine the recommendation probability that one piece of recommended information corresponds to multiple recommended objects, and can determine the recommendation probability that multiple pieces of recommended information correspond to multiple recommended objects.
[0032] After executing S101 to obtain information features of the information to be recommended and object features and object behavior features of the recommended object, this embodiment executes S102 to obtain a target information vector and a target object vector according to the object features and the object behavior features.
[0033] Specifically, when executing S102 in this embodiment to obtain the target information vector and the target object vector based on the object features and the object behavior features, the optional implementation method that can be adopted is: splicing the object features and the object behavior features, and inputting the splicing result into the first processing model; if the features are not vectors, the features can be converted into vectors before splicing; obtaining the output of the last activation layer in the first processing model as the target object vector; obtaining the parameters of the output layer in the first processing model as the target information vector.
[0034] Among them, the first processing model used in this embodiment when executing S102 is pre-trained. The first processing model is a collaborative filtering model based on DNN (Deep Neural Network), which is composed of an input layer, multiple activation layers (such as ReLU activation layer) and an output layer (Softmax layer).
[0035] The first processing model in this embodiment can be pre-trained in the following manner: obtain first training data, where the first training data includes multiple first feature pairs and labeled purchase probabilities of multiple first feature pairs, and each first feature pair includes sample object features and sample object behavior features; after splicing the sample object features and the sample object behavior features in the first feature pairs, input the splicing result into the first neural network model to obtain the predicted purchase probability output by the first neural network model; calculate the first loss function value according to the predicted purchase probability and the labeled purchase probability, and update the parameters of the first neural network model according to the first loss function value until the first neural network model converges to obtain the first processing model.
[0036] After executing S102 to obtain the target information vector and the target object vector, this embodiment executes S103 to obtain the predicted click rate according to the information features and the object features.
[0037] Specifically, when executing S103 in this embodiment to obtain the predicted click-through rate based on the information features and the object features, an optional implementation method that can be used is: splicing the information features and the object features, and inputting the splicing result into the second processing model. If the feature is not a vector, the feature is converted into a vector before splicing; the output result of the second processing model is used as the predicted click-through rate.
[0038] In addition, when executing S103 to splice information features with object features and input the splicing result into the second processing model, this embodiment may also include the following contents: obtaining visitor features of the information to be recommended; splicing the information features of the information to be recommended, the visitor features of the information to be recommended, and the object features of the recommended object, and inputting the splicing result into the second processing model.
[0039] That is to say, this embodiment can also combine the visitor characteristics of the information to be recommended to obtain the predicted click-through rate, thereby improving the accuracy of the obtained predicted click-through rate.
[0040] The second processing model used in this embodiment when executing S103 is pre-trained, and the second processing model may be a click rate prediction model based on DeepFM.
[0041] The second processing model in this embodiment can be pre-trained in the following manner: obtain second training data, where the second training data includes multiple second feature pairs and annotated click-through rates of multiple second feature pairs, each second feature pair includes sample information features and sample object features, and may also include sample visitor features; after splicing the sample information features and sample object features in the second feature pairs, input the splicing result into the second neural network model to obtain the predicted click-through rate output by the second neural network model; calculate the second loss function value based on the predicted click-through rate and the annotated click-through rate, and update the parameters of the second neural network model based on the second loss function value until the second neural network model converges to obtain the second processing model.
[0042] The second neural network model in this embodiment can also be jointly trained with the third neural network model to obtain the second processing model and the third processing model. The specific training process is described below.
[0043] This embodiment introduces a click-through rate prediction model to obtain a predicted click-through rate based on information features, object features, and visitor features. The above features can be fully crossed, so that the obtained predicted click-through rate is used to obtain the final recommendation probability, thereby achieving a higher quality prediction of the recommendation probability, and effectively alleviating the cold start and sparsity problems caused by the lack of interactive information between the information to be recommended and the recommended object.
[0044] After executing S103 to obtain the predicted click rate, this embodiment executes S104 to obtain the recommendation probability of the recommended object corresponding to the information to be recommended according to the predicted click rate, the target information vector and the target object vector.
[0045] After executing S104 to obtain the recommendation probability of the recommended object corresponding to the information to be recommended, this embodiment can compare the obtained recommendation probability with the preset probability threshold. When it is determined that the recommendation probability is greater than or equal to the preset probability threshold, the information to be recommended (for example, the identification information of the information to be recommended) is sent to the recommended object. After the recommended object decides to purchase the information to be recommended, the specific content contained in the information to be recommended, especially the contact information contained in the information to be recommended, is sent to the recommended object.
[0046] When executing S104 in this embodiment to obtain the recommendation probability of the recommended object corresponding to the information to be recommended according to the predicted click-through rate, the target information vector and the target object vector, an optional implementation method that can be adopted is: splicing the predicted click-through rate, the target information vector and the target object vector, and inputting the splicing result into the third processing model; if the predicted click-through rate, the target information vector and the target object vector are not vectors, converting the features into vectors before splicing them; and using the output result of the third processing model as the recommendation probability.
[0047] In this embodiment, when executing S104 to concatenate the predicted click rate, the target information vector and the target object vector, the target information vector and the target object vector may be firstly subjected to a dot product process, and then the dot product result may be concatenated with the predicted click rate.
[0048] The third processing model used in executing S103 of this embodiment can be pre-trained in the following manner: obtaining third training data, where the third training data includes multiple third feature pairs and labeled recommendation probabilities of multiple third feature pairs, and each third feature pair includes a sample click rate, a sample information vector, and a sample object vector; concatenating the sample click rate, the sample information vector, and the sample object vector in the third feature pair, and inputting the concatenated result into the third neural network model to obtain the predicted recommendation probability output by the third neural network model; calculating the third loss function value according to the predicted recommendation probability and the labeled recommendation probability, and updating the parameters of the third neural network model according to the calculated third loss function value until the third neural network model converges to obtain the third processing model.
[0049] After obtaining the third loss function value, this embodiment may also include the following content: updating the parameters of the second neural network model according to the third loss function value.
[0050] That is to say, this embodiment jointly trains the second neural network model and the third neural network model, and simultaneously updates the parameters of the third neural network model and the second neural network model according to the third loss function value, so that when the third neural network model converges, the parameters of the second neural network model are stopped from being updated, which can enhance the correlation between the click-through rate prediction and the recommendation probability prediction, and thus improve the accuracy of the obtained predicted click-through rate and recommendation probability.
[0051] Figure 2 is a schematic diagram according to a second embodiment of the present disclosure. Figure 2 The flowchart of information recommendation in this embodiment is shown in FIG. ; information features, visitor features, and object features and object behavior features of the information to be recommended are obtained; the object features and object behavior features are converted into vectors and then spliced, and the splicing results are input into the first processing model to obtain the target information vector and target object vector output by the first processing model; the information features, visitor features, and object features are converted into vectors and then spliced, and the splicing results are input into the second processing model to obtain the predicted click-through rate output by the second processing model; the predicted click-through rate, the target information vector, and the target object vector are spliced and input into the third processing model to obtain the recommendation probability output by the third processing model, and then determine whether to recommend the information to be recommended to the recommendation object according to the recommendation probability.
[0052] Figure 3 is a schematic diagram according to a third embodiment of the present disclosure. Figure 3 A schematic diagram of obtaining a target information vector and a target object vector according to a first processing model in this embodiment is shown; the first processing model includes an input layer, three ReLu activation layers, and a Softmax layer; after the object features and the object behavior features are input into the first processing model, the output of the last ReLU activation layer is obtained as the target object vector, and the parameters of the Softmax layer are obtained as the target information vector; the obtained target object vector and target information vector can be stored in a Redis cache for subsequent use.
[0053] Figure 4 is a schematic diagram according to the fourth embodiment of the present disclosure. Figure 4 As shown, the information recommendation device 400 of this embodiment includes:
[0054] The acquisition unit 401 is used to acquire information features of the information to be recommended, and object features and object behavior features of the recommended object;
[0055] A first processing unit 402 is used to obtain a target information vector and a target object vector according to the object characteristics and the object behavior characteristics;
[0056] The second processing unit 403 obtains a predicted click rate according to the information feature and the object feature;
[0057] The third processing unit 404 obtains a recommendation probability that the information to be recommended corresponds to the recommended object according to the predicted click rate, the target information vector and the target object vector.
[0058] The information features of the information to be recommended obtained by the obtaining unit 401 may include features such as product brand, product industry, product description, etc. of the product required by the user in the information to be recommended.
[0059] When the acquisition unit 401 acquires information features of the information to be recommended, it can further acquire user portrait features such as age, gender, and occupation according to the contact information of the user in the information to be recommended as visitor features of the information to be recommended.
[0060] The object features of the recommended object acquired by the acquisition unit 401 may include features such as the industry to which the recommended object belongs and the keywords for which the promotional information of the recommended object has been placed.
[0061] The object behavior characteristics of the recommended object obtained by the acquisition unit 401 include the purchase characteristics and search characteristics of the recommended object; wherein the purchase characteristics include the identification information of the product demand information purchased by the recommended object within a certain time range (for example, within a week, within a month, etc.), the product brand of the product required by the user in the purchased product demand information, the industry to which the product belongs, and other characteristics; the search characteristics include the search keywords entered by the recommended object when searching within a certain time range, the industry to which the search keywords belong, the geographical location at the time of search, and other characteristics.
[0062] After acquiring the information features, object features, object behavior features, and visitor features, the acquisition unit 401 may further process the features into vectors according to a preset processing method.
[0063] It can be understood that this embodiment does not limit the number of recommended information and recommended objects, and can determine the recommendation probability that one piece of recommended information corresponds to one recommended object, can determine the recommendation probability that multiple pieces of recommended information correspond to one recommended object, can determine the recommendation probability that one piece of recommended information corresponds to multiple recommended objects, and can determine the recommendation probability that multiple pieces of recommended information correspond to multiple recommended objects.
[0064] In this embodiment, after the acquisition unit 401 acquires the information features of the information to be recommended and the object features and object behavior features of the recommended object, the first processing unit 402 obtains the target information vector and the target object vector according to the object features and the object behavior features.
[0065] Specifically, when the first processing unit 402 obtains the target information vector and the target object vector based on the object features and the object behavior features, the optional implementation method that can be adopted is: splicing the object features and the object behavior features, and inputting the splicing result into the first processing model; if the features are not vectors, the features can be converted into vectors before splicing; obtaining the output of the last activation layer in the first processing model as the target object vector; obtaining the parameters of the output layer in the first processing model as the target information vector.
[0066] Among them, the first processing model used by the first processing unit 402 is pre-trained. The first processing model is a collaborative filtering model based on DNN (Deep Neural Network), which consists of an input layer, multiple activation layers (such as ReLU activation layer) and an output layer (Softmax layer).
[0067] The information recommendation device 400 of this embodiment may further include a first training unit 405, which is used to train the first processing model in the following manner: obtaining first training data, the first training data including a plurality of first feature pairs and labeled purchase probabilities of the plurality of first feature pairs, each first feature pair including a sample object feature and a sample object behavior feature; after splicing the sample object feature and the sample object behavior feature in the first feature pair, inputting the splicing result into the first neural network model to obtain a predicted purchase probability output by the first neural network model; calculating a first loss function value according to the predicted purchase probability and the labeled purchase probability, and updating the parameters of the first neural network model according to the first loss function value until the first neural network model converges to obtain the first processing model.
[0068] In this embodiment, after the first processing unit 402 obtains the target information vector and the target object vector, the second processing unit 403 obtains the predicted click rate according to the information features and the object features.
[0069] Specifically, when the second processing unit 403 obtains the predicted click-through rate based on the information features and the object features, an optional implementation method that can be adopted is: splicing the information features and the object features, inputting the splicing result into the second processing model, and if the feature is not a vector, converting the feature into a vector before splicing; using the output result of the second processing model as the predicted click-through rate.
[0070] In addition, when the second processing unit 403 splices the information features with the object features and inputs the splicing result into the second processing model, it may also include the following contents: obtaining the visitor features of the information to be recommended; splicing the information features of the information to be recommended, the visitor features of the information to be recommended and the object features of the recommended object, and inputting the splicing result into the second processing model.
[0071] That is to say, the second processing unit 403 may also obtain the predicted click rate in combination with the visitor characteristics of the information to be recommended, thereby improving the accuracy of the obtained predicted click rate.
[0072] The second processing model used by the second processing unit 403 is pre-trained, and the second processing model may be a click rate prediction model based on DeepFM.
[0073] The information recommendation device 400 of this embodiment may further include a second training unit 406, which is used to train a second processing model in the following manner: obtaining second training data, wherein the second training data includes a plurality of second feature pairs and annotated click-through rates of the plurality of second feature pairs, each second feature pair includes a sample information feature and a sample object feature, and may also include a sample visitor feature; after splicing the sample information feature and the sample object feature in the second feature pair, inputting the splicing result into a second neural network model to obtain a predicted click-through rate output by the second neural network model; calculating a second loss function value according to the predicted click-through rate and the annotated click-through rate, and updating the parameters of the second neural network model according to the second loss function value until the second neural network model converges to obtain a second processing model.
[0074] The second neural network model in this embodiment can also be jointly trained with the third neural network model to obtain the second processing model and the third processing model. The specific training process is described below.
[0075] The second processing unit 403 introduces a click-through rate prediction model to obtain a predicted click-through rate based on information features, object features, and visitor features, and can fully cross the above features, so that the obtained predicted click-through rate is used to obtain the final recommendation probability, thereby achieving a higher quality recommendation probability prediction.
[0076] In this embodiment, after the second processing unit 403 obtains the predicted click rate, the third processing unit 404 obtains the recommendation probability of the recommended object corresponding to the information to be recommended according to the predicted click rate, the target information vector and the target object vector.
[0077] After obtaining the recommendation probability of the recommended object corresponding to the information to be recommended, the third processing unit 404 can compare the obtained recommendation probability with the preset probability threshold, and when it is determined that the recommendation probability is greater than or equal to the preset probability threshold, send the information to be recommended (such as identification information of the information to be recommended) to the recommended object, and then after the recommended object decides to purchase the information to be recommended, send the specific content contained in the information to be recommended, especially the contact information contained in the information to be recommended, to the recommended object.
[0078] When the third processing unit 404 obtains the recommendation probability of the recommended object corresponding to the information to be recommended based on the predicted click-through rate, the target information vector and the target object vector, an optional implementation method that can be adopted is: splicing the predicted click-through rate, the target information vector and the target object vector, and inputting the splicing result into the third processing model; if the predicted click-through rate, the target information vector and the target object vector are not vectors, converting the features into vectors before splicing them; and using the output result of the third processing model as the recommendation probability.
[0079] The information recommendation device 400 of this embodiment may further include a third training unit 407, which is used to train a third processing model in the following manner: obtaining third training data, wherein the third training data includes multiple third feature pairs and annotated recommendation probabilities of multiple third feature pairs, and each third feature pair includes a sample click rate, a sample information vector, and a sample object vector; concatenating the sample click rate, the sample information vector, and the sample object vector in the third feature pair, and inputting the concatenated result into a third neural network model to obtain a predicted recommendation probability output by the third neural network model; calculating a third loss function value according to the predicted recommendation probability and the annotated recommendation probability, and updating the parameters of the third neural network model according to the calculated third loss function value until the third neural network model converges to obtain the third processing model.
[0080] After obtaining the third loss function value, the third training unit 407 may further include the following content: updating the parameters of the second neural network model according to the third loss function value.
[0081] That is to say, the third training unit 407 jointly trains the second neural network model and the third neural network model, and simultaneously updates the parameters of the third neural network model and the second neural network model according to the third loss function value, so that when the third neural network model converges, the parameters of the second neural network model are stopped from being updated, which can enhance the correlation between the click-through rate prediction and the recommendation probability prediction, and thus improve the accuracy of the obtained predicted click-through rate and recommendation probability.
[0082] It can be understood that the first training unit 405, the second training unit 406 and the third training unit 407 in this embodiment can be the same training unit, which is used to train and obtain the first processing model, the second processing model and the third processing model.
[0083] In the technical solution disclosed herein, the acquisition, storage and application of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.
[0084] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.
[0085] like Figure 5, is a block diagram of an electronic device according to an information recommendation method according to an embodiment of the present disclosure. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.
[0086] like Figure 5 As shown, the device 500 includes a computing unit 501, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 502 or a computer program loaded from a storage unit 508 into a random access memory (RAM) 503. In the RAM 503, various programs and data required for the operation of the device 500 can also be stored. The computing unit 501, the ROM 502, and the RAM 503 are connected to each other via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.
[0087] A number of components in the device 500 are connected to the I / O interface 505, including: an input unit 506, such as a keyboard, a mouse, etc.; an output unit 507, such as various types of displays, speakers, etc.; a storage unit 508, such as a disk, an optical disk, etc.; and a communication unit 509, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 509 allows the device 500 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.
[0088] The computing unit 501 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 501 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 501 performs the various methods and processes described above, such as the information recommendation method. For example, in some embodiments, the information recommendation method may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 508.
[0089] In some embodiments, part or all of the computer program may be loaded and / or installed on the device 500 via the ROM 502 and / or the communication unit 509. When the computer program is loaded into the RAM 503 and executed by the computing unit 501, one or more steps of the information recommendation method described above may be performed. Alternatively, in other embodiments, the computing unit 501 may be configured to perform the information recommendation method in any other appropriate manner (e.g., by means of firmware).
[0090] Various implementations of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.
[0091] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable information recommendation device, so that when the program code is executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine as a stand-alone software package and partially on a remote machine, or entirely on a remote machine or server.
[0092] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
[0093] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).
[0094] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.
[0095] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship between the client and the server is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or a cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services ("Virtual Private Server", or "VPS" for short). The server may also be a server for a distributed system, or a server combined with a blockchain.
[0096] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.
[0097] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.
Claims
1. An information recommendation method, comprising: Acquire information features of the information to be recommended, and object features and object behavior features of the recommended object; Obtaining a target information vector and a target object vector according to the object characteristics and the object behavior characteristics; Obtaining a predicted click rate according to the information features and the object features; A recommendation probability of the information to be recommended corresponding to the recommended object is obtained according to the predicted click rate, the target information vector and the target object vector.
2. The method according to claim 1, wherein: The obtaining of a target information vector and a target object vector according to the object feature and the object behavior feature comprises: splicing the object feature with the object behavior feature, and inputting the splicing result into a first processing model; Obtaining the output of the last activation layer in the first processing model as the target object vector; Obtain parameters of the output layer in the first processing model as the target information vector.
3. The method according to claim 1, wherein: The obtaining of the predicted click rate according to the information feature and the object feature includes: splicing the information feature with the object feature, and inputting the splicing result into a second processing model; The output result of the second processing model is used as the predicted click rate.
4. The method according to claim 3, wherein: The step of combining the information feature with the object feature and inputting the combined result into a second processing model comprises: Obtain visitor characteristics of the information to be recommended; The information features, the visitor features and the object features are concatenated, and the concatenated result is input into the second processing model.
5. The method according to claim 1, wherein: The obtaining, according to the predicted click rate, the target information vector and the target object vector, of the recommendation probability of the information to be recommended corresponding to the recommended object comprises: splicing the predicted click rate, the target information vector and the target object vector, and inputting the splicing result into a third processing model; The output result of the third processing model is used as the recommendation probability.
6. The method according to claim 1, further comprising: When it is determined that the recommendation probability is greater than or equal to a preset probability threshold, the information to be recommended is recommended to the recommendation object.
7. The method according to claim 1, further comprising: Acquire second training data, where the second training data includes a plurality of second feature pairs and annotated click rates of the plurality of second feature pairs, and each second feature pair includes a sample information feature and a sample object feature; After splicing the sample information feature and the sample object feature in the second feature pair, inputting the splicing result into the second neural network model to obtain the predicted click rate output by the second neural network model; A second loss function value is calculated according to the predicted click rate and the annotated click rate, and the parameters of the second neural network model are updated according to the second loss function value to obtain the second processing model.
8. The method according to claim 1, further comprising: Acquire third training data, wherein the third training data includes a plurality of third feature pairs and a plurality of labeling recommendation probabilities of the third feature pairs, and each third feature pair includes a sample click rate, a sample information vector, and a sample object vector; Concatenate the sample click rate, sample information vector and sample object vector in the third feature pair, input the concatenation result into a third neural network model, and obtain the predicted recommendation probability output by the third neural network model; A third loss function value is calculated according to the predicted recommendation probability and the labeled recommendation probability, and the parameters of the third neural network model are updated according to the third loss function value to obtain the third processing model.
9. The method according to claim 8, further comprising: According to the third loss function value, update the parameters of the second neural network model.
10. An information recommendation device, comprising: An acquisition unit, used to acquire information features of the information to be recommended, and object features and object behavior features of the recommended object; A first processing unit, configured to obtain a target information vector and a target object vector according to the object feature and the object behavior feature; A second processing unit, configured to obtain a predicted click rate according to the information feature and the object feature; The third processing unit is used to obtain the recommendation probability of the information to be recommended corresponding to the recommended object according to the predicted click rate, the target information vector and the target object vector.
11. The device according to claim 10, wherein: When the first processing unit obtains the target information vector and the target object vector according to the object feature and the object behavior feature, the first processing unit specifically performs: splicing the object feature with the object behavior feature, and inputting the splicing result into a first processing model; Obtaining the output of the last activation layer in the first processing model as the target object vector; Obtain parameters of the output layer in the first processing model as the target information vector.
12. The device according to claim 10, wherein: When the second processing unit obtains the predicted click rate according to the information feature and the object feature, the second processing unit specifically performs: splicing the information feature with the object feature, and inputting the splicing result into a second processing model; The output result of the second processing model is used as the predicted click rate.
13. The device according to claim 12, wherein: When the second processing unit splices the information feature with the object feature and inputs the splicing result into the second processing model, specifically: Obtain visitor characteristics of the information to be recommended; The information features, the visitor features and the object features are concatenated, and the concatenated result is input into the second processing model.
14. The device according to claim 10, wherein: When the third processing unit obtains the recommendation probability of the information to be recommended corresponding to the recommended object according to the predicted click rate, the target information vector and the target object vector, the third processing unit specifically performs: splicing the predicted click rate, the target information vector and the target object vector, and inputting the splicing result into a third processing model; The output result of the third processing model is used as the recommendation probability.
15. The device according to claim 10, wherein: The third processing unit is further configured to execute: When it is determined that the recommendation probability is greater than or equal to a preset probability threshold, the information to be recommended is recommended to the recommendation object.
16. The apparatus according to claim 10, further comprising a second training unit, configured to train the second processing model in the following manner: Acquire second training data, where the second training data includes a plurality of second feature pairs and annotated click rates of the plurality of second feature pairs, and each second feature pair includes a sample information feature and a sample object feature; After splicing the sample information feature and the sample object feature in the second feature pair, inputting the splicing result into the second neural network model to obtain the predicted click rate output by the second neural network model; A second loss function value is calculated according to the predicted click rate and the annotated click rate, and the parameters of the second neural network model are updated according to the second loss function value to obtain the second processing model.
17. The apparatus according to claim 10, further comprising a third training unit, configured to train the second processing model in the following manner: Acquire third training data, wherein the third training data includes a plurality of third feature pairs and a plurality of labeling recommendation probabilities of the third feature pairs, and each third feature pair includes a sample click rate, a sample information vector, and a sample object vector; Concatenate the sample click rate, sample information vector and sample object vector in the third feature pair, input the concatenation result into a third neural network model, and obtain the predicted recommendation probability output by the third neural network model; A third loss function value is calculated according to the predicted recommendation probability and the labeled recommendation probability, and the parameters of the third neural network model are updated according to the third loss function value to obtain the third processing model.
18. The device according to claim 17, wherein: The third training unit is further configured to execute: According to the third loss function value, update the parameters of the second neural network model.
19. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 9.
20. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1 to 9.
21. A computer program product comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 9.
Citation Information
Patent Citations
Content recommendation method and device, electronic equipment and storage medium
CN111177575A
Information pushing method and device, electronic equipment and storage medium
CN112925972A
Recommendation method and device, storage medium and electronic equipment
CN113742590A
Information recommendation method and device, electronic equipment and computer readable storage medium
CN115858911A
Recommendation method, model generation method and apparatus, medium and device
WO2021135588A1
