Information pushing method, computer equipment and computer program product

Abstract

The invention discloses an information pushing method, computer equipment and a computer program product. The information pushing method comprises the steps of determining candidate materials matched with user feature data of a target user from multiple materials; the transformation probability of the target user for the candidate materials pushed through different channels is predicted to serve as the transformation probability of the candidate materials under different channels, and the transformation probability is used for representing the probability of a transformation behavior; determining a target channel and a target material meeting a target condition based on the conversion probability of the candidate material under different channels; and pushing the target material to equipment corresponding to the target user through the target channel. According to the method, the pushing effect of information pushing can be improved.

Description

Technical Field

[0001] The present application relates to the field of Internet technology, and more specifically, to an information push method, a computer device, and a computer program product. Background Art

[0002] With the rapid advancement of technology and living standards, electronic devices (such as smart phones, tablet computers, etc.) have become one of the commonly used electronic products in people's lives. In addition, with the development of Internet technology, products are often pushed to users' electronic devices by sending text messages, emails, etc. However, the current method is not accurate enough in pushing information, resulting in poor push effects. Summary of the invention

[0003] The present application proposes an information push method, a computer device and a computer program product, which can improve the push effect of information push.

[0004] In a first aspect, an embodiment of the present application provides an information push method, the method comprising: determining candidate materials that match user feature data of a target user from a plurality of materials; predicting the conversion probability of the target user for the candidate materials pushed through different channels as the conversion probability of the candidate materials in different channels, the conversion probability being used to characterize the probability of conversion behavior occurring; determining a target channel and a target material that meet a target condition based on the conversion probability of the candidate materials in different channels; and pushing the target material to a device corresponding to the target user through the target channel.

[0005] In the second aspect, an embodiment of the present application provides an information push device, which includes: a first determination module, a probability prediction module, a second determination module and a material push module, wherein the first determination module is used to determine candidate materials that match user feature data of a target user from a plurality of materials; the probability prediction module is used to predict the conversion probability of the target user for the candidate materials pushed through different channels, as the conversion probability of the candidate materials in different channels, and the conversion probability is used to characterize the probability of conversion behavior; the second determination module is used to determine a target channel and a target material that meet the target conditions based on the conversion probability of the candidate materials in different channels; the material push module is used to push the target material to the device corresponding to the target user through the target channel.

[0006] In a third aspect, an embodiment of the present application provides a computer device comprising: one or more processors; a memory; one or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, and the one or more applications are configured to execute the information push method provided in the first aspect above.

[0007] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, in which a program code is stored. The program code can be called by a processor to execute the information push method provided in the first aspect above.

[0008] In a fifth aspect, an embodiment of the present application provides a computer program product, which includes a computer program, and the computer program is stored in a storage medium; a processor of a computer device reads the computer program from the storage medium, and the processor executes the computer program, so that the computer device executes the information push method provided in the first aspect above.

[0009] The solution provided by the present application determines the candidate material that matches the user characteristic data of the target user from a variety of materials, predicts the conversion probability of the target user for the candidate material pushed through different channels, and uses the conversion probability of the candidate material in different channels as the conversion probability. The conversion probability is used to characterize the probability of the conversion behavior. Based on the conversion probability of the candidate material in different channels, the target channel and target material that meet the target conditions are determined, and the target material is pushed to the device corresponding to the target user through the target channel. Therefore, when pushing information, it is possible to accurately determine the channel and material for the candidate material that matches the user characteristic data, combined with the push effect of the material and the channel (i.e., the conversion probability of the above candidate material in different channels), thereby improving the accuracy of information push, and then improving the push effect of information push. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0011] Figure 1 A schematic diagram of an application environment provided by an embodiment of the present application is shown.

[0012] Figure 2 A schematic flow chart of an information push method according to an embodiment of the present application is shown.

[0013] Figure 3 A flow chart of an information push method according to another embodiment of the present application is shown.

[0014] Figure 4 A flow chart of an information push method according to another embodiment of the present application is shown.

[0015] Figure 5A flow chart of an information push method according to another embodiment of the present application is shown.

[0016] Figure 6 A schematic flow chart of an information push method according to yet another embodiment of the present application is shown.

[0017] Figure 7 A schematic diagram showing the principle of the information push method provided in an embodiment of the present application is shown.

[0018] Figure 8 A block diagram of an information push device according to an embodiment of the present application is shown.

[0019] Fig. 9 It is a block diagram of a computer device according to an embodiment of the present application for executing the information push method according to an embodiment of the present application.

[0020] Fig.10 It is a storage unit of an embodiment of the present application for storing or carrying program codes for implementing the information push method according to an embodiment of the present application. DETAILED DESCRIPTION

[0021] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0022] At present, in Internet services, corresponding marketing business is often carried out through active marketing. For example, common active marketing methods include SMS marketing, notification bar message push, red dots, coupon distribution, etc. The common operation mode of active marketing is: based on the user's behavior data, interest preferences and historical activities, the system generates user tags; business personnel select specific groups of people by selecting these tags. For example, a user may have tags such as "high-frequency shopper" and "fashion lover"; business personnel select target user groups and push marketing information according to predefined rules and strategies. For example, the latest fashion discount information is pushed to "fashion lovers" at a specific time; in order to avoid excessive disturbance to users, the system sets strict frequency control rules. For example, each user receives a maximum of a certain number of notification bar messages or SMS messages per day; by monitoring the effectiveness indicators of marketing activities (such as open rate, click-through rate, conversion rate, etc.), business personnel constantly adjust the tag combination and delivery strategy to optimize the marketing effect.

[0023] In addition, the scenario of active marketing is different from the common in-feed advertising and recommendation systems. In-feed advertising, users actively enter the advertising display location (such as news push on social media), and then the system recommends advertising content based on user characteristics and behavior data. This method is usually external advertisers using platform traffic for promotion, and the user's click behavior is active; the recommendation system is when users actively browse content, the system recommends related content based on the user's interests and historical behavior. This method is also the user actively triggering the system recommendation, the main purpose is to improve user experience and platform stickiness; while active marketing is the application developer actively pushing marketing information to users, usually through SMS, in-app notifications and other channels, the purpose is to encourage users to perform specific payment behaviors within the application. In this scenario, the user's reception is passive, but the accuracy and timeliness of marketing are key.

[0024] In the related art, the core problem of active marketing is how to effectively determine "which users (Which Users)" at "what time (When)" and in "what way (How)" to push "which marketing materials (Which Marketing Materials)", but the related art cannot solve this problem well when conducting active marketing, resulting in insufficient accuracy of marketing information push and poor push effect.

[0025] In response to the above problems, the inventors have proposed the information push method, computer device and computer program product provided in the embodiments of the present application, which can accurately determine the channel and material for the candidate material matching the user feature data and the push effect of the material and channel (i.e., the conversion probability of the above candidate material under different channels) when pushing information, thereby improving the accuracy of information push and further improving the push effect of information push. Among them, the specific information push method is described in detail in the subsequent embodiments.

[0026] The following first introduces the scenarios involved in the embodiments of the present application.

[0027] like Figure 1 As shown, in Figure 1The scene shown includes a server 100 and a client 200. The server 100 may be a server (network access server), a server cluster (cloud server) consisting of several servers, or a cloud computing center (database server). The client 200 may be any electronic device with communication and storage functions, including but not limited to a PC (Personal Computer), a PDA (tablet computer), a smart TV, a smart phone, a smart wearable device, or other smart communication devices with network connection functions.

[0028] The server 100 and the client 200 can interact through the network to receive or send information. In the embodiment of the present application, the server 100 can determine the candidate material that matches the user feature data of the target user corresponding to the client 200 from a variety of materials, predict the conversion probability of the target user for the candidate material pushed through different channels, as the conversion probability of the candidate material under different channels, and then determine the target channel and target material that meet the target conditions based on the conversion probability of the candidate material under different channels, and finally push the target material to the client 200 through the determined target channel.

[0029] The information push method provided in the embodiment of the present application is described in detail below with reference to the accompanying drawings.

[0030] See also Figure 2 , Figure 2 The following is a flow chart of an information push method provided by an embodiment of the present application. Figure 2 The process shown is described in detail, and the information push method may specifically include the following steps:

[0031] Step S110: determining candidate materials matching the user characteristic data of the target user from the multiple materials.

[0032] In an embodiment of the present application, when a computer device pushes information to a target user, it can determine the material that matches the target user from the material pool. The computer device can obtain the user characteristic data of the target user, and then determine the candidate material that matches the user characteristic data of the target user from the multiple materials in the material pool. Among them, the target user can be any user among all users; the computer device that executes the information push method provided in the embodiment of the present application can be the aforementioned server; the user characteristic data can be data used to reflect the characteristics of the user, so by determining the material that matches the user characteristic data, the candidate material that matches the target user can be obtained. Material refers to the material that constitutes the work or expresses the content. The material can be used to promote and publicize products, services or brands. The information in the material can include text, pictures, videos, audio and other forms to attract the attention of the target audience, convey specific marketing information, and prompt the audience to take action (such as purchase, registration, learn more information, etc.).

[0033] By first determining candidate materials that match the user's feature data from a variety of materials, some materials that obviously do not meet the user's needs can be filtered out, which can ensure the accuracy of the materials that need to be pushed in the subsequent determination, and reduce the amount of processing in the subsequent processing. It should be noted that the number of candidate materials determined can be one or more.

[0034] Optionally, the application scenario of the information push method provided in the embodiment of the present application can be an active marketing scenario. Active marketing refers to the application developer (App Developer) actively pushing marketing information to users through its own platform, and its core lies in the effective use and conversion of private traffic (Private Traffic). In this scenario, the material mentioned in the embodiment of the present application can be a marketing material, and the channel for pushing the marketing material can be a marketing channel.

[0035] In some embodiments, the computer device may collect user characteristic data of the user through various channels. Optionally, the user characteristic data may include, but is not limited to, the user's historical behavior, real-time interaction data, and contextual information. The historical behavior may include browsing behavior (e.g., browsing history, dwell time, click behavior, etc. of the user on the website or application), search behavior (user's search keywords, search frequency, search time, etc.), purchase behavior (user's purchase record, purchase preference, purchase frequency, purchase amount, etc.), and usage habits (the time, frequency, and method of the user's use of the product or service). The real-time interaction data may include click data, browsing data, search data, form data, evaluation data, consultation data, geographic location, information of the currently used device, an identifier of the user session, etc., and the contextual information may refer to background information related to the user's current behavior, demand, or status, such as time context information, space context information, social context information, device context information, task context information, etc.

[0036] Among them, for the above real-time interactive data, click data can be a record of users clicking on elements such as pages, links, buttons, etc., including click time, click location, click frequency, etc.; browsing data can be the user's browsing trajectory on the page, including browsing time, scrolling depth, page dwell time, etc.; search data can be the keywords entered by the user in the search box, search time, search result selection, etc.; form data can be the information entered by the user when filling in the form, such as registration information, feedback, etc.; evaluation data can be the user's evaluation, rating, comments, etc. on the product or service; consultation data can be consultations and questions raised by users through online customer service, message boards, social media and other channels; geographic location can be the geographic location information when the user visits; information on the currently used device can be the type of device used by the user, operating system, browser version, etc.; the identifier of the user session can be a unique identifier used to identify the user session, which can track the user's behavior throughout the session.

[0037] For the above context information, temporal context information can be the time point or time period when the user performs an activity or accesses a service. Temporal information is of great significance for understanding user behavior patterns, interest changes, etc. For example, in a recommendation system, temporal context information can help identify users' seasonal interests or periodic behavior patterns. Spatial context information can be the user's geographical location or environment. Location information is crucial for understanding users' localized needs and interests. For example, users in different regions may pay more attention to local news, weather, catering and other information. Social context information can be the user's social network and social interaction. Social context information helps to understand the user's social influence, interest preferences and social behavior patterns. For example, on social media platforms, the user's follow list, friend relationships, etc. can all be part of the social context information. Device context information can be the type of device, operating system, browser version, etc. used by the user. Device information is of great significance for optimizing user experience and providing highly adaptable services or content. Task context information can be the task that the user is currently performing or hopes to complete. Task context information helps to provide users with more targeted help and guidance.

[0038] In a possible implementation, after obtaining the above user feature data, the computer device may preprocess the user feature data, and the preprocessing may include data cleaning, feature extraction, data normalization, etc. Data cleaning may be to remove noise data and abnormal data from the user feature data to ensure data quality; feature extraction may be to extract key behavioral features for the obtained user feature data, such as recent payment records, browsing history, etc.; data normalization may standardize data of different dimensions so that it can be used for subsequent processing by related models.

[0039] In some embodiments, the computer device can, based on user feature data, target multiple materials in the material pool through one or more of feature matching, similarity calculation, multi-level filtering and scoring mechanisms to determine the material that best meets the current user needs from the multiple materials as the above candidate material.

[0040] Step S120: predicting the conversion probability of the target user for the candidate material pushed through different channels as the conversion probability of the candidate material in different channels, wherein the conversion probability is used to characterize the probability of conversion behavior.

[0041] In the embodiment of the present application, after determining the above candidate materials, the conversion probability of the target user for the candidate materials pushed through different channels can be predicted for each candidate material in the candidate materials, and the predicted conversion probability of the candidate materials pushed through different channels is used as the conversion probability of the candidate materials in different channels. Among them, the above conversion probability represents the probability of the target user performing a conversion behavior for the pushed candidate materials, and the conversion behavior can be an operation behavior such as clicking, sharing, and purchasing the pushed materials.

[0042] It is understandable that candidate materials may be pushed through different channels. When the same candidate material is pushed through different channels, the probability of users converting (such as clicking) to it is different. Therefore, the conversion probability can be predicted for the situation where the candidate material is pushed through different channels, so as to more accurately determine the channel and material for information push to the target user this time. Moreover, since the channel and material are combined at the same time when determining the above conversion probability, it is convenient to more accurately determine the channel and material based on the determined conversion probability in the future. Among them, the channel can be SMS marketing, notification bar message push, red dot, coupon distribution and other channels.

[0043] In some embodiments, the computer device may predict the conversion probability of the candidate materials pushed by the target user under different channels based on the candidate materials and different channels, and according to the above user characteristic data, based on the pre-trained prediction model for predicting the conversion probability. Among them, the above prediction model may be a deep learning model (Deep Learning Models), a logistic regression model (Logistic Regression Models), etc., and the specific type of the prediction model may not be limited in the embodiments of the present application.

[0044] Exemplarily, the candidate materials include material 1, material 2 and material 3, and the channels include channel 1, channel 2 and channel 3. The conversion probability of each material in different channels can be shown in the following table:

[0045] Channel 1 Channel 2 Channel 3 Material 1 Probability a Probability d Probability g Material 2 Probability b Probability e Probability h Material 3 Probability c Probability f Probability i

[0046] In the above table, probability a represents the conversion probability of the target user for material 1 pushed through channel 1, probability b represents the conversion probability of the target user for material 2 pushed through channel 1, probability b represents the conversion probability of the target user for material 3 pushed through channel 1, probability d represents the conversion probability of the target user for material 1 pushed through channel 2, probability g represents the conversion probability of the target user for material 1 pushed through channel 2, and so on.

[0047] Step S130: Based on the conversion probabilities of the candidate materials in different channels, determine the target channels and target materials that meet the target conditions.

[0048] In the embodiment of the present application, after determining the conversion probability of the above candidate materials under different channels for the candidate materials, the channels and materials to be pushed for the information push for the target user can be determined. The computer device can determine the target channel and target material that meet the target conditions based on the conversion probability of the candidate materials under different channels. The target channel is the channel to be used for the information push for the target user this time, and the target material is the material to be pushed for the information push for the target user this time. Among them, the above target material is determined in the candidate material, the target channel is determined in the multiple channels corresponding to the multiple conversion probabilities corresponding to the target material, and the above target conditions can be the conditions set for the needs of the push task (such as marketing business) this time, that is, the above target channels and target materials determined can meet the needs of the push task (such as marketing business) this time. It can be understood that the above conversion probability can reflect the push effect to a certain extent, and the above conversion probability is corresponding to the material and channel, so the above target channel and target material can be determined based on the above conversion probability.

[0049] In some embodiments, the target conditions may include at least a set value expectation. The computer device may calculate the value expectation of each material being pushed through different channels based on the conversion probability of each material in different channels, and then determine the target channels and target materials that meet the above target conditions based on the obtained value expectations.

[0050] Step S140: Pushing the target material to the device corresponding to the target user through the target channel.

[0051] In the embodiment of the present application, after determining the target channel and target material that meet the target conditions, the target material can be pushed to the device corresponding to the target user through the target channel. The device corresponding to the target user can be the aforementioned client, for example, the target user's smartphone, smart wearable device, etc.

[0052] The information push method provided in the embodiment of the present application can, when pushing information, accurately determine the channel and the material for the candidate material that matches the user's characteristic data, combined with the material and the push effect of the channel (i.e., the conversion probability of the above candidate material under different channels), and then push the information based on the determined channel and material, thereby improving the accuracy of the information push and further improving the push effect of the information push.

[0053] See also Figure 3 , Figure 3A schematic diagram of a flow chart of an information push method provided by another embodiment of the present application is shown. Figure 3 The process shown is described in detail, and the information push method may specifically include the following steps:

[0054] Step S210: determining candidate materials matching the user characteristic data of the target user from the multiple materials.

[0055] In the embodiment of the present application, step S210 can refer to the contents of other embodiments and will not be repeated here.

[0056] Step S220: Determine each channel that the candidate material matches.

[0057] In the embodiment of the present application, different prediction models can be used to predict conversion probabilities for different channels. When predicting the conversion probability of target users for candidate materials pushed through different channels for each candidate material, each channel that the candidate material matches can be determined for the candidate material. It can be understood that different materials can be applied to different channels. For example, if a material contains graphic information, then its applicable channels include push notifications. Therefore, the channels that the candidate material is applicable to can be determined first, as each channel that the candidate material matches.

[0058] Step S230: Utilizing the prediction model corresponding to each channel and based on the user characteristic data, predict the conversion probability of the target user for the candidate material pushed through each channel as the conversion probability of the candidate material under different channels, and the conversion probability is used to characterize the probability of conversion behavior occurring.

[0059] In the embodiment of the present application, considering that the push effects of different channels are different, different prediction models can be used to predict the conversion probability for different channels. After determining the channels that match the candidate materials, the prediction model corresponding to each channel can be used, and based on the user feature data, the conversion probability of the target user for the candidate materials pushed through each channel can be predicted. Among them, the prediction model can be a deep learning model, a logistic regression model, etc., and the specific type of the prediction model may not be limited in the embodiment of the present application.

[0060] In some embodiments, for the case where a candidate material is pushed under each channel that matches it, the above user behavior feature data and the information of the channel can be input into the prediction model corresponding to each channel to obtain the result output by the prediction model corresponding to each channel. The result output by the prediction model corresponding to each channel is the conversion probability of pushing the candidate material through each channel.

[0061] In some embodiments, considering that after materials of different business types are pushed, the conversion behaviors of users are different, the prediction sub-model corresponding to the corresponding prediction task can be used to predict the conversion probability for different materials. Among them, the computer device can use the target prediction sub-model among the multiple prediction sub-models corresponding to each channel, and based on the user characteristic data, predict the conversion probability of the target user for the candidate material pushed through each channel as the conversion probability of the candidate material under different channels, wherein the multiple prediction sub-models correspond to multiple prediction tasks one by one, and the prediction task corresponding to the target prediction sub-model matches the business type of the material. Among them, the prediction task can be to predict the click conversion probability, registration conversion probability, purchase conversion probability, download conversion probability, etc.

[0062] In the above implementation, the prediction sub-model can be a click-through rate prediction model (CTRM), a conversion rate prediction model (CVRM), a user behavior sequence model (UBSM), etc. Among them, the click-through rate prediction model is a very important model in the fields of recommendation systems and online advertising, which is used to predict the probability of users clicking on pushed content (such as advertisements, search results, recommended products, etc.); the conversion rate prediction model is a model used to predict the conversion probability of users performing specific behaviors (such as purchases, registrations, downloads, etc.); the user behavior sequence model is a model used to analyze and predict the user's behavior sequence in a specific environment, and can output the probability of the behavior occurring.

[0063] Through the above implementation method, when determining the conversion probability of candidate materials in different channels, it is possible to route to the prediction model corresponding to the corresponding channel according to the material, and further route to the prediction sub-model corresponding to the prediction task that matches the business type of the material, so as to predict the user's conversion behavior (i.e., conversion probability) for materials in different channels, and not only consider the characteristics of different channels (such as SMS marketing, notification bar message push, red dots, coupon distribution, etc.), but also realize refined prediction through multi-model combination (combination of channels and prediction tasks), and finally obtain the conversion probability, thereby improving the accuracy of the predicted conversion probability.

[0064] In a possible implementation, when using the target prediction sub-model among the multiple prediction sub-models corresponding to each channel to predict the conversion probability of the target user for the candidate material pushed through each channel, the user feature data and the material information can be input into the target prediction sub-model to obtain the conversion probability output by the target prediction sub-model.

[0065] Step S240: Calibrate the conversion probability of the candidate material in different channels based on the conversion probability calibration model.

[0066] In an embodiment of the present application, after determining the conversion probability of the candidate material under different channels, when determining the corresponding target channel and target material based on the obtained conversion probability, the conversion probability of the candidate material under different channels can also be calibrated based on the conversion probability calibration model to further improve the accuracy of the conversion probability, thereby ensuring that the subsequently determined target channels and target materials are more accurate.

[0067] In some embodiments, the computer device can calibrate each conversion probability obtained above through an isotonic regression model, so that the calibrated conversion probability is closer to the actual conversion rate. In addition, the calibrated conversion probability can not only reflect the ranking information, but also be used for subsequent calculations and decisions. Among them, isotonic regression, also known as monotonic regression, is a special form of regression analysis. It performs non-parametric estimation on given data in a monotonic function space. The basic idea of ​​isotonic regression is to find a set of non-decreasing fragmented continuous linear functions (i.e., isotonic functions) under the premise of maintaining the monotonicity of the function, so that these functions are as close as possible to the sample data.

[0068] Step S250: Based on the conversion probabilities of the candidate materials in different channels, determine the target channels and target materials that meet the target conditions.

[0069] Step S260: Pushing the target material to the device corresponding to the target user through the target channel.

[0070] In the embodiment of the present application, step S250 and step S260 can refer to the contents of other embodiments and will not be repeated here.

[0071] The information push method provided in the embodiment of the present application can, when pushing information, accurately determine the channel and the material for the candidate material matching the user feature data, combined with the influence of the material and the channel on the push effect (i.e., the conversion probability of the above candidate material in different channels), and then push the information according to the determined channel and material, thereby improving the accuracy of the information push, and further improving the push effect of the information push; and, when determining the conversion probability of the candidate material in different channels, the corresponding prediction sub-model is determined in combination with the channel and the material, and the conversion probability is predicted, thereby ensuring the accuracy of the obtained conversion probability, thereby ensuring that the target channel and target material determined subsequently are more accurate; in addition, the conversion probability of the candidate material in different channels is calibrated, further improving the accuracy of the conversion probability, thereby further improving the accuracy of the determined target channel and target material.

[0072] See also Figure 4 , Figure 4 A schematic diagram of a flow chart of an information push method provided by another embodiment of the present application is shown. Figure 4 The process shown is described in detail, and the information push method may specifically include the following steps:

[0073] Step S310: determining candidate materials matching the user characteristic data of the target user from the multiple materials.

[0074] Step S320: predicting the conversion probability of the target user for the candidate material pushed through different channels as the conversion probability of the candidate material in different channels, wherein the conversion probability is used to characterize the probability of conversion behavior.

[0075] In the embodiment of the present application, step S310 and step S320 can refer to the contents of other embodiments and will not be described in detail here.

[0076] Step S330: Based on the conversion probabilities of the candidate materials in different channels, the expected values ​​of pushing the candidate materials through different channels are determined.

[0077] In the embodiment of the present application, when determining the target channel and the target material based on the conversion probability of the candidate material in different channels, the expected value (Expected Marketing Value, EMV) of the candidate material pushed through different channels can be determined based on the conversion probability of the candidate material in different channels, so as to determine the target channel and the target material based on the determined expected value. The expected value is the expected value obtained by pushing information.

[0078] In some embodiments, for each material pushed under different channels, its expected value can be calculated, and the expected value is the product of the conversion probability and the conversion revenue. Since the above conversion probability is for one of the materials under a certain channel, it can be aimed at different combinations of materials and channels, so it is possible to calculate the expected value according to different channels and different materials, so as to facilitate the subsequent determination of the target channel and target material according to the calculated expected value. For example, although push notifications and rich text SMS may belong to the same business, their sending levels and user acceptance are different. Therefore, through the above method, their expected values ​​can be calculated separately, and the exclusivity between channels is taken into account.

[0079] Step S340: Based on the expected value of pushing the candidate material through different channels and the constraints, determine the target channel and target material that meet the target conditions.

[0080] In the embodiment of the present application, the target channel and target material that meet the target condition can be determined based on the expected value of the candidate material being pushed through different channels and the constraint condition. The constraint condition can be the constraint condition set for this push task.

[0081] In some implementations, the above constraints may include frequency control rules, channel level constraints, user experience optimization conditions, etc. Frequency control rules are used to constrain the push frequency of materials. For example, users are not allowed to receive too many materials of the same service within a period of time to avoid user disgust. Frequency control rules need to set frequency control thresholds; channel level constraints are used to constrain the number of pushes through a certain channel. For example, the sending level of rich text messages is relatively small compared to push notifications; user experience optimization conditions are to avoid high-frequency touches by considering the user's reception history and interaction records to enhance the user experience. For example, the same service is not repeatedly touched in a short period of time.

[0082] In some implementations, based on the expected value of pushing candidate materials through different channels and constraints, multiple push schemes can be sorted to obtain a sorting result, where the push scheme includes channels and materials; a target push scheme is determined from the sorting result, and the channel in the target push scheme is used as the target channel, and the material in the target scheme is used as the target material. It can be understood that the above value expectations are calculated for the situation when the material is pushed through different channels, so the combination of each material and each channel can be regarded as a push scheme, and the push schemes are sorted according to the expected value and constraints, and then the above target channels and target materials are determined from the sorting results.

[0083] In a possible implementation, a solution that maximizes the expected value under the above constraints can be found through optimization algorithms such as linear programming (LP) and integer programming (IP). For example, a mixed-integer programming (MIP) model is used to comprehensively consider various conditions in the constraints and generate the optimal sorting result.

[0084] In a possible implementation, after obtaining the above sorting results, the top push scheme can be determined from the sorting results as the target push scheme, and the channel in the target push scheme can be used as the target channel, and the material in the target push scheme can be used as the target material. Of course, in some push scenarios, the top N push schemes can also be determined from the sorting results as the above target push schemes, and the channel in each target push scheme can be used as a target channel, and the material in each target push scheme can be used as a target material.

[0085] In a possible implementation, after obtaining the above sorting results, the above sorting results can be adjusted according to the high priority rule, and the high priority rule can be set according to operational needs. For example, in the push scenario of marketing information, the priority can be dynamically adjusted for the materials that need to be sent first for a specific marketing campaign to ensure that high priority materials are processed first.

[0086] Optionally, the above high-priority rules can be customized by operations personnel through the policy and rule engine, such as priority sending of specific businesses, special holiday activities and other strategies. In this way, the above sorting results can be intelligently sorted according to the set rules; in addition, rule combinations can also be set to achieve complex rule combinations. For example, high-value expected materials for specific businesses are given priority without exceeding the daily push quantity limit.

[0087] In some embodiments, in the process of determining the target material and the target channel for the above target user, it can also be determined whether the expected value of the above push schemes is greater than the target value, and whether the cost of each push scheme is greater than the target cost; if the expected value of each push scheme is not greater than the target value, and / or the cost of each push scheme is greater than the target cost, it means that the target user does not meet the push conditions this time, and therefore the material is not pushed to the target user, and the next user is traversed from all users, and the next traversed user is taken as the target user, and the above step S310 is returned; if there is a push scheme whose expected value is greater than the target value, and the cost of the push scheme is less than the target cost, then the above target push scheme can be determined from these push schemes.

[0088] Step S350: Pushing the target material to the device corresponding to the target user through the target channel.

[0089] In the embodiment of the present application, step S350 can refer to the contents of other embodiments and will not be repeated here.

[0090] The information push method provided in the embodiment of the present application can, when pushing information, accurately determine the channel and the material for the candidate material that matches the user feature data, combined with the influence of the material and the channel on the push effect (i.e., the conversion probability of the above candidate material in different channels), and then push the information based on the determined channel and material, thereby improving the accuracy of the information push, and then improving the push effect of the information push; and, based on the conversion probability of the candidate material in different channels, determine the value expectations of different push plans (materials and channels), and determine the target channel and target material based on the value expectations and constraints, so that the push plan that maximizes the push value expectations and satisfies the constraints can be determined, further improving the accuracy of the information push.

[0091] See also Figure 5 , Figure 5 A schematic diagram of a flow chart of an information push method provided by another embodiment of the present application is shown. Figure 5 The process shown is described in detail, and the information push method may specifically include the following steps:

[0092] Step S410: determining candidate materials matching the user characteristic data of the target user from the multiple materials.

[0093] Step S420: predicting the conversion probability of the target user for the candidate material pushed through different channels as the conversion probability of the candidate material in different channels, wherein the conversion probability is used to characterize the probability of conversion behavior.

[0094] Step S430: Based on the conversion probabilities of the candidate materials in different channels, determine the target channels and target materials that meet the target conditions.

[0095] In the embodiment of the present application, steps S410 to S430 may refer to the contents of other embodiments and will not be described in detail here.

[0096] Step S440: Determine a push time corresponding to the target user based on the user behavior log data, historical interaction data, and real-time context information of the target user.

[0097] In an embodiment of the present application, before pushing the material, the push time can also be determined for the target user, so as to push the material at an appropriate time for the target user, reduce disturbance to the user, and improve the user experience. Among them, the computer device can determine the push time corresponding to the target user based on the user behavior log data, historical interaction data and real-time context information of the target user. The above user behavior log data, historical interaction data and real-time context information can be obtained from user feature data.

[0098] In some embodiments, the time period prediction model can be used to predict the user's active probability and interaction tendency in different time periods. The time period prediction model can be a random forest, gradient boosting tree or other models, and the specific model type can be unlimited; the personalized sending timing model can be trained based on the user's personalized data and behavior patterns, so as to determine the sending time suitable for the target user according to the model. In combination with the results output by the above model, real-time context information can also be integrated, such as geographic location, current device status (such as battery level, network connection status), user's current activity (such as whether driving), etc., to ensure that the material is sent in a suitable environment; the sending time can also be dynamically adjusted according to the real-time context. For example, when it is detected that the user is driving, the push notification is delayed to ensure safety.

[0099] In a possible implementation, for the time determined in the above manner, a multi-objective optimization algorithm can be used to balance the push effect and user experience to determine a push time that maximizes the conversion rate and does not disturb the user.

[0100] In one possible implementation, the computer device can also collect feedback from target users on different sending times, such as click-through rate, conversion rate, unsubscribe rate, etc., to evaluate the push effect; based on the evaluated push effect, the above models for determining the push time can be regularly updated and optimized, so that they can continuously learn and improve, thereby improving the accuracy and effect of the sending time selection.

[0101] Step S450: Pushing the target material to the device corresponding to the target user through the target channel at the push time.

[0102] In an embodiment of the present application, after the above push time is determined, the target material can be pushed to the device corresponding to the target user through the target channel when the current time is the determined push time.

[0103] In some implementations, for materials that need to be pushed to multiple users at the same time, a batch sending strategy may be adopted, such as sending in different time periods and batches, to avoid excessive system load and excessive user interference.

[0104] The information push method provided in the embodiment of the present application can, when pushing information, accurately determine the channel and the material for the candidate material that matches the user's characteristic data, combined with the influence of the material and the channel on the push effect (i.e., the conversion probability of the above candidate material under different channels), and then push the information according to the determined channel and material, thereby improving the accuracy of the information push, and then improving the push effect of the information push; in addition, a push time suitable for the target user is determined for the target user, and the determined material is pushed through the determined channel at the push time, thereby avoiding sending the material at a time that is not suitable for the user, reducing the disturbance to the user, and improving user satisfaction.

[0105] See also Figure 6 , Figure 6 A schematic diagram showing a flow chart of an information push method provided by another embodiment of the present application is shown. Figure 6 The process shown is described in detail, and the information push method may specifically include the following steps:

[0106] Step S510: Filter out materials that do not meet the push conditions from multiple materials to obtain filtered materials.

[0107] In the embodiment of the present application, when determining candidate materials for target users, materials that do not meet the push conditions can be filtered out from multiple materials to obtain filtered materials. The push conditions can be set according to laws and regulations, business rules, user preferences and taboos.

[0108] In some implementations, unqualified materials can be filtered out based on user feature data and in accordance with laws, regulations, and business rules. For example, minors cannot receive credit-related materials. Then, the types of materials that users explicitly indicate they do not want to receive can be further filtered to ensure user experience.

[0109] Step S520: If the number of the filtered materials is greater than or equal to the target number, then determining candidate materials that match the user feature data from the filtered materials.

[0110] In an embodiment of the present application, after filtering multiple materials to obtain filtered materials, the number of filtered materials can be compared with a target number, for example, the target number is 3, 5, 10, etc.; according to the comparison result, if the number of filtered materials is greater than or equal to the target number, it means that there are more materials left, so candidate materials that match the user feature data can be further determined from the filtered materials.

[0111] In some implementations, for the filtered materials, one or more of feature matching, similarity calculation, multi-level filtering and scoring mechanisms can be used to determine the material that best meets the current user needs from the filtered materials as the above candidate material.

[0112] Step S530: If the number of the filtered materials is less than the target number, the filtered materials are used as the candidate materials.

[0113] In an embodiment of the present application, after comparing the number of filtered materials with the target number, according to the comparison result, if the number of filtered materials is less than the target number, it means that the filtered materials are small, so the filtered materials can be directly used as candidate materials and processed in subsequent steps.

[0114] Step S540: predicting the conversion probability of the target user for the candidate material pushed through different channels as the conversion probability of the candidate material in different channels, wherein the conversion probability is used to characterize the probability of conversion behavior.

[0115] Step S550: Based on the conversion probabilities of the candidate materials in different channels, determine the target channels and target materials that meet the target conditions.

[0116] Step S560: Pushing the target material to the device corresponding to the target user through the target channel.

[0117] In the embodiment of the present application, steps S540 to S560 may refer to the contents of other embodiments and will not be described in detail here.

[0118] In some implementations, the computer device can also collect target users' interactive data on the material in real time, including open rate, click-through rate, conversion rate, etc., and record detailed logs for each transmission; analyze target users' feedback on different channels, push times, and pushed material content, and identify user behavior patterns and preferences.

[0119] Among them, the behavior path of the target user's electronic device after receiving the material, such as page dwell time, click behavior, purchase behavior, etc., can be fully tracked through event tracking technology and third-party data interfaces; the delivery results of the material can also be monitored in real time, abnormal situations such as delivery failure and delay can be detected, and the delivery strategy can be adjusted in time. Computer equipment can clean, deduplicate, and format the collected raw data to ensure the accuracy and integrity of the data, and then store the data, for example, using distributed databases (such as Hadoop and Cassandra) to store the processed data to provide support for subsequent analysis and model training. Computer equipment can use statistical analysis, machine learning and other technologies to conduct in-depth analysis of the collected feedback data and identify the advantages and disadvantages of the push strategy. Computer equipment can also automatically generate detailed push effect reports based on the analysis results, provide visual charts and key indicators (such as ROI, user retention rate, etc.), and help the operation team formulate optimization strategies.

[0120] The information push method provided in the embodiment of the present application can, when pushing information, accurately determine the channels and materials for the candidate materials that match the user's characteristic data, combined with the influence of the materials and channels on the push effect (i.e., the conversion probability of the above candidate materials under different channels), and then push the information based on the determined channels and materials, thereby improving the accuracy of the information push and further improving the push effect of the information push.

[0121] Next, we will combine the push scenario of marketing information. Figure 7 The information push method involved in the above embodiment is illustrated by examples.

[0122] See also Figure 7In the push scenario of marketing information, the original behavior data of users (i.e., user feature data) can be collected, and then pre-processed such as data cleaning, feature extraction, and data normalization can be performed; for multiple marketing materials in the marketing material pool, they can be filtered through the business baffle, that is, the marketing materials that do not meet the push conditions are filtered out; if the number of filtered marketing materials is greater than the threshold, the intelligent recall logic is further used to determine the candidate marketing materials that match the user feature data from the filtered marketing materials; if the number of filtered marketing materials is less than the threshold, the filtered marketing materials are directly used as candidate marketing materials; then, for the combination of each marketing material and different marketing channels, intelligent routing is performed, and the target prediction sub-model among the multiple prediction sub-models corresponding to the marketing channel is determined through the channel hierarchical model and the functional differentiation model, and the prediction sub-model and the user behavior characteristics are used to determine the target prediction sub-model. Data can be used to predict the conversion probability of marketing materials under different marketing channels, and then the conversion probability can be calibrated to obtain the calibrated conversion probability; based on the calibrated conversion probability, the expected marketing value can be calculated, and then according to the calculated expected marketing value and multiple constraints (i.e., channel information, frequency control rules, channel magnitude constraints, and user behavior data), the combination of marketing materials and marketing channels can be sorted through an intelligent sorting algorithm. The sorting results can also be adjusted based on marketing strategies and rules, and the target marketing materials and target marketing channels can be determined based on the adjusted sorting results; then, based on the user behavior log data, historical interaction data, and real-time context information of the target users, the push time corresponding to the target users is determined; finally, according to the sending plan (i.e., push time), the target marketing materials are pushed to the client in accordance with the determined target marketing channel. In addition, it can also feedback and collect target users' interactive data on marketing materials, track the behavioral paths of target users' electronic devices after receiving marketing materials, monitor the sending results of marketing materials in real time, detect abnormal situations such as sending failures and delays, and adjust sending strategies in a timely manner; clean, deduplicate, and format the collected raw data to ensure the accuracy and completeness of the data, and then store the data; analyze the collected feedback data, and automatically generate a detailed marketing effectiveness report based on the analysis results.

[0123] See also Figure 8, which shows a structural block diagram of an information push device 600 provided in an embodiment of the present application. The information push device 600 uses the above-mentioned computer device, and the information push device 600 includes: a first determination module 610, a probability prediction module 620, a second determination module 630 and a material push module 640. Among them, the first determination module 610 is used to determine the candidate material that matches the user feature data of the target user from a variety of materials; the probability prediction module 620 is used to predict the conversion probability of the target user for the candidate material pushed through different channels, as the conversion probability of the candidate material under different channels, and the conversion probability is used to characterize the probability of conversion behavior; the second determination module 630 is used to determine the target channel and target material that meet the target condition based on the conversion probability of the candidate material under different channels; the material push module 640 is used to push the target material to the device corresponding to the target user through the target channel.

[0124] In some embodiments, the probability prediction module 620 can be specifically used to determine each channel that matches the candidate material; utilize the prediction model corresponding to each channel, and based on the user characteristic data, predict the conversion probability of the target user for the candidate material pushed through each channel as the conversion probability of the candidate material under different channels.

[0125] In a possible implementation, the probability prediction module 620 can also be used to utilize a target prediction sub-model among the multiple prediction sub-models corresponding to each channel, and based on the user characteristic data, predict the conversion probability of the target user for the candidate material pushed through each channel as the conversion probability of the candidate material under different channels, wherein the multiple prediction sub-models correspond one-to-one to multiple prediction tasks, and the prediction task corresponding to the target prediction sub-model matches the business type of the material.

[0126] In some implementations, the information push device 600 may further include a probability calibration module. The probability calibration module may calibrate the conversion probability of the candidate material in different channels based on a conversion probability calibration model.

[0127] In some embodiments, the second determination module 630 can be specifically used to: determine the expected value of pushing the candidate material through different channels based on the conversion probability of each material in different channels; determine the target channel and target material that meet the target conditions based on the expected value of pushing the candidate material through different channels and constraints.

[0128] In a possible implementation, the second determination module 630 can also be used to sort multiple push schemes based on the expected value of pushing each candidate material through different channels, as well as constraints, to obtain a sorting result, wherein the push scheme includes channels and materials; determine a target push scheme from the sorting result, and use the channel in the target push scheme as the target channel, and use the material in the target push scheme as the target material.

[0129] In some implementations, the information push device 600 may further include a time determination module. The time determination module is used to determine the push time corresponding to the target user based on the user behavior log data, historical interaction data, and real-time context information of the target user before pushing the target material to the device corresponding to the target user through the target channel; the material push module 640 may be specifically used to push the target material to the device corresponding to the target user through the target channel at the push time.

[0130] In some embodiments, the first determination module 610 can be specifically used to: filter out materials that do not meet the push conditions from multiple materials to obtain filtered materials; if the number of the filtered materials is greater than or equal to the target number, determine candidate materials that match the user feature data from the filtered materials; if the number of the filtered materials is less than the target number, use the filtered materials as the candidate materials.

[0131] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices and modules can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here.

[0132] In several embodiments provided in the present application, the coupling between modules may be electrical, mechanical or other forms of coupling.

[0133] In addition, each functional module in each embodiment of the present application can be integrated into a processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The above integrated modules can be implemented in the form of hardware or software functional modules.

[0134] In summary, the solution provided by the present application determines the candidate material that matches the user characteristic data of the target user from a variety of materials, predicts the conversion probability of the target user for the candidate material pushed through different channels, and uses the conversion probability of the candidate material under different channels as the conversion probability. The conversion probability is used to characterize the probability of the conversion behavior. Based on the conversion probability of each material under different channels, the target channel and target material that meet the target conditions are determined, and the target material is pushed to the device corresponding to the target user through the target channel. Therefore, when pushing information, it is possible to accurately determine the channel and material for the candidate material that matches the user characteristic data, combined with the push effect of the material and the channel (i.e., the conversion probability of the above candidate material under different channels), thereby improving the accuracy of information push, and then improving the push effect of information push.

[0135] Please refer to Fig. 9 , which shows a structural block diagram of a computer device provided in an embodiment of the present application. The computer device 100 may be a server. The computer device 100 in the present application may include one or more of the following components: a processor 110, a memory 120, and one or more application programs, wherein the one or more application programs may be stored in the memory 120 and configured to be executed by one or more processors 110, and the one or more application programs are configured to execute the method described in the aforementioned method embodiment.

[0136] The processor 110 may include one or more processing cores. The processor 110 uses various interfaces and lines to connect various parts of the entire computer device 100, and executes various functions and processes data of the computer device 100 by running or executing instructions, programs, code sets or instruction sets stored in the memory 120, and calling data stored in the memory 120. Optionally, the processor 110 can be implemented in at least one hardware form of digital signal processing (DSP), field-programmable gate array (FPGA), and programmable logic array (PLA). The processor 110 can integrate one or a combination of a central processing unit (CPU), a graphics processing unit (GPU), and a modem. Among them, the CPU mainly processes the operating system, user interface, and application programs; the GPU is responsible for rendering and drawing display content; and the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor 110, but may be implemented separately through a communication chip.

[0137] The memory 120 may include a random access memory (RAM) or a read-only memory (ROM). The memory 120 may be used to store instructions, programs, codes, code sets or instruction sets. The memory 120 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for implementing at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the following various method embodiments, etc. The data storage area may also store data (such as a phone book, audio and video data, chat record data) created by the computer device 100 during use.

[0138] Please refer to Fig.10 , which shows a structural block diagram of a computer-readable storage medium provided in an embodiment of the present application. The computer-readable medium 800 stores program codes, which can be called by a processor to execute the method described in the above method embodiment.

[0139] The computer readable storage medium 800 may be an electronic memory such as a flash memory, an EEPROM (electrically erasable programmable read-only memory), an EPROM, a hard disk, or a ROM. Optionally, the computer readable storage medium 800 includes a non-transitory computer-readable storage medium. The computer readable storage medium 800 has storage space for program code 810 that performs any method step in the above method. These program codes can be read from or written to one or more computer program products. The program code 810 can be compressed, for example, in an appropriate form.

[0140] An embodiment of the present application also provides a computer program product, which includes a computer program, and the computer program is stored in a storage medium; a processor of a computer device reads the computer program from the storage medium, and the processor executes the computer program, so that the computer device executes the information push method provided in the above embodiment.

[0141] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An information push method, characterized in that: The method comprises: Determine candidate materials matching user characteristic data of target users from a plurality of materials; Predicting the conversion probability of the target user for the candidate material pushed through different channels as the conversion probability of the candidate material under different channels, wherein the conversion probability is used to represent the probability of conversion behavior; Based on the conversion probabilities of the candidate materials in different channels, determine target channels and target materials that meet target conditions; The material is pushed to the device corresponding to the target user through the target channel.

2. The method according to claim 1, characterized in that The predicting the conversion probability of the target user for the candidate material pushed through different channels as the conversion probability of the candidate material in different channels includes: Determine each channel that the candidate material matches; The prediction model corresponding to each channel is used, and based on the user characteristic data, the conversion probability of the target user for the candidate material pushed through each channel is predicted as the conversion probability of the candidate material under different channels.

3. The method according to claim 2, characterized in that The using the prediction model corresponding to each channel and based on the user feature data, predicting the conversion probability of the target user for the candidate material pushed through each channel as the conversion probability of the candidate material in different channels includes: Utilizing the target prediction sub-model among the multiple prediction sub-models corresponding to each channel and based on the user characteristic data, the conversion probability of the target user for the candidate material pushed through each channel is predicted as the conversion probability of the candidate material under different channels, wherein the multiple prediction sub-models correspond one-to-one to multiple prediction tasks, and the prediction task corresponding to the target prediction sub-model matches the business type of the material.

4. The method according to claim 1, characterized in that: Before determining the target channel and target material that meet the target condition based on the conversion probability of the candidate material in different channels, the method further includes: Based on the conversion probability calibration model, the conversion probability of the candidate material in different channels is calibrated.

5. The method according to claim 1, characterized in that The determining of target channels and target materials that meet target conditions based on the conversion probabilities of the candidate materials in different channels includes: Based on the conversion probabilities of the candidate materials in different channels, determining the expected value of pushing the candidate materials through different channels; Based on the expected value of pushing the candidate materials through different channels and the constraints, target channels and target materials that meet the target conditions are determined.

6. The method according to claim 5, characterized in that The step of determining target channels and target materials that meet target conditions based on the expected value of pushing the candidate materials through different channels and the constraints includes: Based on the expected value of pushing the candidate materials through different channels and the constraints, multiple push schemes are sorted to obtain a sorting result, where the push schemes include channels and materials; A target push scheme is determined from the sorting results, and the channel in the target push scheme is used as the target channel, and the material in the target push scheme is used as the target material.

7. The method according to any one of claims 1 to 6, characterized in that: Before pushing the target material to the device corresponding to the target user through the target channel, the method further includes: Determine a push time corresponding to the target user based on the user behavior log data, historical interaction data, and real-time context information of the target user; The pushing the target material to the device corresponding to the target user through the target channel includes: At the push time, the target material is pushed to the device corresponding to the target user through the target channel.

8. The method according to any one of claims 1 to 6, characterized in that: The step of determining a candidate material matching the user characteristic data of the target user from a plurality of materials includes: Filter out materials that do not meet the push conditions from multiple materials to obtain filtered materials; If the number of the filtered materials is greater than or equal to the target number, determining candidate materials matching the user feature data from the filtered materials; If the number of the filtered materials is less than the target number, the filtered materials are used as the candidate materials.

9. A computer device, characterized in that: include: one or more processors; Memory; One or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs are configured to execute the method according to any one of claims 1 to 8.

10. A computer program product, characterized in that The computer program product comprises a computer program, wherein the computer program is stored in a storage medium; The processor of the computer device reads the computer program from the storage medium, and the processor executes the computer program, so that the computer device executes the method according to any one of claims 1 to 8.

Images