Live broadcast room recommendation method and device, electronic equipment and storage medium

Abstract

The invention provides a live broadcast room recommendation method and device, electronic equipment and a storage medium, and relates to the field of artificial intelligence, in particular to the technical fields of NLP, deep learning, big data, cloud computing and the like. The specific implementation scheme is as follows: obtaining at least one target live broadcast room to be recommended to a target object; determining a display score of each target live broadcast room based on the interaction behavior information of the promotion article in each target live broadcast room in the plurality of first time periods; wherein the display score is used for indicating the display priority of the target live broadcast room; sorting the target live broadcasting rooms based on the display scores of the target live broadcasting rooms to obtain a live broadcasting room list; and recommending the live broadcast room list to the target object.

Description

Technical Field

[0001] The present disclosure relates to the field of AI (Artificial Intelligence), specifically to technical fields such as NLP (Natural Language Processing), deep learning, big data, cloud computing, and especially to methods, devices, electronic devices, and storage media for recommending live broadcast rooms. Background Art

[0002] With the rapid development of the times and the continuous enrichment and convenience of life shopping scenes, online consumption has surpassed the traditional static picture advertising form and ushered in a new interactive shopping model of live e-commerce. Unlike static product display, live e-commerce significantly enhances consumers' sense of participation and willingness to buy through real-time interaction, especially in the live broadcast room, where consumers can obtain more comprehensive and detailed information about items through the host's vivid introduction and instant interaction, which greatly promotes the conversion of promoted items.

[0003] In the unique environment of the live broadcast room, consumers can not only see the display of the promoted items intuitively, but also enhance their trust in and desire to buy the promoted items through the host's professional recommendations and real-time Q&A. This immersive shopping experience not only broadens the exposure channels for promoted items, but also effectively improves the conversion rate of promoted items, making live broadcast e-commerce an important force in promoting sales growth.

[0004] However, in this emerging consumption scenario, it is particularly important to optimize the recommendation ranking of the live broadcast room to enhance the user experience, help e-commerce platforms better understand market trends, and ensure the effective allocation of resources. Summary of the invention

[0005] The present disclosure provides a method, device, electronic device and storage medium for recommending a live broadcast room.

[0006] According to one aspect of the present disclosure, a live broadcast room recommendation method is provided, comprising:

[0007] Obtain at least one target live broadcast room to be recommended to the target object;

[0008] Determine a display score of each target live broadcast room based on the interactive behavior information of the promoted items in each target live broadcast room within a plurality of first time periods; wherein the display score is used to indicate the display priority of the target live broadcast room;

[0009] Based on the display scores of the target live broadcast rooms, the target live broadcast rooms are sorted to obtain a live broadcast room list;

[0010] Recommend the live broadcast room list to the target object.

[0011] According to another aspect of the present disclosure, a live broadcast room recommendation device is provided, comprising:

[0012] An acquisition module, used for acquiring at least one target live broadcast room to be recommended to a target object;

[0013] A determination module, configured to determine a display score of each of the target live broadcast rooms based on the interactive behavior information of the promoted items in each of the target live broadcast rooms within a plurality of first time periods; wherein the display score is used to indicate the display priority of the target live broadcast room;

[0014] A sorting module, used to sort the target live broadcast rooms based on the display scores of the target live broadcast rooms to obtain a live broadcast room list;

[0015] A recommendation module is used to recommend the live broadcast room list to the target object.

[0016] According to another aspect of the present disclosure, there is provided an electronic device, comprising:

[0017] at least one processor; and

[0018] a memory communicatively connected to the at least one processor; wherein,

[0019] 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 so that the at least one processor can execute the live broadcast room recommendation method proposed in the above aspect of the present disclosure.

[0020] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium of computer instructions is provided, wherein the computer instructions are used to enable the computer to execute the live broadcast room recommendation method proposed in the above-mentioned first aspect of the present disclosure.

[0021] According to another aspect of the present disclosure, a computer program product is provided, including a computer program, which, when executed by a processor, implements the live broadcast room recommendation method proposed in the above aspect of the present disclosure.

[0022] 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

[0023] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0024] Figure 1 A flowchart of a live broadcast room recommendation method provided in the first embodiment of the present disclosure;

[0025] Figure 2 A flowchart of a live broadcast room recommendation method provided in Embodiment 2 of the present disclosure;

[0026] Figure 3 This is a flowchart of a live broadcast room recommendation method provided in Embodiment 3 of the present disclosure;

[0027] Figure 4 A flowchart of a live broadcast room recommendation method provided in Embodiment 4 of the present disclosure;

[0028] Figure 5 A flowchart of a live broadcast room recommendation method provided in Embodiment 5 of the present disclosure;

[0029] Figure 6 A schematic diagram of the prediction principle of the commodity conversion amount (i.e., the total commodity transaction amount) provided in the embodiment of the present disclosure;

[0030] Figure 7 This is a schematic diagram of the structure of the live broadcast room recommendation device provided in the sixth embodiment of the present disclosure;

[0031] Figure 8 A schematic block diagram of an example electronic device that can be used to implement embodiments of the present disclosure is shown. DETAILED DESCRIPTION

[0032] 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, descriptions of well-known functions and structures are omitted in the following description.

[0033] In the consumption scenario of live e-commerce, the display score of each live room can be calculated according to the predicted conversion amount of the promoted items in each live room, and the live rooms can be ranked based on the display scores of each live room to enhance user experience and improve the overall efficiency of the e-commerce platform.

[0034] Taking promoted items as an example, the GMV (Gross Merchandise Volume) of the products can be predicted, and the live broadcast rooms can be ranked based on the GMV of the products.

[0035] However, how to predict the predicted conversion amount (such as GMV) of promoted items has become an urgent problem to be solved. The main challenges include at least:

[0036] 1. Multi-dimensional data sources: E-commerce live broadcasting involves many factors, such as the price differences of promotional items mounted by different anchors, and the discrete multi-peak distribution of GMV in different live broadcast rooms. This makes the sample learning difficulty of different live broadcast rooms different, and the prediction difficulty is relatively high.

[0037] 2. The integration of real-time data and historical data: During live broadcasts, data changes very quickly. Real-time data reflects the current situation and usually has strong volatility and short-term impact. Although historical data can provide stable long-term trends, it often cannot reflect short-term changes in a timely manner. How to effectively combine the advantages of both is the key to solving accurate predictions and personalized recommendations.

[0038] Specifically, real-time data includes audience real-time behavior, anchor performance, interaction frequency, flash sales, etc., reflecting short-term changes and immediate market dynamics; historical data shows user behavior habits, seasonal fluctuations, and sales trends through long-term accumulation. The characteristics of the two are different. Real-time data is highly volatile and may contain noise, while historical data has a lag and is difficult to capture short-term sudden changes in time. Therefore, how to effectively integrate these two types of data to improve prediction accuracy has become a huge challenge.

[0039] In the related technologies, in order to cope with the difficulties in predicting the conversion amount of promotional items in the live broadcast room, the main solutions include:

[0040] The first one is the weighted fusion method: by assigning different weights to real-time data and historical data, the two are weightedly fused to balance their respective impacts.

[0041] However, this approach cannot flexibly respond to the rapidly changing needs in live e-commerce scenarios. Specifically, the relationship between real-time data and historical data is dynamic, and the influence of different anchors, different products, and different time periods varies. Using fixed weights cannot accurately reflect this change, which may cause the model to be overly dependent on a certain type of data, thereby affecting the accuracy and stability of the prediction results.

[0042] The second type is time series modeling and deep learning methods: Time series modeling techniques such as LSTM (Long Short-Term Memory) and GRU (Gated Recurrent Unit) can capture long-term dependencies in historical data and reflect short-term fluctuations when real-time data is updated. LSTM maintains long-term historical information through a gating mechanism while adapting to rapidly changing real-time data, so it is suitable for GMV prediction tasks.

[0043] However, although time series models such as LSTM and GRU can capture long-term dependencies, these time series models often respond lags when processing highly volatile and rapidly changing real-time data. In addition, the existing time series models are time-consuming in their training and reasoning processes, so there is a bottleneck in real-time performance and may not meet the needs of real-time prediction.

[0044] The third method is the model integration method: through the collaborative work of multiple models, the impact of historical trends and real-time data are processed separately, and finally the prediction results are integrated through weighted fusion or voting mechanism.

[0045] However, this method requires independent calculation of historical data and real-time data, resulting in higher computing costs and training time, and may have certain practical problems in real-time and subsequent maintenance.

[0046] The fourth method is reinforcement learning: through interaction with the environment, reinforcement learning can adjust model parameters in real time, optimize recommendation strategies and resource allocation. Combining historical data with real-time data, reinforcement learning methods can better respond to market changes dynamically and optimize forecasting results.

[0047] However, since the results of reinforcement learning are deeply affected by the input data, the challenges of real-time data, policy stability issues and sparse rewards may lead to unstable performance in practical applications.

[0048] Therefore, in response to at least one of the above-mentioned problems, the present disclosure proposes a live broadcast room recommendation method, device, electronic device and storage medium.

[0049] The following describes the live broadcast room recommendation method, device, electronic device and storage medium of the embodiments of the present disclosure with reference to the accompanying drawings.

[0050] Figure 1 This is a flow chart of the live broadcast room recommendation method provided in the first embodiment of the present disclosure.

[0051] The embodiment of the present disclosure takes the example of the live broadcast room recommendation method being configured in a live broadcast room recommendation device. The live broadcast room recommendation device can be applied to any electronic device so that the electronic device can perform the live broadcast room recommendation function.

[0052] Among them, the electronic device can be any device with computing capabilities, such as a personal computer, a mobile terminal, a server, etc. The mobile terminal can be, for example, a mobile phone, a tablet computer, a personal digital assistant, a wearable device, and other hardware devices with various operating systems, touch screens and / or display screens.

[0053] like Figure 1 As shown, the live broadcast room recommendation method may include the following steps S101 to S104:

[0054] Step S101, obtaining at least one target live broadcast room to be recommended to a target object.

[0055] The target object may be any user object. For example, the target object may be a user object that is currently logged into the e-commerce platform and the e-commerce platform is in the foreground running state.

[0056] In an embodiment of the present disclosure, a target live broadcast room to be recommended to a target object can be determined from multiple live broadcast rooms on an e-commerce platform based on a resource recommendation algorithm; wherein the target live broadcast room can also be referred to as a recommended live broadcast room. Exemplarily, the target live broadcast room can be a live broadcast room that the target object is interested in.

[0057] Step S102, determining the display score of each target live broadcast room based on the interactive behavior information of the promoted items in each target live broadcast room within multiple first time periods.

[0058] Among them, promotional items include but are not limited to: physical goods, virtual goods and services, etc. For example, promotional items may include online courses, subscription services, digital products, etc.

[0059] The first period includes the historical period before the current period. It should be noted that the present disclosure does not limit the length of the first period. Exemplarily, the length of the first period may be half a day, one day, etc.

[0060] The multiple first time periods may be continuous time periods or discontinuous time periods, and the embodiments of the present disclosure do not limit this.

[0061] The number of the first time periods may be pre-set. For example, the number of the first time periods may be 6, 7, 8, etc.

[0062] The display score is used to indicate the display priority of the target live broadcast room, and the display score is positively correlated with the display priority, that is, the higher the display score, the higher the display priority, and conversely, the lower the display score, the lower the display priority.

[0063] The interactive behavior information includes various details of interactions between viewers in the target live broadcast room and the promoted items, including but not limited to browsing behavior, clicking behavior, following behavior, collecting behavior, purchasing behavior or conversion behavior, etc. In addition, the interactive behavior information also includes various details of interactions between the anchor in the target live broadcast room and the promoted items, including but not limited to: display and introduction of the promoted items, real-time interaction and Q&A, recommendation and guidance, feedback and response, etc.

[0064] In the disclosed embodiment, for any target live broadcast room, the display score of the target live broadcast room can be calculated based on the interactive behavior information of all promotional items in the live broadcast room during multiple first time periods.

[0065] Exemplarily, based on deep learning technology, the display score of the target live broadcast room can be calculated according to the interactive behavior information of all promotional items in each live broadcast room during multiple first time periods.

[0066] Step S103, sorting the target live broadcast rooms based on the display scores of the target live broadcast rooms to obtain a live broadcast room list.

[0067] Exemplarily, each target live broadcast room may be sorted from large to small according to the display score to obtain a live broadcast room list.

[0068] Step S104, recommending a list of live broadcast rooms to the target object.

[0069] Exemplarily, a list of live broadcast rooms may be pushed to the target object's account on the e-commerce platform.

[0070] The live broadcast room recommendation method of the disclosed embodiment obtains the target live broadcast room to be recommended to the target object, and the target live broadcast room can be the live broadcast room that the target object is interested in, so as to ensure that the recommended content is highly matched with the preferences and needs of the target object. This precise recommendation can improve the satisfaction and viewing experience of the target object. Based on the interactive behavior information of the promoted items in each target live broadcast room in multiple first time periods, the display score of each target live broadcast room is determined, and the live broadcast room list is obtained by sorting accordingly. The live broadcast rooms that are more popular and more likely to attract the attention of the target object can be placed in the front, which not only helps the target object to quickly find the live broadcast room of interest, improves viewing efficiency, and increases the exposure rate and conversion rate of the live broadcast room, but also the e-commerce platform can optimize resource allocation and allocate more traffic and exposure opportunities to those live broadcast rooms that are more likely to produce positive results, thereby improving the overall benefits and user satisfaction of the e-commerce platform.

[0071] It should be noted that in the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision and disclosure of user personal information are all carried out with the user's consent, and are in compliance with the relevant laws and regulations and do not violate public order and good morals.

[0072] In order to clearly explain how any embodiment of the present disclosure determines the display score of each target live broadcast room based on the interactive behavior information of the promoted items in each target live broadcast room within multiple first time periods, the present disclosure also proposes a live broadcast room recommendation method.

[0073] Figure 2 This is a flow chart of the live broadcast room recommendation method provided in the second embodiment of the present disclosure.

[0074] like Figure 2 As shown, the live broadcast room recommendation method may include the following steps S201 to S206:

[0075] Step S201, obtaining at least one target live broadcast room to be recommended to a target object.

[0076] It should be noted that the explanation of step S201 can refer to the relevant description in any embodiment of the present disclosure, and will not be repeated here.

[0077] Step S202, for any target live broadcast room, based on the interactive behavior information of any promotional item in any target live broadcast room in multiple first time periods, predict the first indicator value of the first key indicator of any promotional item in the second time period.

[0078] The plurality of first time periods are located before the second time period. Exemplarily, the second time period includes the current time period, wherein the second time period may be the same as the first time period, or may be different in length, which is not limited in the embodiment of the present disclosure.

[0079] There is no limit on the number of promotional items in each target live broadcast room, that is, the number of promotional items in each target live broadcast room can be one or more.

[0080] The first indicator value of the first key indicator of the promotional item in the second time period is used to indicate the sales potential and market acceptance of the promotional item in the second time period.

[0081] In the embodiments of the present disclosure, for any promotional item in any target live broadcast room, deep learning technology can be used to predict the first indicator value of the first key indicator of the promotional item in the second time period based on the interactive behavior information of the promotional item in the target live broadcast room in multiple first time periods.

[0082] As an example, taking the promoted item as a commodity and the first key indicator as GMV, a conversion amount prediction model (such as the GMV tower) can be used to predict the first indicator value of the GMV of the promoted item in the second time period (i.e., the predicted conversion amount) based on the interactive behavior information of the promoted item in multiple first time periods.

[0083] Step S203, determining the traffic value score of any promoted item according to the first indicator value of the first key indicator.

[0084] Among them, the traffic value score can be used to indicate the attractiveness and interactivity of the promoted items, the user's willingness to purchase the promoted items, the matching degree between the user and the promoted items, the monetization ability and sales potential of the promoted items, the cost-effectiveness and long-term value of the promoted items, etc.

[0085] In the disclosed embodiment, the traffic value score of the promotional item can be calculated according to the first indicator value of the first key indicator of the promotional item in the second time period. The traffic value score is positively correlated with the first indicator value of the first key indicator, that is, the larger the first indicator value, the higher the traffic value score, and vice versa, the smaller the first indicator value, the lower the traffic value score.

[0086] Taking the promoted items as commodities and the first key indicator as GMV as an example, the traffic value score can be eCPM (effective cost per mille, effective revenue per thousand impressions), among which the values ​​of eCPM and GMV are positively correlated.

[0087] Step S204, determining the display score of any target live broadcast room according to the traffic value score of each promoted item in any target live broadcast room.

[0088] Among them, the display score is used to indicate the display priority of the target live broadcast room.

[0089] In the disclosed embodiment, for any target live broadcast room, the display score of the target live broadcast room can be calculated according to the traffic value scores of all promoted items in the target live broadcast room, wherein the display score is positively correlated with the traffic value score of each promoted item.

[0090] As a possible implementation method, when there are multiple promotional items in the target live broadcast room, the average of the traffic value scores of the promotional items in the target live broadcast room can be used as the display score of the target live broadcast room.

[0091] As another possible implementation method, when there are multiple promotional items in the target live broadcast room, the sum of the traffic value scores of the promotional items in the target live broadcast room can be used as the display score of the target live broadcast room.

[0092] As another possible implementation method, when there are multiple promotional items in the target live broadcast room, the traffic value score of each promotional item in the target live broadcast room can be weighted and summed to obtain the display score of the target live broadcast room.

[0093] Therefore, different methods can be used to calculate the display score of each target live broadcast room to improve the flexibility and applicability of the method.

[0094] Step S205: sort the target live broadcast rooms based on the display scores of the target live broadcast rooms to obtain a live broadcast room list.

[0095] Step S206, recommending a list of live broadcast rooms to the target object.

[0096] It should be noted that the explanation of steps S205 to S206 can be found in the relevant description in any embodiment of the present disclosure and will not be repeated here.

[0097] The live broadcast room recommendation method of the disclosed embodiment calculates the first indicator value of the first key indicator of the same promotional item in the second time period after the multiple first time periods based on the interactive behavior information of the same promotional item in the target live broadcast room in multiple first time periods, and calculates the traffic value score of the same promotional item accordingly, which can improve the stability and reliability of the traffic value score calculation, and then calculate the display score of the target live broadcast room based on the accurate traffic value score corresponding to each promotional item in the target indicator room, which can reflect the overall attractiveness and conversion efficiency of the target live broadcast room. This helps the e-commerce platform or operation team to prioritize among multiple target live broadcast rooms, give priority to displaying those live broadcast rooms with greater potential and appeal, improve resource utilization efficiency, and enhance user experience and satisfaction.

[0098] In order to clearly illustrate how the above embodiment predicts the first indicator value of the first key indicator of any promotional item in the target live broadcast room in a second time period based on the interactive behavior information of the promotional item in multiple first time periods, the present disclosure also proposes a live broadcast room recommendation method.

[0099] Figure 3 This is a flowchart of the live broadcast room recommendation method provided in the third embodiment of the present disclosure.

[0100] like Figure 3 As shown, the live broadcast room recommendation method may include the following steps S301 to S308:

[0101] Step S301, obtaining at least one target live broadcast room to be recommended to the target object.

[0102] It should be noted that the explanation of step S301 can refer to the relevant description in any embodiment of the present disclosure, and will not be repeated here.

[0103] Step S302, for any promotional item in any target live broadcast room, vectorize the interactive behavior information of any promotional item in multiple first time periods to obtain a first behavior vector.

[0104] Exemplarily, for any promotional item in any target live broadcast room, an embedding network can be used to vectorize the interactive behavior information of the promotional item in multiple first time periods to obtain a first behavior vector.

[0105] Step S303: Based on the gating mechanism, weighted processing is performed on the first behavior vector to obtain a second behavior vector.

[0106] In the embodiment of the present disclosure, different elements in the first behavior vector may be weighted based on a gating mechanism to obtain a processed first behavior vector, which is recorded as a second behavior vector in the present disclosure.

[0107] In any embodiment of the present disclosure, the weighting process of the first behavior vector may include the following steps A to B:

[0108] Step A: Inputting the first behavior vector into a neural gate to obtain at least one gating signal output by the neural gate; wherein the gating signal is used to indicate the weighting degree of different elements in the first behavior vector.

[0109] The value range of the gating signal may be between 0 and 1, indicating the weighting degree of different elements in the first behavior vector, that is, the gating signal determines which elements should be emphasized or weakened in subsequent processing.

[0110] Step B: Perform weighted processing on different elements in the first behavior vector based on the gating signal to obtain a second behavior vector.

[0111] As a result, the gating mechanism can dynamically adjust the weights of each element in the behavior vector according to the context, enhancing the flexibility and adaptability of the model; the gating mechanism allows the model to selectively focus on the most important features in the behavior vector and ignore irrelevant or interfering information, which enables the model to more effectively capture the most relevant information to the task and improve the accuracy and stability of the prediction. Since the gating mechanism can quickly identify and focus on key features, it helps to accelerate the convergence speed during training and save computing resources.

[0112] Step S304: predicting a first indicator value of a first key indicator of any promotional item in a second time period according to the second behavior vector.

[0113] Among them, a plurality of first time periods are located before the second time period.

[0114] In the embodiments of the present disclosure, for any promotional item in any target live broadcast room, deep learning technology can be used to predict the first indicator value of the first key indicator of the promotional item in the second time period based on the second behavior vector of the promotional item in the target live broadcast room.

[0115] As an example, taking the promoted item as a commodity and the first key indicator as GMV, a conversion amount prediction model (such as the GMV tower) can be used to predict the first indicator value of the GMV of the promoted item in the second time period (i.e., the predicted conversion amount) based on the second behavior vector of the promoted item.

[0116] Step S305, determining the traffic value score of any promoted item according to the first indicator value of the first key indicator.

[0117] Step S306, determining the display score of any target live broadcast room according to the traffic value score of each promoted item in any target live broadcast room.

[0118] Among them, the display score is used to indicate the display priority of the target live broadcast room.

[0119] Step S307: sort the target live broadcast rooms based on the display scores of the target live broadcast rooms to obtain a live broadcast room list.

[0120] Step S308, recommending a list of live broadcast rooms to the target object.

[0121] It should be noted that the explanation of steps S305 to S308 can be found in the relevant description in any embodiment of the present disclosure and will not be repeated here.

[0122] The live broadcast room recommendation method of the disclosed embodiment converts the interactive behavior information of the promoted items in multiple first time periods into vector form, which can greatly improve the efficiency of data processing, more accurately capture the behavioral characteristics related to the promoted items, and use a gating mechanism to dynamically adjust the importance of different behavioral characteristics. It can highlight the impact of key features on the prediction results, while weakening the impact of non-key features, thereby improving the accuracy of the prediction.

[0123] In order to clearly illustrate how any embodiment of the present disclosure determines the traffic value score of a promoted item based on the first indicator value of the first key indicator of the promoted item in the second time period, the present disclosure also proposes a live broadcast room recommendation method.

[0124] Figure 4 This is a flowchart of the live broadcast room recommendation method provided in the fourth embodiment of the present disclosure.

[0125] like Figure 4 As shown, the live broadcast room recommendation method may include the following steps S401 to S407:

[0126] Step S401, obtaining at least one target live broadcast room to be recommended to the target object.

[0127] Step S402, for any target live broadcast room, based on the interactive behavior information of any promotional item in any target live broadcast room in multiple first time periods, predict the first indicator value of the first key indicator of any promotional item in the second time period.

[0128] Among them, a plurality of first time periods are located before the second time period.

[0129] It should be noted that the explanation of steps S401 to S402 can be found in the relevant description in any embodiment of the present disclosure and will not be repeated here.

[0130] Step S403, determining the mean and standard deviation of the first key indicator of any promotional item in the multiple first time periods according to the interactive behavior information of any promotional item in the multiple first time periods.

[0131] In the disclosed embodiment, for any promotional item in any target live broadcast room, the mean and standard deviation of the first key indicator of the promotional item in multiple first time periods can be calculated based on the interactive behavior information of the promotional item in multiple first time periods.

[0132] That is, the index value of the first key indicator of the promotional item in each first time period can be determined based on the interactive behavior information of the promotional item in each first time period, and the mean and standard deviation of the index value of the first key indicator of the promotional item in multiple first time periods can be calculated accordingly.

[0133] Step S404, determining the traffic value score of any promoted item according to the first indicator value, mean value and standard deviation of the first key indicator.

[0134] Exemplarily, the first indicator value of the first key indicator can be corrected according to the mean and standard deviation of the first key indicator to obtain a corrected first indicator value, which is recorded as the second indicator value in the present disclosure, so that the traffic value score of the above-mentioned promotional item can be calculated according to the second indicator value of the first key indicator.

[0135] Among them, the traffic value score is positively correlated with the second indicator value.

[0136] In any embodiment of the present disclosure, the traffic value score of the promoted item can be calculated by following steps C to E:

[0137] Step C: According to the mean and standard deviation of the first key indicator, the first indicator value of the first key indicator is corrected to obtain the second indicator value of the first key indicator.

[0138] For example, taking the promoted item as a commodity and the first key indicator as GVM, the standardized transformation z-score technology can be used to correct the GVM, that is, the corrected GMV is:

[0139] Among them, μ represents the mean and δ represents the standard deviation.

[0140] Step D: Obtain the third indicator value of the second key indicator of the above-mentioned promotional item in the second time period.

[0141] Among them, the second key indicator includes but is not limited to: CTR (Click-Through-Rate), CVR (Conversion Rate) and ROI (Return on Investment).

[0142] For example, taking the promoted items as commodities, the ROI may be given by the merchant, and the CTR and CVR may be obtained through real-time monitoring.

[0143] Step E: Calculate the traffic value score of the above-mentioned promotional item according to the second indicator value of the first key indicator and the third indicator value of the second key indicator.

[0144] As an example, when the first key indicator includes the total transaction amount of the item, and the second key indicator includes CTR, CVR and ROI, the display conversion rate of the above-mentioned promoted item can be determined according to the product of the third indicator value of CTR and the third indicator value of CVR, and the total transaction amount corresponding to the unit return on investment of the promoted item can be determined according to the ratio of the second indicator value of the total transaction amount of the item and the third indicator value of ROI. Therefore, the traffic value score of the promoted item can be determined according to the product of the display conversion rate and the total transaction amount corresponding to the unit return on investment.

[0145] Taking the promoted item as a commodity as an example, the total transaction amount of the item can be GMV, and the traffic value score can be eCPM. The following formula can be used to calculate the eCPM of the promoted item:

[0146]

[0147] Therefore, by comprehensively considering the total transaction amount, CTR, CVR and ROI of the items to calculate the traffic value score of the promoted items, the rationality and reliability of the traffic value score calculation can be improved, thereby improving the rationality of the live broadcast room sorting and improving the user experience in the live broadcast room recommendation scenario.

[0148] In summary, the z-score standardization technology is used to adjust the distribution of the first key indicator, so that the distribution of the first key indicator is closer to the normal distribution, thereby improving the accuracy of data fitting, reducing the calculation deviation caused by factors such as different anchors and promotional items, and thus improving the accuracy of the calculation of the traffic value score of the promotional item. Furthermore, the traffic value score of the promotional item is calculated not only based on the first key indicator, but also based on the second key indicator of the promotional item in the second time period, which can improve the rationality and reliability of the calculation of the traffic value score of the promotional item.

[0149] Step S405, determining the display score of any target live broadcast room according to the traffic value score of each promoted item in any target live broadcast room.

[0150] Step S406: sort the target live broadcast rooms based on the display scores of the target live broadcast rooms to obtain a live broadcast room list.

[0151] Step S407, recommending a live broadcast room list to the target object.

[0152] It should be noted that the explanation of steps S405 to S407 can be found in the relevant description in any embodiment of the present disclosure and will not be repeated here.

[0153] The live broadcast room recommendation method of the disclosed embodiment comprehensively calculates the traffic value score of the promoted item by combining the mean and standard deviation of the first key indicator of the promoted item in multiple first time periods and the first indicator value of the first key indicator of the promoted item in the second time period, which can improve the rationality and accuracy of the calculation results.

[0154] In order to clearly explain how to obtain a target live broadcast room to be recommended to a target object in any embodiment of the present disclosure, the present disclosure also proposes a live broadcast room recommendation method.

[0155] Figure 5 This is a flowchart of the live broadcast room recommendation method provided in the fifth embodiment of the present disclosure.

[0156] like Figure 5 As shown, the live broadcast room recommendation method may include the following steps S501 to S506:

[0157] Step S501, obtain the portrait label system of the target object, and determine the candidate live broadcast room in the live broadcast state from multiple live broadcast rooms.

[0158] Among them, the portrait label system is a set of label information used to describe the user characteristics and behavior patterns of the target object, including but not limited to: basic attributes (such as demographic characteristics, social attributes, etc.), statistical characteristics (such as the target object's activity on the e-commerce platform, whether the target object is a new or old user of the e-commerce platform, etc.), the group information to which the target object belongs (such as exquisite women, etc.), predictive indicators (for example, when the target object is a new user of the e-commerce platform, indicators predicted by deep learning technology), etc.

[0159] Among them, the live broadcast room can be a live broadcast room of an e-commerce platform, and the candidate live broadcast room can be a live broadcast room in the e-commerce platform that is in a live broadcast state.

[0160] In any embodiment of the present disclosure, the image label system of the target object can be obtained by following steps F to H:

[0161] Step F: Obtain historical behavior data of the target object in multiple business fields associated with the e-commerce platform.

[0162] Among them, historical behavior data includes but is not limited to: the target object's purchase history, browsing history, search history, etc. on the e-commerce platform, which is used to indicate the target object's short-term and long-term resource preferences, item preferences, etc.

[0163] Among them, multiple business fields include but are not limited to: live e-commerce, video e-commerce, search e-commerce, search, feed (information flow) and other fields.

[0164] Exemplarily, the multiple business fields may include at least one first business field of an e-commerce platform and at least one second business field of an information service platform, and the information service platform is used to provide a traffic entrance for the e-commerce platform. The first business field includes but is not limited to: live e-commerce, video e-commerce, search e-commerce and other fields; the second business field includes but is not limited to: search, feed (information flow) and other fields.

[0165] Among them, the information service platform is not only used to provide information services, but also to provide traffic entrances for e-commerce platforms, promoting the conversion of users from information browsing to shopping behavior.

[0166] Therefore, when constructing a user portrait label system on an e-commerce platform, a more comprehensive perspective can be obtained to understand the interests and needs of users by comprehensively considering the historical behavior data of users in various business fields within the e-commerce platform (such as live e-commerce, video e-commerce, search e-commerce, etc.) and various business fields of the external information service platform that provides traffic entrances (such as search, information flow, etc.). This includes not only the purchase intentions directly expressed by users on the e-commerce platform, but also the browsing, searching and interactive behaviors of users on external platforms, which helps to discover users' potential needs, thereby providing users with more accurate recommendations and services, and improving user experience and satisfaction.

[0167] Step G: Obtain an initial label system of the target object based on the basic information of the target object.

[0168] Among them, basic information includes but is not limited to: demographic characteristics, social attributes, etc.

[0169] Step H: Based on the historical behavior data of multiple business scenarios, the initial label system is updated to obtain a portrait label system of the target object in multiple dimensions.

[0170] Among them, multiple dimensions include but are not limited to the following dimensions: basic attribute dimension, statistical feature dimension, user grouping dimension, predictive indicator dimension, etc.

[0171] In the embodiments of the present disclosure, the historical behavior data of the target object in multiple business fields associated with the e-commerce platform can be processed based on statistical analysis methods and deep learning methods to obtain portrait labels of the target object in multiple dimensions, and the initial label system of the target object can be updated using the portrait labels of the first object in multiple dimensions to obtain a portrait label system of the target object in multiple dimensions.

[0172] In summary, the initial label system is generated based on the basic information of the target object, which can provide a basic user profile. Through the historical behavior data of the target object in multiple business fields associated with the e-commerce platform, such as purchase records, browsing habits, search keywords, etc., the initial label system can be further updated, which can refine the interest preferences and consumption patterns of the target object, thereby more accurately locating the needs of the target object and improving the target object's usage experience in the live broadcast room recommendation scenario.

[0173] Step S502, obtaining the live broadcast room features of multiple candidate live broadcast rooms and the item features of the promoted items in the multiple candidate live broadcast rooms.

[0174] Among them, the live broadcast room features are feature vectors obtained by extracting features from candidate live broadcast rooms. The live broadcast room features include but are not limited to content features, behavior features, context features, theme features (such as the main topics or categories of the live broadcast room), and target groups of the candidate live broadcast rooms.

[0175] Among them, content features include but are not limited to: visual elements (including visual information such as the host's appearance, background settings, displayed products or scenes, etc.), audio elements (including the host's voice characteristics, music usage, and the sound of audience interaction, etc.), and text elements (including the live broadcast room title, description, and text information in barrage comments, etc.). Behavioral features include but are not limited to: user interaction, viewing behavior, transaction behavior, etc. Contextual features include but are not limited to: time features (such as the start time of the live broadcast, the duration of the live broadcast, etc.), spatial features (such as the host's location), and social network features (such as the host's social influence, fan base, and their activity, etc.).

[0176] Among them, the item feature is a feature vector obtained by extracting features from the promoted item.

[0177] In any embodiment of the present disclosure, the item features of any promotional item can be obtained by following steps I to K:

[0178] Step I: Query the knowledge graph associated with the promotional items in multiple candidate live broadcast rooms; wherein the nodes in the knowledge graph are used to indicate the promotional items, and the edges are used to indicate the association relationship between the nodes.

[0179] The knowledge graph is constructed based on the item information of all promoted items on the e-commerce platform. The nodes in the knowledge graph are used to indicate the promoted items, and the edges are used to indicate the association between different nodes. Taking the promoted items as an example, the knowledge graph can also be called the commodity knowledge graph.

[0180] Step J: For any candidate live broadcast room among multiple candidate live broadcast rooms, extract a sub-graph associated with the promotional items in the candidate live broadcast room from the knowledge graph.

[0181] The sub-graph includes the promotional items in the candidate live broadcast room and other promotional items that are associated with the promotional items in the candidate live broadcast room.

[0182] Step K: Based on the feature extraction algorithm, feature extraction is performed on the above sub-graph to obtain the item features of the promoted item in the candidate live broadcast room.

[0183] In summary, by extracting sub-graphs, we can capture various information related to the promoted items, including direct and indirect associations. This correlation analysis helps to reveal the intrinsic connections between promoted items, such as substitutes, complements or competitive relationships, thereby providing richer, more diverse and accurate item features. It should be understood that with the addition of new promoted items, the knowledge graph and sub-graph can be continuously updated and improved, which enables the feature extraction process to keep pace with the times and adapt to the ever-changing market environment.

[0184] Step S503, determining a target live broadcast room from multiple candidate live broadcast rooms based on the portrait label system, the live broadcast room features of the multiple candidate live broadcast rooms, and the item features of the promoted items in the multiple candidate live broadcast rooms.

[0185] As an example, the object features of the target object can be determined based on the portrait label system, and the deep learning technology can be used to predict the recommendation probabilities of multiple candidate live broadcast rooms based on the item features of the promoted items in multiple candidate live broadcast rooms, the room features of multiple candidate live broadcast rooms, and the object features of the target object, wherein the recommendation probability is used to indicate the target object's preference for the candidate live broadcast rooms and the promoted items in the candidate live broadcast rooms. Therefore, in the present disclosure, the target live broadcast room that the target object is interested in can be determined from multiple candidate live broadcast rooms based on the recommendation probabilities of multiple candidate live broadcast rooms.

[0186] Exemplarily, a candidate live broadcast room with a recommendation probability higher than a set probability threshold may be selected as a target live broadcast room.

[0187] Exemplarily, multiple candidate live broadcast rooms can be sorted from large to small according to the recommendation probability to obtain a sorting sequence, so that in the present disclosure, a set number of candidate live broadcast rooms ranked first in the sorting sequence can be used as target live broadcast rooms. For example, taking the set number as N as an example, the top N candidate live broadcast rooms in the sorting sequence can be used as target live broadcast rooms.

[0188] Exemplarily, multiple candidate live broadcast rooms can be sorted from large to small according to the recommendation probability to obtain a sorting sequence. Therefore, in the present disclosure, a set proportion of candidate live broadcast rooms ranked at the front in the sorting sequence can be used as target live broadcast rooms.

[0189] Step S504, determining the display score of each target live broadcast room based on the interactive behavior information of the promoted items in each target live broadcast room within multiple first time periods.

[0190] Among them, the display score is used to indicate the display priority of the target live broadcast room.

[0191] Step S505: sort the target live broadcast rooms based on the display scores of the target live broadcast rooms to obtain a live broadcast room list.

[0192] Step S506, recommending a list of live broadcast rooms to the target object.

[0193] The live broadcast room recommendation method of the disclosed embodiment can ensure that the recommended content is highly consistent with the actual preferences of the target object, reduce the interference of irrelevant information, and enhance the target object's trust and stickiness to the e-commerce platform.

[0194] In any embodiment of the present disclosure, the present disclosure proposes a sliding window method and combines the standardized transformation z-score technology to overcome the problems existing in the related art, wherein the z-score technology makes the data conform to the standard normal distribution by subtracting the mean of the data and dividing by the standard deviation. The labeled variable is x, and the variable transformed by the z-score technology is x'. The mean of the variable x in the sliding window is μ, and the standard deviation in the sliding window is δ, then:

[0195] As an example, taking the promoted item as a commodity, the GMV of the commodity can be expressed as follows: Figure 6 The implementation process shown is used to predict that, Figure 6 It mainly includes the following three modules: sample processing module, sample mapping module, and prediction module based on posterior data transformation.

[0196] 1. The function of the sample processing module is to complete the screening of samples, add click tags to samples with click behavior, and add conversion tags and the actual conversion price (GMV) of the product to samples with conversion behavior.

[0197] Among them, the samples include interactive behavior information of products in the live broadcast room.

[0198] The sample processing module can mark different labels by judging whether the sample has been subjected to corresponding behaviors, so as to facilitate different processing later. For example, in the sample screening stage, the input samples are screened out to select samples with click behaviors and samples with conversion behaviors, and then in the labeling stage, samples with click behaviors are labeled with clicks, and samples with conversion behaviors are labeled with conversions and the actual conversion amount of the product.

[0199] 2. The function of the sample mapping module is to map the input labeled sample feature values ​​from high-dimensional sparse to low-dimensional dense feature vectors through the embedding network, and use the neural gate to weight the feature vectors.

[0200] Considering that the present disclosure aims to improve the accuracy of GMV prediction, the role of the sample mapping module is to map high-dimensional sparse feature values ​​into low-dimensional dense feature vectors, thereby reducing the amount of calculation and improving the training efficiency of the model. Then, the low-dimensional dense feature vectors are weighted to extract useful features and enhance the prediction ability of the model.

[0201] 3. The function of the prediction module based on posterior data transformation is to calculate the mean and standard deviation of the GMV of the products in each live broadcast room in the sliding window (including multiple first time periods), and then use the GMV tower to predict the GMV of the products in each live broadcast room in the second time period ( Figure 6 The predicted GMV is recorded as the predicted GMV in the figure, and the predicted GMV is transformed through the z-score technology. After such distribution adjustment, the distribution of the predicted GMV is closer to the normal distribution, thereby improving the accuracy of data fitting.

[0202] For example, taking the number of the first period as 7, the sliding window can be selected as the 7 days in which the products in the live broadcast room have conversion behavior within the past 30 days. Among them, the prediction target of the GMV tower is: (the predicted conversion amount of the product in the second period - the mean of the GMV of the past seven days) / the standard deviation of the GMV of the past seven days.

[0203] Therefore, by using z-score standardization technology to adjust the distribution of historical data, the distribution of commodity conversion prices is closer to normal distribution, which greatly reduces the difficulty of fitting, improves the accuracy of data fitting, and reduces the calculation deviation caused by different anchors, commodities, etc. In other words, the introduction of z-score standardization transformation based on posterior data can effectively balance the stability of historical data and the sensitivity of real-time data, provide more accurate and stable GMV predictions, and significantly improve the real-time and accuracy of predictions.

[0204] It should be noted that with the booming global e-commerce, the advertising recommendation system has become an important tool for merchants to increase sales and for e-commerce platforms to optimize user experience. The advertising recommendation system determines the display position of products by sorting them, and the fierce competition for advertising positions directly affects the efficiency of merchants in acquiring traffic. With the continuous evolution of the advertising bidding mechanism, the ROI-based bidding model has gradually become the mainstream.

[0205] In this mode, the ranking index of product competition mainly depends on the estimated eCPM, and its calculation formula is: Among them, eCPM represents the amount that advertisers are willing to pay for every thousand impressions, and is an important basis for measuring advertising revenue and display effects. Advertising systems usually sort ads by calculating the eCPM value of each ad. The higher the eCPM value, the higher the display priority of the ad.

[0206] Therefore, accurately predicting the conversion amount GMV of a product not only determines the advertiser’s return on investment, but also directly affects the ranking effect of the ad space.

[0207] In the present disclosure, the eCPM of a product can be calculated based on the GMV of the product calculated based on the posterior data transformation prediction module, and the display score of the live broadcast room where the product is located can be calculated based on the eCPM of the product, so that the live broadcast rooms can be reasonably sorted according to the display scores of each live broadcast room.

[0208] With the above Figures 1 to 5 Corresponding to the live broadcast room recommendation method provided in the embodiment, the present disclosure also provides a live broadcast room recommendation device. Figures 1 to 5 The live broadcast room recommendation method provided in the embodiment corresponds to the live broadcast room recommendation method, so the implementation method of the live broadcast room recommendation method is also applicable to the live broadcast room recommendation device provided in the embodiment of the present disclosure, and will not be described in detail in the embodiment of the present disclosure.

[0209] Figure 7 This is a structural diagram of the live broadcast room recommendation device provided in Example 6 of the present disclosure.

[0210] like Figure 7 As shown, the live broadcast room recommendation device 700 may include: an acquisition module 710, a determination module 720, a sorting module 730 and a recommendation module 740.

[0211] The acquisition module 710 is used to acquire at least one target live broadcast room to be recommended to the target object;

[0212] The determination module 720 is used to determine the display score of each target live broadcast room based on the interactive behavior information of the promoted items in each target live broadcast room in multiple first time periods; wherein the display score is used to indicate the display priority of the target live broadcast room;

[0213] A sorting module 730 is used to sort the target live broadcast rooms based on the display scores of the target live broadcast rooms to obtain a live broadcast room list;

[0214] The recommendation module 740 is used to recommend a list of live broadcast rooms to the target object.

[0215] In a possible implementation of the embodiment of the present disclosure, the determination module 720 is used to: for any target live broadcast room, predict the first indicator value of the first key indicator of any promotional item in any target live broadcast room in a second time period based on the interactive behavior information of any promotional item in any target live broadcast room in multiple first time periods; wherein the multiple first time periods are located before the second time period; determine the traffic value score of any promotional item based on the first indicator value of the first key indicator; determine the display score of any target live broadcast room based on the traffic value scores of each promotional item in any target live broadcast room.

[0216] In a possible implementation of the embodiment of the present disclosure, the determination module 720 is used to: determine the mean and standard deviation of the first key indicator of any promotional item in multiple first time periods based on the interactive behavior information of any promotional item in multiple first time periods; determine the traffic value score of any promotional item based on the first indicator value, mean and standard deviation of the first key indicator.

[0217] In a possible implementation of the embodiment of the present disclosure, the determination module 720 is used to: correct the first indicator value of the first key indicator according to the mean and standard deviation of the first key indicator to obtain the second indicator value of the first key indicator; obtain the third indicator value of the second key indicator of any promotional item in the second time period; determine the traffic value score of any promotional item according to the second indicator value of the first key indicator and the third indicator value of the second key indicator.

[0218] In a possible implementation of the embodiment of the present disclosure, the first key indicator includes the total transaction amount of the item, and the second key indicator includes the click-through rate CTR, the conversion rate CVR and the return on investment ROI. The determination module 720 is used to: determine the display conversion rate of any promoted item based on the product of the third indicator value of CTR and the third indicator value of CVR; determine the total transaction amount corresponding to the unit return on investment of any promoted item based on the ratio of the second indicator value of the total transaction amount of the item and the third indicator value of ROI; determine the traffic value score of any promoted item based on the product of the display conversion rate and the total transaction amount corresponding to the unit return on investment.

[0219] In a possible implementation of the embodiments of the present disclosure, the determination module 720 is used to: vectorize the interactive behavior information of any promotional item in multiple first time periods to obtain a first behavior vector; based on a gating mechanism, weight the first behavior vector to obtain a second behavior vector; and predict the first indicator value of the first key indicator of any promotional item in the second time period based on the second behavior vector.

[0220] In a possible implementation of the embodiment of the present disclosure, the determination module 720 is used to: input the first behavior vector into a neural gate to obtain at least one gating signal output by the neural gate; wherein the gating signal is used to indicate the degree of weighting of different elements in the first behavior vector; and perform weighted processing on different elements in the first behavior vector based on the gating signal to obtain a second behavior vector.

[0221] In a possible implementation of the embodiment of the present disclosure, there are multiple promotional items in any target live broadcast room, and the determination module 720 is used to execute any of the following items: taking the average of the traffic value scores of each promotional item in any target live broadcast room as the display score of any target live broadcast room; taking the sum of the traffic value scores of each promotional item in any target live broadcast room as the display score of any target live broadcast room; performing weighted summation of the traffic value scores of each promotional item in any target live broadcast room to obtain the display score of any target live broadcast room.

[0222] In a possible implementation of the embodiment of the present disclosure, the acquisition module 710 is used to: acquire a portrait label system of the target object, and determine a candidate live broadcast room in a live broadcast state from multiple live broadcast rooms; acquire room features of multiple candidate live broadcast rooms, and item features of promoted items in multiple candidate live broadcast rooms; determine a target live broadcast room from multiple candidate live broadcast rooms based on the portrait label system, and based on the room features of multiple candidate live broadcast rooms and item features of promoted items in multiple candidate live broadcast rooms.

[0223] In a possible implementation of the embodiment of the present disclosure, the candidate live broadcast room is a live broadcast room of an e-commerce platform, and the acquisition module 710 is used to: obtain historical behavior data of the target object in multiple business fields associated with the e-commerce platform; obtain an initial label system of the target object based on the basic information of the target object; update the initial label system based on the historical behavior data of multiple business fields to obtain a portrait label system of the target object in multiple dimensions.

[0224] In a possible implementation manner of the embodiment of the present disclosure, the multiple business domains include at least one first business domain of the e-commerce platform and at least one second business domain of the information service platform, and the information service platform is used to provide a traffic entrance for the e-commerce platform.

[0225] The live broadcast room recommendation device of the disclosed embodiment can ensure that the recommended content is highly matched with the preferences and needs of the target object by obtaining the target live broadcast room to be recommended to the target object, which can be the live broadcast room that the target object is interested in. This accurate recommendation can improve the satisfaction and viewing experience of the target object. Based on the interactive behavior information of the promoted items in each target live broadcast room in multiple first time periods, the display score of each target live broadcast room is determined, and the live broadcast room list is obtained by sorting accordingly. The live broadcast rooms that are more popular and more likely to attract the attention of the target object can be placed in the front, which not only helps the target object to quickly find the live broadcast room of interest, improves viewing efficiency, and increases the exposure rate and conversion rate of the live broadcast room, but also the e-commerce platform can optimize resource allocation and allocate more traffic and exposure opportunities to those live broadcast rooms that are more likely to produce positive results, thereby improving the overall benefits and user satisfaction of the e-commerce platform.

[0226] In order to implement the above embodiments, the present disclosure also provides an electronic device, which may include 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 live broadcast room recommendation method proposed in any of the above embodiments of the present disclosure.

[0227] In order to implement the above embodiments, the present disclosure also provides a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable a computer to execute the live broadcast room recommendation method proposed in any of the above embodiments of the present disclosure.

[0228] In order to implement the above embodiments, the present disclosure also provides a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements the live broadcast room recommendation method proposed in any of the above embodiments of the present disclosure.

[0229] 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.

[0230] Figure 8 A schematic block diagram of an example electronic device that can be used to implement embodiments of the present disclosure is shown.

[0231] Among them, the electronic device may include the server and the client in the above-mentioned embodiment. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can 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 only examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0232] like Figure 8 As shown, the electronic device 800 includes a computing unit 801, which can perform various appropriate actions and processes according to a computer program stored in a ROM (Read-Only Memory) 802 or a computer program loaded from a storage unit 807 to a RAM (Random Access Memory) 803. In the RAM 803, various programs and data required for the operation of the device 800 can also be stored. The computing unit 801, the ROM 802, and the RAM 803 are connected to each other via a bus 804. An I / O (Input / Output) interface 805 is also connected to the bus 804.

[0233] A number of components in the device 800 are connected to the I / O interface 805, including: an input unit 806, such as a keyboard, a mouse, etc.; an output unit 807, such as various types of displays, speakers, etc.; a storage unit 808, such as a disk, an optical disk, etc.; and a communication unit 809, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 809 allows the device 800 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0234] The computing unit 801 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a CPU (Central Processing Unit), a GPU (Graphic Processing Units), various dedicated AI (Artificial Intelligence) computing chips, various computing units running machine learning model algorithms, a DSP (Digital Signal Processor), and any appropriate processor, controller, microcontroller, etc. The computing unit 801 performs the various methods and processes described above, such as the live broadcast room recommendation method described above. For example, in some embodiments, the live broadcast room recommendation method described above may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 808. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 800 via the ROM 802 and / or the communication unit 809. When the computer program is loaded into the RAM 803 and executed by the computing unit 801, one or more steps of the live broadcast room recommendation method described above may be performed. Alternatively, in other embodiments, the computing unit 801 may be configured to execute the above-mentioned live broadcast room recommendation method in any other appropriate manner (for example, by means of firmware).

[0235] Various embodiments of the systems and techniques described above herein may be implemented in digital electronic circuit systems, integrated circuit systems, FPGAs (Field Programmable Gate Arrays), ASICs (Application-Specific Integrated Circuits), ASSPs (Application Specific Standard Products), SOCs (System On Chips), CPLDs (Complex Programmable Logic Devices), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include: being implemented in one or more computer programs that may be executed and / or interpreted on a programmable system including at least one programmable processor that may be a dedicated or general-purpose programmable processor that may 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.

[0236] 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 data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0237] 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 device, or any suitable combination of the foregoing. More specific examples of machine-readable storage media may include electrical connections based on one or more lines, portable computer disks, hard disks, RAM, ROM, EPROM (Electrically Programmable Read-Only-Memory) or flash memory, optical fiber, CD-ROM (Compact Disc Read-Only Memory), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0238] 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 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).

[0239] The systems and techniques described herein may be implemented in a computing system that includes backend 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 frontend 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 backend components, middleware components, or frontend 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: LAN (Local Area Network), WAN (Wide Area Network), the Internet, and blockchain networks.

[0240] 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). The server may also be a server of a distributed system, or a server combined with a blockchain.

[0241] It should be noted that artificial intelligence is a discipline that studies how computers can simulate certain human thought processes and intelligent behaviors (such as learning, reasoning, thinking, planning, etc.), and includes both hardware-level and software-level technologies. Artificial intelligence hardware technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, and big data processing; artificial intelligence software technologies mainly include computer vision technology, speech recognition technology, natural language processing technology, as well as machine learning / deep learning, big data processing technology, knowledge graph technology, and other major directions.

[0242] According to the technical solution of the disclosed embodiment, by obtaining the target live broadcast room to be recommended to the target object, the target live broadcast room can be the live broadcast room that the target object is interested in, which can ensure that the recommended content is highly matched with the preferences and needs of the target object. This precise recommendation can improve the satisfaction and viewing experience of the target object. Based on the interactive behavior information of the promoted items in each target live broadcast room in multiple first time periods, the display score of each target live broadcast room is determined, and the live broadcast room list is obtained by sorting accordingly. The more popular live broadcast rooms that are more likely to attract the attention of the target object can be placed in the front, which not only helps the target object to quickly find the live broadcast room of interest, improves viewing efficiency, and increases the exposure rate and conversion rate of the live broadcast room, but also the e-commerce platform can optimize resource allocation and allocate more traffic and exposure opportunities to those live broadcast rooms that are more likely to produce positive results, thereby improving the overall benefits and user satisfaction of the e-commerce platform.

[0243] 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.

[0244] 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. A live broadcast room recommendation method, comprising: Obtain at least one target live broadcast room to be recommended to the target object; Determine a display score of each target live broadcast room based on the interactive behavior information of the promoted items in each target live broadcast room within a plurality of first time periods; wherein the display score is used to indicate the display priority of the target live broadcast room; Based on the display scores of the target live broadcast rooms, the target live broadcast rooms are sorted to obtain a live broadcast room list; Recommend the live broadcast room list to the target object.

2. The method according to claim 1, wherein: The determining of the display score of each target live broadcast room based on the interactive behavior information of the promoted items in each target live broadcast room within a plurality of first time periods includes: For any target live broadcast room, predicting a first indicator value of a first key indicator of any promotional item in the target live broadcast room in a second time period according to the interactive behavior information of any promotional item in the first time periods; wherein the first time periods are before the second time period; Determining a traffic value score of any one of the promoted items according to a first indicator value of the first key indicator; The display score of any target live broadcast room is determined according to the traffic value score of each promotional item in any target live broadcast room.

3. The method according to claim 2, wherein: Determining the traffic value score of any one of the promoted items according to the first indicator value of the first key indicator includes: Determine, according to the interactive behavior information of any promotional item in the plurality of first time periods, a mean and a standard deviation of the first key indicator of any promotional item in the plurality of first time periods; The traffic value score of any one of the promoted items is determined according to the first indicator value, mean value and standard deviation of the first key indicator.

4. The method according to claim 3, wherein: Determining the traffic value score of any one of the promoted items according to the first indicator value, mean value and standard deviation of the first key indicator includes: Correcting a first indicator value of the first key indicator according to a mean value and a standard deviation of the first key indicator to obtain a second indicator value of the first key indicator; Obtaining a third indicator value of a second key indicator of any promotional item in the second time period; The traffic value score of any one of the promoted items is determined according to the second indicator value of the first key indicator and the third indicator value of the second key indicator.

5. The method according to claim 4, wherein: The first key indicator includes the total transaction amount of the item, and the second key indicator includes the click-through rate CTR, the conversion rate CVR and the return on investment ROI. Determining the traffic value score of any promotional item according to the second indicator value of the first key indicator and the third indicator value of the second key indicator includes: Determine the display conversion rate of any one of the promotional items according to the product of the third indicator value of CTR and the third indicator value of CVR; Determine the total transaction amount corresponding to the unit investment return of any of the promoted items according to the ratio of the second indicator value of the total transaction amount of the item and the third indicator value of ROI; The traffic value score of any one of the promoted items is determined according to the product of the display conversion rate and the total transaction amount corresponding to the unit investment return.

6. The method according to claim 2, wherein: The predicting, based on the interactive behavior information of any promotional item in any target live broadcast room in the plurality of first time periods, a first indicator value of a first key indicator of any promotional item in the second time period includes: Vectorizing the interactive behavior information of any promotional item in the plurality of first time periods to obtain a first behavior vector; Based on the gating mechanism, weighting the first behavior vector is performed to obtain a second behavior vector; Predicting a first indicator value of a first key indicator of any promotional item in the second time period according to the second behavior vector.

7. The method according to claim 6, wherein: The step of performing weighted processing on the first behavior vector based on a gating mechanism to obtain a second behavior vector includes: Inputting the first behavior vector into a neural gate to obtain at least one gating signal output by the neural gate; wherein the gating signal is used to indicate the weighting degree of different elements in the first behavior vector; The second behavior vector is obtained by performing weighted processing on different elements in the first behavior vector based on the gating signal.

8. The method according to any one of claims 2 to 7, wherein: The number of promotional items in any target live broadcast room is multiple, and the determination of the display score of any target live broadcast room according to the traffic value score of each promotional item in any target live broadcast room includes any of the following: The average of the traffic value scores of the promoted items in any target live broadcast room is used as the display score of any target live broadcast room; The sum of the traffic value scores of the promoted items in any target live broadcast room is used as the display score of any target live broadcast room; The traffic value scores of the promoted items in any target live broadcast room are weightedly summed to obtain the display score of any target live broadcast room.

9. The method according to any one of claims 1 to 7, wherein: The obtaining of at least one target live broadcast room to be recommended to the target object includes: Acquire a portrait tag system of the target object, and determine a candidate live broadcast room in a live broadcast state from multiple live broadcast rooms; Acquire live broadcast room features of multiple candidate live broadcast rooms and item features of promoted items in the multiple candidate live broadcast rooms; The target live broadcast room is determined from the multiple candidate live broadcast rooms according to the portrait label system, the live broadcast room characteristics of the multiple candidate live broadcast rooms, and the item characteristics of the promoted items in the multiple candidate live broadcast rooms.

10. The method according to claim 9, wherein: The candidate live broadcast room is a live broadcast room of an e-commerce platform, and the obtaining of the portrait label system of the target object includes: Obtaining historical behavior data of the target object in multiple business fields associated with the e-commerce platform; Acquire an initial label system of the target object according to the basic information of the target object; Based on the historical behavior data of the multiple business scenarios, the initial label system is updated to obtain a portrait label system of the target object in multiple dimensions.

11. The method according to claim 10, wherein: The multiple business domains include at least one first business domain of the e-commerce platform and at least one second business domain of an information service platform, and the information service platform is used to provide a traffic entrance for the e-commerce platform.

12. A live broadcast room recommendation device, comprising: An acquisition module, used for acquiring at least one target live broadcast room to be recommended to a target object; A determination module, configured to determine a display score of each of the target live broadcast rooms based on the interactive behavior information of the promoted items in each of the target live broadcast rooms within a plurality of first time periods; wherein the display score is used to indicate the display priority of the target live broadcast room; A sorting module, used to sort the target live broadcast rooms based on the display scores of the target live broadcast rooms to obtain a live broadcast room list; A recommendation module is used to recommend the live broadcast room list to the target object.

13. The device according to claim 12, wherein: The determining module is used to: For any target live broadcast room, predicting a first indicator value of a first key indicator of any promotional item in the target live broadcast room in a second time period according to the interactive behavior information of any promotional item in the first time periods; wherein the first time periods are before the second time period; Determining a traffic value score of any one of the promoted items according to a first indicator value of the first key indicator; The display score of any target live broadcast room is determined according to the traffic value score of each promotional item in any target live broadcast room.

14. The device according to claim 13, wherein: The determining module is used to: Determine, according to the interactive behavior information of any promotional item in the plurality of first time periods, a mean and a standard deviation of the first key indicator of any promotional item in the plurality of first time periods; The traffic value score of any one of the promoted items is determined according to the first indicator value, mean value and standard deviation of the first key indicator.

15. The device according to claim 14, wherein: The determining module is used to: Correcting a first indicator value of the first key indicator according to a mean value and a standard deviation of the first key indicator to obtain a second indicator value of the first key indicator; Obtaining a third indicator value of a second key indicator of any promotional item in the second time period; The traffic value score of any one of the promoted items is determined according to the second indicator value of the first key indicator and the third indicator value of the second key indicator.

16. The device according to claim 15, wherein: The first key indicator includes the total transaction amount of the item, and the second key indicator includes the click-through rate CTR, the conversion rate CVR and the return on investment ROI. The determining module is used to: Determine the display conversion rate of any one of the promotional items according to the product of the third indicator value of CTR and the third indicator value of CVR; Determine the total transaction amount corresponding to the unit investment return of any of the promoted items according to the ratio of the second indicator value of the total transaction amount of the item and the third indicator value of ROI; The traffic value score of any one of the promoted items is determined according to the product of the display conversion rate and the total transaction amount corresponding to the unit investment return.

17. The device according to claim 13, wherein: The determining module is used to: Vectorizing the interactive behavior information of any promotional item in the plurality of first time periods to obtain a first behavior vector; Based on the gating mechanism, weighting the first behavior vector is performed to obtain a second behavior vector; Predicting a first indicator value of a first key indicator of any promotional item in the second time period according to the second behavior vector.

18. The device according to claim 17, wherein: The determining module is used to: Inputting the first behavior vector into a neural gate to obtain at least one gating signal output by the neural gate; wherein the gating signal is used to indicate the weighting degree of different elements in the first behavior vector; The second behavior vector is obtained by performing weighted processing on different elements in the first behavior vector based on the gating signal.

19. The device according to any one of claims 13 to 18, wherein: The number of promotional items in any target live broadcast room is multiple, and the determination module is used to perform any of the following: The average of the traffic value scores of the promoted items in any target live broadcast room is used as the display score of any target live broadcast room; The sum of the traffic value scores of the promoted items in any target live broadcast room is used as the display score of any target live broadcast room; The traffic value scores of the promoted items in any target live broadcast room are weightedly summed to obtain the display score of any target live broadcast room.

20. The device according to any one of claims 12 to 18, wherein: The acquisition module is used to: Acquire a portrait tag system of the target object, and determine a candidate live broadcast room in a live broadcast state from multiple live broadcast rooms; Acquire live broadcast room features of multiple candidate live broadcast rooms and item features of promoted items in the multiple candidate live broadcast rooms; The target live broadcast room is determined from the multiple candidate live broadcast rooms according to the portrait label system, the live broadcast room characteristics of the multiple candidate live broadcast rooms, and the item characteristics of the promoted items in the multiple candidate live broadcast rooms.

21. The device according to claim 20, wherein: The candidate live broadcast room is a live broadcast room of an e-commerce platform, and the acquisition module is used to: Obtaining historical behavior data of the target object in multiple business fields associated with the e-commerce platform; Acquire an initial label system of the target object according to the basic information of the target object; Based on the historical behavior data of the multiple business scenarios, the initial label system is updated to obtain a portrait label system of the target object in multiple dimensions.

22. The device according to claim 21, wherein The multiple business domains include at least one first business domain of the e-commerce platform and at least one second business domain of an information service platform, and the information service platform is used to provide a traffic entrance for the e-commerce platform.

23. 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 so that the at least one processor can execute the live broadcast room recommendation method described in any one of claims 1-11.

24. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to enable the computer to execute the live broadcast room recommendation method according to any one of claims 1-11.

25. A computer program product, comprising a computer program, which, when executed by a processor, implements the steps of the live broadcast room recommendation method according to any one of claims 1-11.

Images