Live broadcast pushing method and device, computer equipment and storage medium

Abstract

The invention relates to a live broadcast pushing method and apparatus, a computer device and a storage medium. The method comprises the steps of obtaining historical interaction data between a historical user and live broadcasts of multiple live broadcast types; the historical interaction data comprises an index value under a preset interaction index; generating a scatter diagram corresponding to each live broadcast type based on the historical interaction data; the scatter diagram is used for representing a relationship between the preset interaction index and the interestingness index for each live broadcast type; performing unified quantization processing on the scatter diagrams corresponding to the live broadcast types to obtain recommendation degrees of the live broadcast types; and pushing the live broadcast of each live broadcast type based on the recommendation degree of each live broadcast type. By adopting the method, natural differences between different types of live broadcasting rooms can be eliminated, and the pushing quality is improved.

Description

Technical Field

[0001] The present application relates to the field of communication technology, and in particular to a live broadcast push method, apparatus, computer equipment, storage medium and computer program product. Background Art

[0002] With the development of new media technology, live push has become a major communication method in the new media business field. At present, the push strategies used for live push are mean conversion quantization or quantile conversion quantization, etc. Specifically, the behavioral characteristics of users in each live broadcast room are quantitatively compared using their mean or quantile. For example, the average viewing time of users in different types of live broadcast rooms is used to determine user preferences, so as to push live broadcasts.

[0003] However, there are natural differences in the behavioral characteristics of users in different types of live broadcast rooms (such as audio, video, and socializing rooms). For example, users naturally stay longer in video live broadcast rooms than in audio live broadcast rooms. Therefore, using the mean conversion quantization or quantile conversion quantization method for live broadcast push is likely to result in the most recommended live broadcast types accounting for only a small part of all live broadcast types, while the push rate of the remaining majority of live broadcast types is low, which is likely to lead to a long-tail effect. Summary of the invention

[0004] Based on this, it is necessary to provide a live broadcast push method, apparatus, computer equipment, computer-readable storage medium and computer program product to address the technical problem that the above-mentioned live broadcast push method of mean conversion quantization or quantile conversion quantization is prone to cause a long-tail effect.

[0005] In a first aspect, the present application provides a live broadcast push method. The method comprises:

[0006] Acquire historical interaction data between historical users and live broadcasts of multiple live broadcast types; the historical interaction data includes indicator values ​​under preset interaction indicators;

[0007] Based on the historical interaction data, a scatter plot corresponding to each live broadcast type is generated; the scatter plot is used to represent the relationship between the preset interaction index and the interest index for each live broadcast type;

[0008] Performing unified quantitative processing on the scatter plots corresponding to the various live broadcast types to obtain the recommendation degree of the various live broadcast types;

[0009] Based on the recommendation degree of each live broadcast type, the live broadcasts of each live broadcast type are pushed.

[0010] In one embodiment, generating a scatter plot corresponding to each live broadcast type based on the historical interaction data includes:

[0011] For each live broadcast type, determining different indicator values ​​of the preset interaction indicator;

[0012] Based on the principle of survival analysis, statistical processing is performed on the historical interaction data of each live broadcast type to obtain interest index values ​​corresponding to the different index values; the interest index value represents the index value of the interest index;

[0013] Taking each indicator value of the preset interaction indicator as an independent variable and the interest indicator value corresponding to each indicator value as a dependent variable, a scatter plot corresponding to each live broadcast type is generated.

[0014] In one embodiment, based on the survival analysis principle, statistical processing is performed on the historical interaction data of each live broadcast type to obtain the interest index values ​​corresponding to the different index values, including:

[0015] Based on the principle of survival analysis, statistical processing is performed on the historical interaction data of each live broadcast type to obtain the survival probability under the different indicator values;

[0016] Based on the survival probabilities under the different index values, the interest index values ​​corresponding to the different index values ​​are obtained.

[0017] In one embodiment, the scatter plots corresponding to the various live broadcast types are subjected to unified quantitative processing to obtain the recommendation degree of the various live broadcast types, including:

[0018] Performing curve fitting on the scatter plots corresponding to the various live broadcast types respectively to obtain fitting relationship expressions corresponding to the various live broadcast types;

[0019] The fitting relationship expressions corresponding to the various live broadcast types are uniformly quantified to obtain the recommendation degree of the various live broadcast types.

[0020] In one embodiment, performing unified quantization processing on the fitting relationship expressions corresponding to the various live broadcast types to obtain the recommendation degree of the various live broadcast types includes:

[0021] Determine a benchmark live broadcast type; the benchmark live broadcast type is any one of the various live broadcast types;

[0022] Under the condition that the interest index values ​​corresponding to the remaining live broadcast types are equal to the interest index value corresponding to the benchmark live broadcast type, a numerical conversion relationship between the remaining live broadcast types and the benchmark live broadcast type on the preset interaction index is obtained based on the fitting relationship formula corresponding to each live broadcast type; the remaining live broadcast types are the live broadcast types other than the benchmark live broadcast type among the various live broadcast types;

[0023] Based on the numerical conversion relationship, the recommendation degree of each live broadcast type is determined.

[0024] In one embodiment, performing curve fitting on the scatter plots corresponding to the various live broadcast types to obtain fitting equations corresponding to the various live broadcast types includes:

[0025] For each live broadcast type, determining an initial fitting function for the live broadcast type;

[0026] According to the initial fitting function, curve fitting is performed on the scatter plot corresponding to the live broadcast type to obtain the goodness of fit of the initial fitting function;

[0027] If the goodness of fit is greater than a threshold, the initial fitting function is determined as a fitting relationship corresponding to the live broadcast type.

[0028] In one embodiment, the method further comprises:

[0029] If the goodness of fit is not greater than the threshold, adjusting the initial fitting function to obtain a new fitting function, and determining the goodness of fit of the new fitting function;

[0030] If the goodness of fit of the new fitting function is still not greater than the threshold, the new fitting function is adjusted again until a target fitting function with a goodness of fit greater than the threshold is obtained, and the target fitting function is determined as the fitting relationship corresponding to the live broadcast type.

[0031] In one embodiment, before pushing the live broadcasts of the live broadcast types based on the recommendation degree of the live broadcast types, the method further includes:

[0032] Determine, according to the scatter plots corresponding to the various live broadcast types, the relationship type between the preset interaction index and the interest index of the various live broadcast types; the relationship type includes a positive correlation and a negative correlation;

[0033] The pushing of the live broadcasts of the respective live broadcast types based on the recommendation degree of the respective live broadcast types includes:

[0034] Determining a recommendation priority for each live broadcast type based on the recommendation degree of each live broadcast type and the relationship type;

[0035] According to the recommendation priority, the live broadcasts of the respective live broadcast types are pushed.

[0036] In a second aspect, the present application also provides a live broadcast push device. The device comprises:

[0037] An acquisition module, used to acquire historical interaction data between historical users and live broadcasts of multiple live broadcast types; the historical interaction data includes indicator values ​​under preset interaction indicators;

[0038] A generating module, configured to generate a scatter plot corresponding to each live broadcast type based on the historical interaction data; the scatter plot is used to characterize the relationship between the preset interaction index and the interest index for each live broadcast type;

[0039] A quantification module, used for performing unified quantification processing on the scatter plots corresponding to the various live broadcast types to obtain the recommendation degree of the various live broadcast types;

[0040] The push module is used to push the live broadcasts of each live broadcast type based on the recommendation degree of each live broadcast type.

[0041] In a third aspect, the present application further provides a computer device. The computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:

[0042] Acquire historical interaction data between historical users and live broadcasts of multiple live broadcast types; the historical interaction data includes indicator values ​​under preset interaction indicators;

[0043] Based on the historical interaction data, a scatter plot corresponding to each live broadcast type is generated; the scatter plot is used to represent the relationship between the preset interaction index and the interest index for each live broadcast type;

[0044] Performing unified quantitative processing on the scatter plots corresponding to the various live broadcast types to obtain the recommendation degree of the various live broadcast types;

[0045] Based on the recommendation degree of each live broadcast type, the live broadcasts of each live broadcast type are pushed.

[0046] In a fourth aspect, the present application further provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0047] Acquire historical interaction data between historical users and live broadcasts of multiple live broadcast types; the historical interaction data includes indicator values ​​under preset interaction indicators;

[0048] Based on the historical interaction data, a scatter plot corresponding to each live broadcast type is generated; the scatter plot is used to represent the relationship between the preset interaction index and the interest index for each live broadcast type;

[0049] Performing unified quantitative processing on the scatter plots corresponding to the various live broadcast types to obtain the recommendation degree of the various live broadcast types;

[0050] Based on the recommendation degree of each live broadcast type, the live broadcasts of each live broadcast type are pushed.

[0051] In a fifth aspect, the present application further provides a computer program product. The computer program product includes a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0052] Acquire historical interaction data between historical users and live broadcasts of multiple live broadcast types; the historical interaction data includes indicator values ​​under preset interaction indicators;

[0053] Based on the historical interaction data, a scatter plot corresponding to each live broadcast type is generated; the scatter plot is used to represent the relationship between the preset interaction index and the interest index for each live broadcast type;

[0054] Performing unified quantitative processing on the scatter plots corresponding to the various live broadcast types to obtain the recommendation degree of the various live broadcast types;

[0055] Based on the recommendation degree of each live broadcast type, the live broadcasts of each live broadcast type are pushed.

[0056] The above-mentioned live broadcast push method, device, computer equipment, storage medium and computer program product generate scatter plots corresponding to each live broadcast type through historical interaction data between historical users and live broadcasts of multiple live broadcast types, which are used to characterize the relationship between preset interaction indicators and interest indicators for each live broadcast type; further, the scatter plots corresponding to each live broadcast type are uniformly quantified to obtain the recommendation degree of each live broadcast type; based on the recommendation degree of each live broadcast type, the live broadcasts of each live broadcast type are pushed. This method constructs a scatter plot between preset interaction indicators and interest indicators of each live broadcast type based on historical interaction data, and realizes an intuitive characterization of the user's preference for live broadcasts of different live broadcast types. Then, the scatter plots of different live broadcast types are uniformly quantified under preset interaction indicators to eliminate the natural differences between different types of live broadcast rooms, so that live broadcasts of different types can be converted to each other on the same basis, and the user's type preference is obtained based on comparable unified indicators, so that multiple types of live broadcasts can be better pushed to users, avoiding the long tail effect and improving the push quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 A schematic diagram of a flow chart of a live broadcast push method in one embodiment;

[0058] Figure 2 A schematic diagram of a process of generating a scatter plot in one embodiment;

[0059] Figure 3A schematic diagram of a scatter plot representing the relationship between viewing time and the cumulative retention rate of users on the next day in one embodiment;

[0060] Figure 4 A schematic diagram of a process of uniformly quantizing a scatter plot in one embodiment;

[0061] Figure 5 A schematic diagram of a process of performing curve fitting on a scatter plot in one embodiment;

[0062] Figure 6 A flowchart of a live broadcast push method in another embodiment;

[0063] Figure 7 It is a structural block diagram of a live broadcast push device in one embodiment;

[0064] Figure 8 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0065] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0066] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.

[0067] It is understandable that due to the natural differences in the behavioral characteristics of users in different types of live broadcast rooms (such as audio, video, and dating rooms), the use of mean conversion quantization or quantile conversion quantization methods for live broadcast push will easily lead to a long-tail effect. Therefore, how to eliminate the biased impact of differences in live broadcast room types on user type preferences, so as to more accurately push preferred live broadcast content to users, has become a key issue in live broadcast push.

[0068] In order to solve this problem, the present application proposes a method for live broadcast push that can unify modeling features and eliminate the biased influence of differences in live broadcast room types on user type preferences, so that users can be more accurately pushed to preferred live broadcast content. This method first selects research features (interaction indicators) and triggering events, such as viewing time features and next-day retention events, and then gives a scatter plot between user features and triggering events under different types of live broadcasts based on survival analysis. Among them, survival analysis takes into account both the cross-sectional information of the user's behavioral characteristics in the live broadcast room and the temporal variation information of the behavioral characteristics. By selecting specific triggering events, it can be refined and unified according to business scenarios. Quantification. Then, the statistical regression method is used to fit the curve and determine the fitting relationship. Finally, the mutual conversion relationship of the features under different types of live broadcasts is determined according to the fitting relationship, so as to achieve the purpose of unified quantitative features. User type preferences are derived based on comparable unified features, and user preferred live broadcast room push can be performed later.

[0069] In one embodiment, Figure 1 As shown, a live broadcast push method is provided. This embodiment uses the method applied to a terminal as an example for illustration. It can be understood that the method can also be applied to a server, and can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. Among them, the terminal can be but is not limited to various personal computers, laptops, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The server can be implemented as an independent server or a server cluster composed of multiple servers. In this embodiment, the method includes the following steps:

[0070] Step S110, obtaining historical interaction data between historical users and live broadcasts of multiple live broadcast types; the historical interaction data includes indicator values ​​under preset interaction indicators.

[0071] Among them, the live broadcast type is used to indicate the difference in live broadcast content. The live broadcast type may include game live broadcast, entertainment live broadcast, sports live broadcast, video live broadcast, music live broadcast and goods live broadcast, etc.

[0072] Among them, historical interaction data represents the historical user interaction data in the live broadcast room. For example, historical interaction data may include the historical user's stay time or viewing time in the live broadcast room, the number of comments, the number of reposts, and the amount of resources invested (such as payment), etc.

[0073] The preset interaction indicator is a direct indicator that can represent the interest of a single user in the live broadcast. For example, the preset interaction indicator can be the user's viewing time of the live broadcast. The longer the viewing time, the more interested the user is. Conversely, the shorter the viewing time, the less interested the user is. For another example, the preset interaction indicator can be the amount of resources invested by the user in the live broadcast room. The more resources invested, the more interested the user is. Conversely, the less resources invested, the less interested the user is.

[0074] Among them, there can be multiple historical users.

[0075] Step S120, generating a scatter plot corresponding to each live broadcast type based on the historical interaction data; the scatter plot is used to characterize the relationship between the preset interaction index and the interest index for each live broadcast type.

[0076] Among them, the interest index value is an index value that is associated with the preset interaction index and can represent an indirect indicator of the interest level of the group of users in the live broadcast type. The interest index value will change over time. For example, the interest index value can be the cumulative retention rate of the user on the next day. Before the end of the previous day, the cumulative retention rate of the user on the next day can change over time. For another example, the interest index value can also be the probability of the user's cumulative investment of resources, which can also change over time.

[0077] Among them, the scatter plot can be a two-dimensional coordinate graph constructed with the indicator value of the preset interaction indicator as the independent variable and the indicator value of the interest index as the dependent variable, that is, the preset interaction indicator can correspond to the x-axis, and the interest index can correspond to the y-axis.

[0078] In a specific implementation, each data point in the scatter plot corresponds to a data pair, one of which is the index value of the preset interaction index, and the other is the index value of the interest index. Therefore, based on the historical interaction data, a scatter plot corresponding to each live broadcast type is generated, including: for each live broadcast type, multiple data pairs can be determined based on the historical interaction data corresponding to the live broadcast type, and the data in each data pair are mapped in a two-dimensional coordinate graph in a manner that the preset interaction index corresponds to the x-axis and the interest index corresponds to the y-axis, so as to obtain a scatter plot corresponding to the live broadcast type.

[0079] For example, taking the preset interaction indicator as viewing time and the interest indicator as the user's next-day cumulative retention rate as an example, for the live broadcast type of music live broadcast, multiple viewing times are determined from the historical interaction data of music live broadcast, and the user's next-day cumulative retention rate corresponding to each viewing time is determined, thereby obtaining multiple data pairs. Then, in the manner of x, y = (viewing time, user's next-day cumulative retention rate), the determined data pairs are plotted in a two-dimensional coordinate graph to obtain a scatter plot corresponding to the music live broadcast type.

[0080] Step S130, performing unified quantitative processing on the scatter plots corresponding to each live broadcast type to obtain the recommendation degree of each live broadcast type.

[0081] Specifically, taking into account the biased impact of type differences in live broadcast rooms on user type preferences, the embodiment of the present application adopts a method of uniformly quantifying the scatter plots of each live broadcast type after determining the scatter plots between the preset interaction indicators and the interest indicators corresponding to each live broadcast type, so as to determine the recommendation degree of each live broadcast type, so that the benchmarks of the obtained recommendation degrees are unified, thereby eliminating the biased impact of type differences on user type preferences.

[0082] More specifically, by performing unified quantitative processing on the scatter plots corresponding to each live broadcast type, the interest index values ​​of each scatter plot can be made equal, and the relationship between each live broadcast type on the preset interaction index can be obtained, so that the live broadcast push of each live broadcast type can be performed based on the relationship between each live broadcast type on the preset interaction index.

[0083] For example, taking the preset interaction indicator as viewing time and the interest index value as the user's cumulative retention rate on the next day as an example, unified quantitative processing of the scatter plots corresponding to video live broadcast and music live broadcast can make the user's cumulative retention rate on the next day of the two scatter plots equal, and obtain the relationship between the two live broadcast types in terms of viewing time. For example, the viewing time of video live broadcast = 0.8 × the viewing time of music live broadcast. If the viewing time of music live broadcast is set to 1, the recommendation degrees of music live broadcast and video live broadcast can be 1 and 0.8 respectively.

[0084] Step S140: Push the live broadcasts of each live broadcast type based on the recommendation degree of each live broadcast type.

[0085] In a specific implementation, after determining the recommendation degree of each live broadcast type, the recommendation priority of each live broadcast type can be determined according to the recommendation degree, and the live broadcasts of each live broadcast type can be pushed according to the recommendation priority.

[0086] In the above live broadcast push method, the historical interaction data between historical users and live broadcasts of multiple live broadcast types are used to generate scatter plots corresponding to each live broadcast type, which are used to characterize the relationship between the preset interaction index and the interest index for each live broadcast type; the scatter plots corresponding to each live broadcast type are further subjected to unified quantification processing to obtain the recommendation degree of each live broadcast type; and the live broadcasts of each live broadcast type are pushed based on the recommendation degree of each live broadcast type. This method constructs a scatter plot between the preset interaction index and the interest index of each live broadcast type based on historical interaction data, so as to achieve an intuitive characterization of the user's preference for live broadcasts of different live broadcast types, and then uniformly quantifies the scatter plots of different live broadcast types under the preset interaction index to eliminate the natural differences between different types of live broadcast rooms, so that live broadcasts of different types of live broadcasts can be converted to each other on the same basis, and the user's type preference is obtained based on comparable unified indicators, so that multiple types of live broadcasts can be better pushed to users to avoid the long-tail effect.

[0087] In an exemplary embodiment, if Figure 2 As shown, in the above step S120, based on the historical interaction data, a scatter plot corresponding to each live broadcast type is generated, including:

[0088] Step S121, determining different indicator values ​​of the preset interaction indicator for each live broadcast type;

[0089] Step S122, based on the survival analysis principle, statistically process the historical interaction data of each live broadcast type to obtain interest index values ​​corresponding to different index values;

[0090] Step S123, taking each indicator value of the preset interaction indicator as an independent variable and the interest indicator value corresponding to each indicator value as a dependent variable, a scatter plot corresponding to each live broadcast type is generated.

[0091] In a specific implementation, the size of the indicator value of the preset interaction indicator can affect the recommendation result of each live broadcast type, and multiple indicator values ​​of the preset interaction indicator are associated with time. Therefore, different indicator values ​​of the preset interaction indicator can be determined based on time. For example, taking the preset interaction indicator as viewing time as an example, multiple indicator values ​​of viewing time can be 1 minute, 2 minutes, 3 minutes, etc. For another example, taking the preset interaction indicator as the amount of invested resources as an example, multiple values ​​of the amount of invested resources can be the amount of invested resources. t＝1分钟 , Amount of resources invested t＝2分钟 , Amount of resources invested t＝3分钟…etc. After determining the different indicator values ​​of the preset interaction indicators, the historical interaction data of each live broadcast type can be statistically processed based on the principle of survival analysis to obtain the survival probability of the event corresponding to the interest index value under different indicator values. Based on the survival probability under different indicator values, the interest index values ​​under different indicator values ​​are obtained. Finally, each indicator value of the preset interaction indicator is used as an independent variable, and the interest index value corresponding to each indicator value is used as a dependent variable. The indicator values ​​of the preset interaction indicator and the corresponding interest index values ​​are plotted in a two-dimensional coordinate graph to obtain a scatter plot corresponding to each live broadcast type.

[0092] For example, refer to Figure 3 , taking the preset interaction indicator as viewing time and the interest index value as the user's cumulative retention rate on the next day as an example, a scatter plot of video live broadcast type and audio live broadcast type is constructed. The horizontal axis represents the viewing time, and the vertical axis represents the user's cumulative retention rate on the next day. It can be seen from the figure that the user's cumulative retention rate on the next day is positively correlated with the viewing time.

[0093] Further, in an exemplary embodiment, step S122, based on the principle of survival analysis, statistically processing the historical interaction data of each live broadcast type to obtain interest index values ​​corresponding to different index values ​​further includes: based on the principle of survival analysis, statistically processing the historical interaction data of each live broadcast type to obtain the survival probability of events corresponding to the interest index values ​​under different index values; based on the survival probabilities under different indicator values, obtaining the interest index values ​​corresponding to different indicator values.

[0094] To facilitate understanding of this embodiment, the principle of survival analysis is first described below.

[0095] Assume that n i Represents the end time t i The number of groups that have the opportunity to trigger an event, d i represents the number of people who are actually triggered by the event at time t. According to the survival analysis principle (Kaplan-Meier Estimate, KM estimation), at t i The survival probability at time (the event has not yet occurred) is:

[0096]

[0097] Based on the above formula, it can be determined that the survival probability is a decreasing function that changes with t. As time goes on, the probability that the event has not occurred decreases.

[0098] In the application scenario of the present application, for any live broadcast type, the historical interaction data of the live broadcast type can be first counted to determine the index value of the preset interaction index at different times, and the total number of people watching the live broadcast at each time (corresponding to n i) and the number of people who actually triggered the event corresponding to the interest index value (corresponding to d i ). Substitute the total number of people watching the live broadcast and the number of people who actually triggered the event at each moment into the survival analysis model, that is, the above formula (1), to obtain the survival probability at each moment. Based on the survival probability at each moment, the interest index value at each moment is obtained. Specifically, the interest index value at each moment can be obtained by subtracting the survival probability from 1. Therefore, at each moment, the index value of the preset interaction index and the interest index value can constitute a data pair, and a scatter plot can be generated based on the data pairs at each moment.

[0099] For example, if the preset interaction indicator is the viewing time, the interest index value is the user's cumulative retention rate on the next day, and the live broadcast type is music live broadcast, the event corresponding to the interest index value can be recorded as the user's live broadcast retention on the next day. Table 1 below shows the data pairs at various moments based on this assumption. As shown in Table 1, the first column in the table represents the time, the second column represents the various indicator values ​​of the preset interaction indicator (i.e., the viewing time of the previous day); the third column represents the total number of people watching the live broadcast as of each moment; the fourth column represents the number of people who actually remain as of each moment, i.e., the number of people retained on the next day; the fifth column represents the survival probability at different moments obtained by substituting the second column data and the third column data into the above formula (1) row by row; the sixth column represents the interest index value obtained based on the survival probability, i.e., the user's cumulative retention rate on the next day.

[0100] Table 1 Relationship between viewing time and user's next-day cumulative retention rate

[0101]

[0102] Based on Table 1, we can get multiple data pairs between viewing time and user next-day cumulative retention rate, namely (1, 0.2), (2, 0.44), (3, 0.664), .... By plotting each data pair in a two-dimensional coordinate graph, we can get a scatter plot of music live broadcast types.

[0103] In this embodiment, after determining the trigger events corresponding to the interaction index and the interest index values, based on survival analysis, a scatter plot between the interaction index and the interest index corresponding to each live broadcast type is determined. Survival analysis takes into account both the cross-sectional information of the user's interaction characteristics (or behavioral characteristics) in the live broadcast room and the temporal variation information of the interaction characteristics. The scatter plot thus determined can more accurately characterize the preferences of group users for different types of live broadcasts. In addition, the interest index can be set according to demand, that is, the trigger event corresponding to the interest index can also be flexibly selected, so that refined unified quantification can be achieved according to specific business goals, overcoming the defect that traditional methods cannot perform differentiated conversion quantification according to specific business goals.

[0104] In an exemplary embodiment, if Figure 4 As shown, in the above step S130, the scatter plots corresponding to each live broadcast type are uniformly quantified to obtain the recommendation degree of each live broadcast type, including:

[0105] Step S131, performing curve fitting on the scatter plots corresponding to each live broadcast type, to obtain a fitting relationship corresponding to each live broadcast type;

[0106] Step S132, performing unified quantitative processing on the fitting relationship corresponding to each live broadcast type to obtain the recommendation degree of each live broadcast type.

[0107] In a specific implementation, before fitting the scatter plot, a fitting function may be determined first, and the scatter plot may be fitted by the fitting function. Specifically, a suitable fitting function may be selected according to the characteristics of the data in the scatter plot. After determining the fitting function, the parameters to be fitted may be determined. For example, the values ​​of the fitting parameters may be estimated by using statistical methods such as the least squares method. Afterwards, the fitting function and the determined parameters are used to perform curve fitting on the data in the scatter plot.

[0108] It can be understood that the distribution of data points on the scatter plot corresponding to each live broadcast type is different. Therefore, the fitting functions determined for different live broadcast types will also be different. Therefore, the corresponding fitting function can be determined in advance for each live broadcast type, and the curve fitting of the scatter plot can be performed to obtain the fitting relationship corresponding to each live broadcast type. Afterwards, the fitting relationship corresponding to each live broadcast type is uniformly quantified to obtain the recommendation degree of each live broadcast type. Specifically, the preset interaction index of any live broadcast type can be used as a benchmark, and the preset interaction indexes of the remaining live broadcast types can be converted according to the fitting relationship of each live broadcast type to obtain the numerical conversion relationship between the remaining live broadcast types and any of the live broadcast types under the preset interaction index. Based on this numerical conversion relationship, the recommendation degree of each live broadcast type is determined.

[0109] In this embodiment, after curve fitting is performed on the scatter plots corresponding to each live broadcast type to obtain the fitting relationship formula corresponding to each live broadcast type, the natural differences between different types of live broadcast rooms can be eliminated by performing unified quantitative processing on the fitting relationship formula corresponding to each live broadcast type, so that live broadcasts of different live broadcast types can be converted to each other on the same basis, avoiding the long-tail effect and improving the accuracy of the recommendation degree of each determined live broadcast type.

[0110] In an exemplary embodiment, in the above step S132, the fitting relationship corresponding to each live broadcast type is uniformly quantified to obtain the recommendation degree of each live broadcast type, further comprising:

[0111] Step S132a, determining the benchmark live broadcast type;

[0112] Step S132b, on the condition that the interest index values ​​corresponding to the remaining live broadcast types are equal to the interest index value corresponding to the benchmark live broadcast type, based on the fitting relationship corresponding to each live broadcast type, obtain the numerical conversion relationship between the remaining live broadcast types and the benchmark live broadcast type on the preset interaction index;

[0113] Step S132c, based on the numerical conversion relationship, determine the recommendation degree of each live broadcast type.

[0114] Among them, the benchmark live broadcast type is any live broadcast type among all live broadcast types.

[0115] Among them, the remaining live broadcast types are the live broadcast types other than the base live broadcast type.

[0116] In a specific implementation, the method of uniformly quantifying the fitting relationship corresponding to each live broadcast type is as follows: arbitrarily select a live broadcast type from each live broadcast type as the benchmark live broadcast type, then make the interest index values ​​of the remaining live broadcast types equal to the interest index value of the benchmark live broadcast type, and establish a numerical relationship between the fitting relationship corresponding to the remaining live broadcast types and the fitting relationship of the benchmark live broadcast type. Further based on the numerical relationship, the numerical conversion relationship between the remaining live broadcast types and the benchmark live broadcast type on the preset interaction index is obtained, and the recommendation degree of each live broadcast type is determined according to the numerical conversion relationship.

[0117] For example, suppose the fitting relationship of the video live broadcast type is: 1 =0.0046x 1 +0.1629;

[0118] The fitting relationship of the audio live broadcast type is: 2 =0.0045x 2 +0.0038.

[0119] Among them, y 1 and 2 Both represent the interest index value, x 1 and x 2 Both represent the indicator values ​​of the preset interaction indicators.

[0120] Taking the video live broadcast type as the benchmark live broadcast type, let y 1 =y 2 , that is: 0.0046x 1 +0.1629=0.0045x 2 +0.0038, after conversion we get:

[0121] x 2 =(0.0046x1+0.1629-0.0038) / 0.0045=1.02x 1 +35.35

[0122] This formula indicates that under the same interest index value, the numerical conversion relationship between the audio live broadcast type and the video live broadcast type on the preset interaction index is: 2 =1.02x 1 +35.35.

[0123] If x 1 =1, then x 2 =36.37. Furthermore, the recommendation degrees of the audio live broadcast type and the video live broadcast type may be 1 and 36.37 respectively.

[0124] Similarly, for other live broadcast types, the numerical conversion relationship between other live broadcast types and video live broadcast types on preset interaction indicators can be determined in the same manner as above, thereby determining the recommendation degree of each live broadcast type.

[0125] In this embodiment, a benchmark live broadcast type is first determined, and then the interest index values ​​corresponding to the remaining live broadcast types are made equal to the interest index value corresponding to the benchmark live broadcast type. In combination with the fitting relationship corresponding to each live broadcast type, the numerical conversion relationship between the remaining live broadcast types and the benchmark live broadcast type on the preset interaction index is obtained, thereby determining the recommendation degree of each live broadcast type, and realizing unified quantification of each live broadcast type on the same interaction index, thereby improving the accuracy of the recommendation degree of each live broadcast type determined based on the quantification result, and further improving the quality of live broadcast push based on the recommendation degree.

[0126] In an exemplary embodiment, if Figure 5 As shown, in the above step S131, curve fitting is performed on the scatter plots corresponding to each live broadcast type to obtain the fitting relationship corresponding to each live broadcast type, including:

[0127] Step S131a, for each live broadcast type, determining an initial fitting function for the live broadcast type;

[0128] Step S131b, performing curve fitting on the scatter plot corresponding to the live broadcast type according to the initial fitting function, and obtaining the goodness of fit of the initial fitting function;

[0129] Step S131c: if the goodness of fit is greater than the threshold, the initial fitting function is determined as the fitting relationship corresponding to the live broadcast type.

[0130] Step S131d, if the goodness of fit is not greater than the threshold, adjust the initial fitting function to obtain a new fitting function, and determine the goodness of fit of the new fitting function; if the goodness of fit of the new fitting function is still not greater than the threshold, adjust the new fitting function again until a target fitting function with a goodness of fit greater than the threshold is obtained, and determine the target fitting function as the fitting relationship corresponding to the live broadcast type.

[0131] In the specific implementation, in order to ensure the accuracy of the fitting relationship of each determined live broadcast type, this embodiment also provides a method for evaluating the accuracy and adaptability of each fitting function through goodness of fit. Specifically, for each live broadcast type, after selecting a fitting function based on the data point distribution characteristics of the scatter plot of the live broadcast type, the parameters of the fitting function are determined, and the fitting function after the parameters are determined is used as the initial fitting function. Through the initial fitting function, the data points in the scatter plot corresponding to the live broadcast type are curve fitted, and the goodness of fit corresponding to the initial fitting function is calculated.

[0132] The goodness of fit is calculated as follows:

[0133] Assume that y is the interest index value up to time t, that is, the variable to be fitted, and its mean is The fitting value calculated by the initial fitting function is The goodness of fit of the initial fitting function is R 2 It can be expressed as:

[0134]

[0135] Among them, SST (Total sum of Squares) is the total sum of squares, and SSR (Residual sum of squares) is the residual sum of squares.

[0136] In order to determine whether the initial fitting function can be used as a fitting relationship corresponding to the live broadcast type, a threshold value, such as 0.9, can be set. By comparing the size relationship between the goodness of fit and the threshold value, it can be determined whether the initial fitting function can be used as a fitting relationship corresponding to the live broadcast type.

[0137] More specifically, the interval of goodness of fit is [0,1], and the goodness of fit R 2 The closer it is to 1, the closer the regression fitting curve is to the real variable. Therefore, when the goodness of fit of the initial fitting function is greater than the threshold, it indicates that the fitting effect is good, and the initial fitting function can be determined as the final fitting relationship corresponding to the live broadcast type. On the contrary, if the goodness of fit is less than or equal to the threshold, it indicates that the fitting effect is not ideal, and the initial fitting function needs to be adjusted. Specifically, the parameters of the initial fitting function can be adjusted, or the initial fitting function can be directly replaced to obtain a new fitting function, and the goodness of fit of the new fitting function is calculated and compared with the threshold again. If the goodness of fit of the new fitting function is still not greater than the threshold, the new fitting function is adjusted again until a target fitting function with a goodness of fit greater than the threshold is obtained, and the target fitting function is determined as the fitting relationship corresponding to the live broadcast type.

[0138] In this embodiment, the fitting function is evaluated by the goodness of fit. If the goodness of fit of the fitting function does not meet the condition, the fitting function is adjusted until a target fitting function whose goodness of fit meets the condition is obtained, thereby ensuring that the fitting relationship corresponding to each determined live broadcast type is the optimal fitting relationship and the best fitting effect is ensured, thereby improving the credibility and accuracy of the results of unified quantification based on the fitting relationship and improving the quality of live broadcast push.

[0139] In an exemplary embodiment, in the above step S140, based on the recommendation degree of each live broadcast type, before pushing the live broadcast of each live broadcast type, it also includes: determining the relationship type between the preset interaction index and the interest index of each live broadcast type according to the scatter plot corresponding to each live broadcast type; the relationship type includes a positive correlation and a negative correlation; correspondingly, in step S140, based on the recommendation degree of each live broadcast type, pushing the live broadcast of each live broadcast type further includes: determining the recommendation priority for each live broadcast type based on the recommendation degree of each live broadcast type and the relationship type; and pushing the live broadcast of each live broadcast type according to the recommendation priority.

[0140] Among them, the recommendation priority can be used to determine the order of recommendation. The higher the recommendation priority, the earlier it is recommended, and the lower the recommendation priority, the later it is recommended.

[0141] Specifically, since the recommendation degree of each live broadcast type is determined based on the preset interaction index, and the recommendation priority of each live broadcast type is affected by the interest index value, before determining the recommendation priority of each live broadcast type based on the recommendation degree of each live broadcast type, it is also necessary to first determine the relationship between the preset interaction index and the interest index, that is, determine whether the two are positively correlated or negatively correlated. Then, combined with the recommendation degree of each live broadcast type and the type of relationship between the two, the recommendation priority for each live broadcast type is determined; according to the recommendation priority, the live broadcasts of each live broadcast type are pushed.

[0142] More specifically, when there is a positive correlation between the preset interaction index and the interest index, that is, the larger the index value of the preset interaction index, the larger the interest index value, and correspondingly, the higher the recommendation priority. The recommendation degree of each live broadcast type is uniformly quantified in the preset interaction index dimension, so the larger the recommendation degree, the larger the index value of the preset interaction index. From this, it can be determined that the recommendation degree is positively correlated with the recommendation priority, that is, the larger the recommendation degree, the higher the recommendation priority. For example, the recommendation degree of music live broadcast is 1.2, the recommendation degree of video live broadcast is 1, and the recommendation degree of game live broadcast is 0.8. The recommendation priority of the three live broadcast types is from high to low: music live broadcast>video live broadcast>game live broadcast.

[0143] On the contrary, when there is a negative correlation between the preset interaction index and the interest index, that is, the larger the index value of the preset interaction index, the smaller the interest index value, and correspondingly, the lower the recommendation priority. Similarly, as above, the greater the recommendation degree, the greater the index value of the preset interaction index. It can be determined that the recommendation degree and the recommendation priority are negatively correlated, that is, the greater the recommendation degree, the lower the recommendation priority. For example, the recommendation degree of music live broadcast is 1.2, the recommendation degree of video live broadcast is 1, and the recommendation degree of game live broadcast is 0.8. The recommendation priority of the three live broadcast types from high to low is: game live broadcast>video live broadcast>music live broadcast.

[0144] It should be noted that the relationship type between the preset interaction indicators and the interest indicators of each live broadcast type can be directly determined based on the distribution pattern of data points in the scatter plot of each live broadcast type, or it can be determined based on the fitting relationship after curve fitting is performed on the scatter plot corresponding to any live broadcast type and the corresponding fitting relationship is obtained. Since the relationship type between the preset interaction indicators corresponding to each live broadcast type and the interest indicators of each live broadcast type is consistent, the relationship type can be determined based on the scatter plot or fitting relationship corresponding to any live broadcast type.

[0145] In this embodiment, based on the scatter plot corresponding to each live broadcast type, the relationship type between the preset interaction indicator and the interest indicator of each live broadcast type is determined, and further based on the recommendation degree and relationship type of each live broadcast type, the recommendation priority for each live broadcast type is determined. This can ensure the accuracy of the determined recommendation priority, thereby improving the push effect of the live broadcasts of each live broadcast type according to the recommendation priority.

[0146] In one embodiment, in order to facilitate those skilled in the art to understand the embodiment of the present application, the following takes the preset interaction indicator as the viewing time and the interest index value as the user's next-day cumulative retention rate as an example. Figure 6 The flowchart shown in the figure specifically illustrates the live broadcast push method provided by this application.

[0147] Figure 6 The flowchart shown includes the following steps:

[0148] Step S610, obtaining historical interaction data between historical users and live broadcasts of multiple live broadcast types; the historical interaction data includes indicator values ​​under viewing time;

[0149] Step S620, determining different indicator values ​​of viewing time for each live broadcast type;

[0150] Step S630, based on the principle of survival analysis, statistically process the historical interaction data of each live broadcast type to obtain the survival probability under different viewing durations;

[0151] Step S640, based on the survival probabilities at different viewing times, obtaining the next-day cumulative retention rates of users at different viewing times;

[0152] Step S650, taking each viewing time as an independent variable and the next-day cumulative retention rate of users corresponding to each viewing time as a dependent variable, generating a scatter plot corresponding to each live broadcast type;

[0153] Step S660, performing curve fitting on the scatter plots corresponding to each live broadcast type, to obtain a fitting relationship corresponding to each live broadcast type;

[0154] Step S670, performing unified quantitative processing on the fitting relationship formula corresponding to each live broadcast type to obtain the recommendation degree of each live broadcast type; and determining the relationship type between the viewing time and the next-day cumulative retention rate of users of each live broadcast type according to the scatter plot corresponding to each live broadcast type, the relationship type including positive correlation and negative correlation;

[0155] Step S680, determining the recommendation priority for each live broadcast type based on the recommendation degree and relationship type of each live broadcast type;

[0156] Step S690: Push live broadcasts of various live broadcast types according to recommendation priorities.

[0157] The live broadcast push method provided in this embodiment constructs a scatter plot between preset interaction indicators and interest indicators of each live broadcast type based on historical interaction data, thereby achieving an intuitive representation of user preferences for live broadcasts of different live broadcast types. The scatter plots of different live broadcast types are then uniformly quantified under preset interaction indicators to eliminate the natural differences between different types of live broadcast rooms, so that live broadcasts of different types can be converted to each other on the same basis. Based on comparable unified indicators, the user's type preference is obtained, thereby better pushing multiple types of live broadcasts to users and avoiding the long-tail effect.

[0158] After experimental verification, after adopting different types of live broadcast push methods based on survival analysis provided in this application, the scenarios of music live broadcast business involving this technology have achieved significant improvement under AB experimental verification. For example, in the recommended Tab and up and down sliding scenarios, the viewing time is selected as the quantitative feature and the retention is used as the trigger event. For video, audio, and dating type live broadcasts, after adopting a unified viewing time quantification scheme, the average number of views per user with high activity in the up and down sliding scenarios increased by 5.74%, and the average viewing time increased by 0.83%. The average viewing time of new users on the same day increased by 3.217%, the average viewing time of returning users on the same day increased by 7.414%, and the average viewing time of low activity users increased by 0.6750%; in the recommended Tab scenario, the average number of views per user with high activity increased by 3.89%, and the average viewing time increased by 4.851%. The uv penetration rate of returning users on the same day increased by 9.646%, and the average viewing time increased by 11.62%. The average viewing time of new users increased by 16.01%, and the uv penetration rate of low activity users increased by 2.02%.

[0159] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0160] Based on the same inventive concept, the embodiment of the present application also provides a live push device for implementing the live push method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more live push device embodiments provided below can refer to the limitations of the live push method above, and will not be repeated here.

[0161] In one embodiment, Figure 7 As shown, a live broadcast push device is provided, including: an acquisition module 710, a generation module 720, a quantification module 730 and a push module 740, wherein:

[0162] An acquisition module 710 is used to acquire historical interaction data between historical users and live broadcasts of multiple live broadcast types; the historical interaction data includes an indicator value under a preset interaction indicator;

[0163] A generating module 720 is used to generate a scatter plot corresponding to each live broadcast type based on the historical interaction data; the scatter plot is used to represent the relationship between the preset interaction index and the interest index for each live broadcast type;

[0164] A quantification module 730 is used to perform unified quantification processing on the scatter plots corresponding to the various live broadcast types to obtain the recommendation degree of the various live broadcast types;

[0165] The push module 740 is used to push the live broadcasts of the respective live broadcast types based on the recommendation degree of the respective live broadcast types.

[0166] In one embodiment, the generation module 720 is also used to determine different indicator values ​​of the preset interaction indicator for each live broadcast type; based on the principle of survival analysis, the historical interaction data of each live broadcast type is statistically processed to obtain the interest indicator values ​​corresponding to the different indicator values; the interest indicator value represents the indicator value of the interest indicator; each indicator value of the preset interaction indicator is used as an independent variable, and the interest indicator value corresponding to each indicator value is used as a dependent variable to generate a scatter plot corresponding to each live broadcast type.

[0167] In one embodiment, the generation module 720 is also used to perform statistical processing on the historical interaction data of each live broadcast type based on the principle of survival analysis to obtain the survival probability under the different indicator values; based on the survival probability under the different indicator values, obtain the interest index value under the different indicator values.

[0168] In one of the embodiments, the quantification module 730 is further used to perform curve fitting on the scatter plots corresponding to the various live broadcast types to obtain the fitting relationship expressions corresponding to the various live broadcast types; and perform unified quantization processing on the fitting relationship expressions corresponding to the various live broadcast types to obtain the recommendation degree of the various live broadcast types.

[0169] In one embodiment, the quantification module 730 is also used to determine a benchmark live broadcast type; the benchmark live broadcast type is any live broadcast type among the various live broadcast types; on the condition that the interest index values ​​corresponding to the remaining live broadcast types are equal to the interest index values ​​corresponding to the benchmark live broadcast type, based on the fitting relationship corresponding to the various live broadcast types, the numerical conversion relationship between the remaining live broadcast types and the benchmark live broadcast type on the preset interaction index is obtained; the remaining live broadcast types are the live broadcast types among the various live broadcast types except the benchmark live broadcast type; based on the numerical conversion relationship, the recommendation degree of each live broadcast type is determined.

[0170] In one of the embodiments, the quantification module 730 is also used to determine, for each live broadcast type, an initial fitting function for the live broadcast type; based on the initial fitting function, curve fitting is performed on the scatter plot corresponding to the live broadcast type to obtain the goodness of fit of the initial fitting function; if the goodness of fit is greater than a threshold, the initial fitting function is determined as the fitting relationship corresponding to the live broadcast type.

[0171] In one embodiment, the quantization module 730 is also used to adjust the initial fitting function to obtain a new fitting function and determine the goodness of fit of the new fitting function if the goodness of fit is not greater than the threshold; if the goodness of fit of the new fitting function is still not greater than the threshold, adjust the new fitting function again until a target fitting function with a goodness of fit greater than the threshold is obtained, and determine the target fitting function as the fitting relationship corresponding to the live broadcast type.

[0172] In one of the embodiments, the push module 740 is also used to determine the relationship type between the preset interaction indicator and the interest indicator of each live broadcast type according to the scatter plot corresponding to each live broadcast type; the relationship type includes a positive correlation and a negative correlation; based on the recommendation degree of each live broadcast type and the relationship type, determine the recommendation priority for each live broadcast type; and push the live broadcast of each live broadcast type according to the recommendation priority.

[0173] Each module in the above-mentioned live push device can be implemented in whole or in part by software, hardware and their combination. Each of the above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory in the computer device in the form of software, so that the processor can call and execute the corresponding operations of each of the above modules.

[0174] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 8As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a live push method is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a button, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0175] Those skilled in the art will understand that Figure 8 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0176] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiments when executing the computer program.

[0177] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0178] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0179] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0180] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0181] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be construed as limiting the scope of the present application. It should be noted that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A live broadcast push method, It is characterized in that The method comprises: Acquire historical interaction data between historical users and live broadcasts of multiple live broadcast types; the historical interaction data includes indicator values ​​under preset interaction indicators; Based on the historical interaction data, a scatter plot corresponding to each live broadcast type is generated; the scatter plot is used to represent the relationship between the preset interaction index and the interest index for each live broadcast type; Performing unified quantitative processing on the scatter plots corresponding to the various live broadcast types to obtain the recommendation degree of the various live broadcast types; Based on the recommendation degree of each live broadcast type, the live broadcasts of each live broadcast type are pushed.

2. The method according to claim 1, It is characterized in that The generating a scatter plot corresponding to each live broadcast type based on the historical interaction data includes: For each live broadcast type, determining different indicator values ​​of the preset interaction indicator; Based on the principle of survival analysis, statistical processing is performed on the historical interaction data of each live broadcast type to obtain interest index values ​​corresponding to the different index values; the interest index value represents the index value of the interest index; Taking each indicator value of the preset interaction indicator as an independent variable and the interest indicator value corresponding to each indicator value as a dependent variable, a scatter plot corresponding to each live broadcast type is generated.

3. The method according to claim 2, It is characterized in that Based on the survival analysis principle, statistical processing is performed on the historical interaction data of each live broadcast type to obtain the interest index values ​​corresponding to the different index values, including: Based on the principle of survival analysis, statistical processing is performed on the historical interaction data of each live broadcast type to obtain the survival probability under the different indicator values; Based on the survival probabilities under the different index values, the interest index values ​​corresponding to the different index values ​​are obtained.

4. The method according to claim 1, It is characterized in that The step of performing unified quantitative processing on the scatter plots corresponding to the various live broadcast types to obtain the recommendation degree of the various live broadcast types includes: Performing curve fitting on the scatter plots corresponding to the various live broadcast types respectively to obtain fitting relationship expressions corresponding to the various live broadcast types; The fitting relationship expressions corresponding to the various live broadcast types are uniformly quantified to obtain the recommendation degree of the various live broadcast types.

5. The method according to claim 4, It is characterized in that The uniformly quantifying the fitting relational expressions corresponding to the various live broadcast types to obtain the recommendation degree of the various live broadcast types includes: Determine a benchmark live broadcast type; the benchmark live broadcast type is any one of the various live broadcast types; Under the condition that the interest index values ​​corresponding to the remaining live broadcast types are equal to the interest index value corresponding to the benchmark live broadcast type, a numerical conversion relationship between the remaining live broadcast types and the benchmark live broadcast type on the preset interaction index is obtained based on the fitting relationship formula corresponding to each live broadcast type; the remaining live broadcast types are the live broadcast types other than the benchmark live broadcast type among the various live broadcast types; Based on the numerical conversion relationship, the recommendation degree of each live broadcast type is determined.

6. The method according to claim 4, It is characterized in that The curve fitting is performed on the scatter plots corresponding to the various live broadcast types to obtain the fitting relationship formulas corresponding to the various live broadcast types, including: For each live broadcast type, determining an initial fitting function for the live broadcast type; According to the initial fitting function, curve fitting is performed on the scatter plot corresponding to the live broadcast type to obtain the goodness of fit of the initial fitting function; If the goodness of fit is greater than a threshold, the initial fitting function is determined as a fitting relationship corresponding to the live broadcast type.

7. The method according to claim 6, It is characterized in that The method further comprises: If the goodness of fit is not greater than the threshold, adjusting the initial fitting function to obtain a new fitting function, and determining the goodness of fit of the new fitting function; If the goodness of fit of the new fitting function is still not greater than the threshold, the new fitting function is adjusted again until a target fitting function with a goodness of fit greater than the threshold is obtained, and the target fitting function is determined as the fitting relationship corresponding to the live broadcast type.

8. The method according to claim 1, It is characterized in that Before pushing the live broadcasts of the respective live broadcast types based on the recommendation degrees of the respective live broadcast types, the method further includes: Determine, according to the scatter plots corresponding to the various live broadcast types, the relationship type between the preset interaction index and the interest index of the various live broadcast types; the relationship type includes a positive correlation and a negative correlation; The pushing of the live broadcasts of the respective live broadcast types based on the recommendation degree of the respective live broadcast types includes: Determining a recommendation priority for each live broadcast type based on the recommendation degree of each live broadcast type and the relationship type; According to the recommendation priority, the live broadcasts of the respective live broadcast types are pushed.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program. It is characterized in that When the processor executes the computer program, the steps of the live broadcast push method described in any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the steps of the live broadcast push method described in any one of claims 1 to 8 are implemented.

Images