CN112200639A - Information flow model construction method, device and medium - Google Patents
Information flow model construction method, device and medium Download PDFInfo
- Publication number
- CN112200639A CN112200639A CN202011189611.7A CN202011189611A CN112200639A CN 112200639 A CN112200639 A CN 112200639A CN 202011189611 A CN202011189611 A CN 202011189611A CN 112200639 A CN112200639 A CN 112200639A
- Authority
- CN
- China
- Prior art keywords
- weight
- information flow
- stay time
- flow model
- sample
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000010276 construction Methods 0.000 title claims abstract description 24
- 238000000034 method Methods 0.000 claims abstract description 42
- 230000003993 interaction Effects 0.000 claims description 59
- 230000006870 function Effects 0.000 claims description 55
- 238000004590 computer program Methods 0.000 claims description 15
- 238000012545 processing Methods 0.000 claims description 12
- 238000004088 simulation Methods 0.000 abstract description 24
- 230000002452 interceptive effect Effects 0.000 abstract description 22
- 230000014759 maintenance of location Effects 0.000 abstract description 12
- 230000000694 effects Effects 0.000 abstract description 8
- 238000004364 calculation method Methods 0.000 description 15
- 238000012886 linear function Methods 0.000 description 10
- 238000010586 diagram Methods 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000004891 communication Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000007704 transition Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 1
- 230000019771 cognition Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000005055 memory storage Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000009877 rendering Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/067—Enterprise or organisation modelling
Abstract
The application discloses a method, a device and a medium for constructing an information flow model, wherein the method comprises the following steps: and after the stay time and the interactive data of each sample are obtained, calculating the weight of each sample according to the corresponding weight of each stay time and the interactive data, and constructing an information flow model according to the corresponding weight of each sample. Because the weights corresponding to the stay time lengths and the interactive data are respectively calculated, the confidence coefficient of the stay time lengths of the samples is improved. If the original stay time is directly added into the information flow model, other core indexes on the line can obviously drop, so that the influence of the directly added original stay time on the prediction effect of the information flow model is reduced, the reliability of the simulation result of the information flow model is improved, the matching degree of recommended pictures and/or video contents is improved, and the stay time and the retention rate of a user are further improved. In addition, the information flow model construction device and the medium correspond to the information flow model construction method, and the effect is the same as that of the information flow model construction method.
Description
Technical Field
The present application relates to the field of information technology, and in particular, to a method, an apparatus, and a medium for constructing an information flow model.
Background
With the development of science and technology, e-commerce is more and more prosperous, and in order to attract the eyes of users, the e-commerce platform often recommends pictures and videos with high click rate and low dwell time to the users by only including a single information flow model, so that the contents of the pictures and/or videos are poor, and the cognition of the users on the e-commerce platform is greatly reduced.
At present, an information flow model is mainly constructed through the stay time of samples, and under the actual condition, the stay time of each sample is various and wide in distribution range, so that the stay time is directly adopted, the information flow model comprises data with too long or too short stay time in the samples, the confidence coefficient of the data is low, the reliability of a simulation result of the information flow model is low, the matching degree of pictures and/or videos recommended according to the simulation result is poor, and the stay time and the retention rate of a user are reduced.
Therefore, how to improve the reliability of the information flow model simulation result and improve the residence time and retention rate of the user is a problem to be solved urgently by the technical personnel in the field.
Disclosure of Invention
The application aims to provide an information flow model construction method, an information flow model construction device and an information flow model construction medium.
In order to solve the above technical problem, the present application provides an information flow model construction method, including:
obtaining the stay time and interaction data of each sample;
determining a dwell time weight corresponding to the dwell time and an interaction weight corresponding to the interaction data to obtain a weight of each sample;
and constructing an information flow model according to the weight corresponding to each sample.
Preferably, determining the corresponding weights of the dwell time and the interaction data to obtain the weight of each sample includes:
dividing the sample with at least one of the stay time and/or the interaction data into a positive sample, and dividing the sample without the stay time and the interaction data into a negative sample;
respectively calculating each interaction weight and each stay time length weight;
calculating the weight corresponding to each positive sample according to each stay time length weight and each interaction weight;
setting the weight corresponding to each negative sample as a first fixed value;
wherein the weight corresponding to the positive sample is greater than the first fixed value.
Preferably, calculating the dwell time weight comprises:
under the condition that the stay time length is not greater than a first threshold value, carrying out linear processing on the stay time length to obtain a first stay time length weight;
and under the condition that the stay time length is larger than the first threshold value, calculating a second stay time length weight according to a first logarithmic function.
Preferably, calculating the dwell time weight comprises:
setting a third staying time length weight as a second fixed value under the condition that the staying time length is not more than a second threshold value;
under the condition that the stay time length is larger than the second threshold value, calculating a fourth stay time length weight according to a second logarithmic function;
wherein the second threshold is less than the first threshold.
Preferably, before the determining the dwell time and the corresponding weight of the interaction data to obtain the weight of each sample, the method further includes:
setting the stay time length as a third threshold value under the condition that the stay time length is greater than the third threshold value;
wherein the third threshold is greater than the first threshold.
Preferably, the interaction data includes: like data, comment data, share data, concern data, and collection data.
Preferably, the first threshold is specifically 50 seconds, the second threshold is specifically 4 seconds, and the third threshold is specifically 300 seconds.
In order to solve the above technical problem, the present application further provides an information flow model building apparatus, including:
the acquisition module is used for acquiring the stay time and the interaction data of each sample;
the determining module is used for determining a dwell time weight corresponding to the dwell time and an interaction weight corresponding to the interaction data to obtain a weight of each sample;
and the construction module is used for constructing an information flow model according to the weight corresponding to each sample.
In order to solve the above technical problem, the present application further provides an information flow model building apparatus, including:
a memory for storing a computer program;
a processor for implementing the steps of the information flow model construction method as described above when executing the computer program.
In order to solve the above technical problem, the present application further provides a computer-readable storage medium, on which a computer program is stored, and the computer program, when executed by a processor, implements the steps of the information flow model building method as described above.
According to the information flow model construction method, after the stay time and the interaction data of each sample are obtained, the weight of each sample is calculated according to the corresponding weight of each stay time and the corresponding weight of the interaction data, and the information flow model is constructed according to the corresponding weight of each sample. Because the weights corresponding to the stay time lengths and the interactive data are respectively calculated, the confidence coefficient of the stay time lengths of the samples is improved. If the original stay time is directly added into the information flow model, other core indexes on the line can obviously drop, so that the influence of the directly added original stay time on the prediction effect of the information flow model is reduced, the reliability of the simulation result of the information flow model is improved, the matching degree of recommended pictures and/or video contents is improved, and the stay time and the retention rate of a user are further improved.
In addition, the information flow model construction device and the medium correspond to the information flow model construction method, and the effect is the same as that of the information flow model construction method.
Drawings
In order to more clearly illustrate the embodiments of the present application, the drawings needed for the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and that other drawings can be obtained by those skilled in the art without inventive effort.
Fig. 1 is a flowchart of an information flow model building method according to an embodiment of the present application;
fig. 2 is a flowchart for calculating weights of samples according to an embodiment of the present disclosure;
FIG. 3 is a flowchart of calculating dwell time duration weights according to an embodiment of the present application;
FIG. 4 is a graph of a first threshold value of a fitted dwell time weight of 30 seconds provided by an embodiment of the present application;
FIG. 5 is a graph of a first threshold value of a fitted dwell time weight of 50 seconds provided by an embodiment of the present application;
FIG. 6 is a graph comparing the weights calculated according to the default logarithmic function and the weights calculated according to the first logarithmic function provided by the embodiments of the present application;
FIG. 7 is a flow chart of another method for calculating dwell time weights according to embodiments of the present application;
FIG. 8 is a graph comparing interaction weights, dwell period weights determined using a first logarithmic function and a linear function, and dwell period weights determined using a second logarithmic function and a second fixed value, provided by an embodiment of the present application;
fig. 9 is a schematic structural diagram of an information flow model building apparatus according to an embodiment of the present application;
fig. 10 is a schematic structural diagram of another information flow model building apparatus according to an embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all the embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments in the present application without any creative effort belong to the protection scope of the present application.
The core of the application is to provide a method, a device and a medium for constructing an information flow model.
In order that those skilled in the art will better understand the disclosure, the following detailed description will be given with reference to the accompanying drawings.
Fig. 1 is a flowchart of an information flow model building method according to an embodiment of the present application. As shown in fig. 1, the method includes:
s10: and acquiring the stay time and the interaction data of each sample.
In the embodiment of the present application, the retention time of each sample is obtained by dotting statistics, and the interactive data is specifically a series of interactive indexes of the sample, and as an optimal embodiment, the interactive data includes: the interactive data comprises more data types, more perfect interactive data of the sample, and higher reliability of a simulation result of finally constructing the information flow model.
S11: and determining the weight of the stay time corresponding to the stay time and the interaction weight corresponding to the interaction data to obtain the weight of each sample.
It should be noted that, the calculation method of the dwell time length weight and the interaction weight is not specifically limited, and in specific implementation, a logarithmic function may be used for calculation, and accordingly, the weight of each sample is the sum of the dwell time length weight and the interaction weight of each sample.
S12: and constructing an information flow model according to the weight corresponding to each sample.
The information flow model is constructed based on a loss function, namely, the stay time and the interaction data of each sample are used as weights in the loss function to influence the transformation of the loss function, and further influence the simulation result of the information flow model. The formula for the loss function is as follows:
sigmoids_weighted＝-y_true*tf.log(y_pred)*reweight-(1-y_true)*tf.log(1-y_pred)
sigmoids_loss＝tf.reduce_mean(sigmoids_weighted)
wherein sigmoids _ weighted is a total loss of a hyper-parameter (i.e. total loss of 1024 samples), y _ true is a real label, tf.log is a logarithmic function, y _ pred is a pre-estimation value of an information flow model, reweight is a weight corresponding to a sample (i.e. a summation value of a stay duration weight and an interaction weight corresponding to the sample), sigmoids _ loss is an average loss of the sample, and tf.
In the method for constructing an information flow model provided in this embodiment, after the staying time and the interaction data of each sample are obtained, the weight of each sample is calculated according to the staying time and the weight corresponding to the interaction data, and an information flow model is constructed according to the weight corresponding to each sample. Because the weights corresponding to the stay time lengths and the interactive data are respectively calculated, the confidence coefficient of the stay time lengths of the samples is improved. If the original stay time is directly added into the information flow model, other core indexes on the line can obviously drop, so that the influence of the directly added original stay time on the prediction effect of the information flow model is reduced, the reliability of the simulation result of the information flow model is improved, the matching degree of recommended pictures and/or video contents is improved, and the stay time and the retention rate of a user are further improved.
Fig. 2 is a flowchart for calculating weights of samples according to an embodiment of the present disclosure. As shown in fig. 2, on the basis of the foregoing embodiment, S11 specifically includes:
s20: and dividing the sample with at least one of the stay time and/or the interaction data into positive samples, and dividing the sample without the stay time and the interaction data into negative samples.
S21: and respectively calculating each interaction weight and each stay time length weight.
It should be noted that, the calculation method of the dwell time weight and the interaction weight is not particularly limited, and in a specific implementation, a logarithmic function may be used for calculation. For example, the interaction weight is calculated as follows:
In＝log1.2(interact_num+1.0)
wherein In is the interaction weight, and interaction _ num is the simple sum or weighted sum of the five events of like, comment, share, concern and collection. For example, if only the comment data and the comment data are in sample a, the interct _ num is 2 by simple summation.
It is understood that the base number of the logarithmic function of 1.2 is only one expression, and in the specific implementation, the base number of the logarithmic function is not limited and can be matched with the requirement.
S22: and calculating the weight corresponding to each positive sample according to each stay time length weight and each interaction weight.
It should be further noted that the weight corresponding to each positive sample is a sum of the dwell time weight and the interaction weight of each positive sample.
S23: and setting the weight corresponding to each negative sample as a first fixed value.
In this embodiment of the application, the weight corresponding to the positive sample is greater than a first fixed value, and in specific implementation, the first fixed value may be selected to be 1.0.
It is understood that S22 and S23 have no sequence, and may be calculated in the sequence shown in fig. 2, or may be calculated at the same time as the weights corresponding to the positive samples and the weights corresponding to the negative samples are set.
In the information flow model construction method provided by this embodiment, since the samples are divided into the positive samples and the negative samples, and the weight of the positive samples is greater than that of the negative samples, the samples with longer dwell time and more interactive data are sufficiently trained in the information flow model training, so that the simulation result of the information flow model tends to recommend pictures and/or videos with long dwell time and more interactive data.
Fig. 3 is a flowchart for calculating a dwell time duration weight according to an embodiment of the present application. As shown in fig. 3, on the basis of the above embodiment, calculating the dwell-time weight includes:
s30: and judging whether the stay time length is larger than a first threshold, if not, entering S31, and if so, entering S32.
S31: and performing linear processing on the stay time to obtain a first stay time weight.
S32: a second dwell-time length weight is calculated according to the first logarithmic function.
In order to make the flowchart for calculating the stay time length weight provided in the embodiment of the present application more clear to those skilled in the art, the embodiment will be specifically described by way of example.
In the case where the stay period is not greater than the first threshold, the first stay period weight is calculated by the following formula:
Ti1＝0.8*time
wherein time is the dwell time of the sample having a dwell time not greater than a first threshold, Ti1Is the first dwell time length weight.
It should be noted that the ratio in the linear function is not particularly limited, and the value of the ratio needs to match the requirement and the logarithmic function, so that the ratio of the linear function of 0.8 is only one representation form, and does not represent only one representation form.
In the case where the stay period is greater than the first threshold, the second stay period weight is calculated by the following formula:
Ti2＝log1.0775(time-t1)
wherein, Ti2Is a second dwell time duration weight, t1Is a fixed value not greater than a first threshold, and time is the stay time of the sample with the stay time greater than the first threshold, wherein the time is greater than t1。
The base number in the logarithmic function is determined by the distribution index such as the mean, the variance, the quantile, and the like of the residence time of each sample, and therefore 1.0775, which is the base number of the logarithmic function, is only one expression form, and does not represent only one expression form.
Fig. 4 is a curve fitted with a dwell time weight of 30 seconds as a first threshold value provided by an embodiment of the present application, and fig. 5 is a curve fitted with a dwell time weight of 50 seconds as a first threshold value provided by an embodiment of the present application. Wherein, the abscissa is the time of the stay time, and the ordinate is the weight of the stay time. In the case where the first threshold value is 30 seconds or the first threshold value is 50 seconds, t1May have a value of 30. As can be seen from fig. 4 and 5, the linear function is tangent to the logarithmic function when the first threshold is 50 seconds, and a smooth transition is achieved.
It will be appreciated that the slope of the linear function is the same as the slope of the logarithmic function at the intersection of the linear function, the smaller the error of the dwell-time weight at the intersection of the two, and therefore, in order to achieve a smooth transition of the curve of the dwell-time weight, the first threshold value may be selected to be 50 seconds as a preferred embodiment.
In addition, the second stay time length weight can be calculated directly through a default logarithmic function, and correspondingly, the ratio of the linear function needs to be matched with the default logarithmic function, so that the default logarithmic function is intersected with the linear function. Wherein the default logarithmic function includes a base-2 logarithmic function, a base-10 logarithmic function, and a base-e logarithmic function.
Fig. 6 is a comparison graph of calculating weights according to a default logarithmic function and calculating weights according to a first logarithmic function provided in an embodiment of the present application. As shown in fig. 6, the abscissa is the number of interaction events in one positive sample, and the ordinate is the weight of the positive sample, where Ti represents the weight of the dwell time calculated by the default logarithmic function. When the weight of the positive sample is calculated according to the default logarithmic function, the weight change of the positive sample is smaller along with the increase of the number of the interaction events, so that the influence of the stay time on the simulation result of the information flow model is larger by the calculation method, and compared with the calculation method for calculating the weight according to the first logarithmic function, the reliability of the simulation result of the information flow model established according to the calculation method for calculating the weight according to the default logarithmic function is lower.
The calculation method of the first logarithmic function is adopted, and the final simulation result of the information flow model established by the first threshold selection for 50 seconds is as follows: the exposure is reduced by 3.1%, the click rate is reduced by 6.29%, the click rate is reduced by 3.29%, the stay time of a single picture and/or video is increased by 6.96%, the per-person stay time is increased by 11.5%, the stay time of an android system is increased by 6.81%, the stay time of an apple system (IOS) is increased by 7.53%, and the stay time of a small window is increased by 14.02%. According to the simulation result, the information flow model established by adopting the first logarithmic function calculation method and selecting 50 seconds as the first threshold can effectively improve the stay time of the user, but the click rate of the user is correspondingly reduced.
According to the information flow model construction method provided by the embodiment, the stay time is divided by the first threshold value, and different processing modes are adopted, so that the method provided by the embodiment can reduce the error of the weight of the calculated stay time when the stay time is low, the accuracy of calculating the stay time is improved, and the reliability of the simulation result of the information flow model is improved.
Fig. 7 is a flowchart of another method for calculating the stay time length weight according to an embodiment of the present application. As shown in fig. 7, on the basis of the above embodiment, calculating the stay time length weight includes:
s40: and judging whether the stay time length is larger than a second threshold, if not, entering S41, and if so, entering S42. Wherein the second threshold is smaller than the first threshold in the above embodiments.
S41: setting the third dwell-time duration weight to a second fixed value.
S42: and calculating a fourth dwell-time length weight according to the second logarithmic function.
It is understood that the second threshold and the second fixed value are not particularly limited, and as a preferred embodiment, the second threshold may be selected to be 4 seconds, and the second fixed value may be selected to be 1.0.
In order to make the flowchart for calculating the stay time length weight provided by the embodiment of the present application more clear to those skilled in the art, the embodiment will be specifically described by taking the example that the second threshold is 4.
In the case where the stay time period is not more than 4 seconds, the third stay time period weight Ti is set3Is 1.0; in the case where the stay time period is longer than 4 seconds, the fourth stay time period weight is calculated by the following formula:
Ti4＝log1.4(time-t2)
wherein, Ti4Is a fourth dwell time duration weight, t2Is a fixed value not greater than a second threshold, time is the dwell time of a sample having a dwell time not greater than 4 seconds, where time is greater than t2. It will be appreciated that when the second threshold is 4, t2The value of (d) may be 3.
It should be noted that the base number in the fourth logarithmic function is determined by the distribution index such as the mean, the variance, the quantile, and the like of the residence time of each sample, and therefore the base number of the logarithmic function of 1.4 is only one representation form, and does not represent only one representation form.
In the embodiment of the present application, the fourth dwell-time weight may also be directly calculated by a default logarithmic function, where the default logarithmic function includes a base-2 logarithmic function, a base-10 logarithmic function, and a base-e logarithmic function.
The final simulation result of the information flow model established by the calculation method of the second logarithm function and the second fixed value is as follows: the exposure is increased by 3.73%, the click rate is increased by 7.21%, the click rate is increased by 3.46%, the number of clicks is increased by 2.95%, the dwell time of individual pictures and/or videos is reduced by 3.8%, and the page dwell time (the total dwell time of the user in the app) is increased by 12%. According to the simulation result, the information flow model established by the calculation method of the second logarithm function and the second fixed value not only further improves the stay time of the user, but also improves the click rate and the number of clicks of the user.
Fig. 8 is a graph comparing an interaction weight, a dwell period weight determined using a first logarithmic function and a linear function, and a dwell period weight determined using a second logarithmic function and a second fixed value, provided by an embodiment of the present application. As shown in fig. 8, the abscissa is the number of interactive events in one positive sample, and the ordinate is the weight value. As can be seen from fig. 8, compared with the dwell time weight determined by using the first logarithmic function and the linear function, the dwell time weight calculated by the method provided in the embodiment of the present application is smaller than the interaction weight, so that the influence of the dwell time on the simulation result of the information flow model is reduced, and the influence of the interaction data on the simulation result of the information flow model is increased, thereby further improving the reliability of the simulation result of the information flow model.
On the basis of the above embodiment, before S11, the method further includes: and under the condition that the stay time length is greater than the third threshold value, setting the stay time length as the third threshold value. Wherein the third threshold is greater than the first threshold in the above embodiment.
In the embodiment of the present application, the third threshold is not specifically limited, and may be matched with the requirement. As a preferred embodiment, the third threshold may take 300 seconds.
In the method for constructing the information flow model, the staying time longer than the third threshold is set as the third threshold, so that the problem that the confidence of the staying time of the total samples is reduced due to the overlong staying time is solved, and the influence of the staying time on the simulation result of the information flow model is further reduced, so that the reliability of the simulation result of the information flow model is further improved, and the staying time and the retention rate of a user are further improved.
In the above embodiments, the information flow model construction method is described in detail, and the present application also provides embodiments corresponding to the information flow model construction device. It should be noted that the present application describes the embodiments of the apparatus portion from two perspectives, one from the perspective of the function module and the other from the perspective of the hardware.
Fig. 9 is a schematic structural diagram of an information flow model building apparatus according to an embodiment of the present application. As shown in fig. 9, the apparatus includes, based on the angle of the function module:
and the acquisition module 10 is used for acquiring the residence time and the interaction data of each sample.
And the determining module 11 is configured to determine a dwell time weight corresponding to the dwell time and an interaction weight corresponding to the interaction data to obtain a weight of each sample.
And the building module 12 is configured to build an information flow model according to the weight corresponding to each sample.
Since the embodiments of the apparatus portion and the method portion correspond to each other, please refer to the description of the embodiments of the method portion for the embodiments of the apparatus portion, which is not repeated here.
As a preferred embodiment, the determining module 11 specifically includes:
the dividing module is used for dividing the sample with at least one of the stay time and/or the interactive data into a positive sample and dividing the sample without the stay time and the interactive data into a negative sample.
And the first calculation module is used for calculating each interaction weight and each stay time length weight respectively.
And the second calculation module is used for calculating the weight corresponding to each positive sample according to each stay time length weight and each interaction weight.
And the first setting module is used for setting the weight corresponding to each negative sample as a first fixed value, wherein the weight of the positive sample is greater than that of the negative sample.
The first calculation module includes:
and the processing module is used for carrying out linear processing on the stay time to obtain a first stay time weight under the condition that the stay time is not greater than the first threshold.
And the third calculating module is used for calculating the weight of the second stay time length according to the first logarithmic function under the condition that the stay time length is greater than the first threshold value.
The first calculation module includes:
and the second setting module is used for setting the third stay time length weight as a second fixed value under the condition that the stay time length is not more than a second threshold value, wherein the second threshold value is less than the first threshold value.
And the fourth calculating module is used for calculating the fourth stay time length weight according to the second logarithmic function under the condition that the stay time length is greater than the second threshold value.
Further comprising:
and the third setting module is used for setting the stay time length as a third threshold value under the condition that the stay time length is greater than the third threshold value, wherein the third threshold value is smaller than the first threshold value.
The information flow model building device provided in this embodiment obtains the retention time and the interaction data of each sample, calculates the weight of each sample according to the weight corresponding to each retention time and interaction data, and builds the information flow model according to the weight corresponding to each sample. Because the weights corresponding to the stay time lengths and the interactive data are respectively calculated, the confidence coefficient of the stay time lengths of the samples is improved. If the original stay time is directly added into the information flow model, other core indexes on the line can obviously drop, so that the influence of the directly added original stay time on the prediction effect of the information flow model is reduced, the reliability of the simulation result of the information flow model is improved, the matching degree of recommended pictures and/or video contents is improved, and the stay time and the retention rate of a user are further improved.
Fig. 10 is a schematic structural diagram of another information flow model building apparatus according to an embodiment of the present application. As shown in fig. 10, the apparatus includes, from the perspective of the hardware configuration:
a memory 20 for storing a computer program;
a processor 21, configured to implement the steps of the information flow model building method in the above embodiments when executing the computer program.
The processor 21 may include one or more processing cores, such as a 4-core processor, an 8-core processor, and the like. The processor 21 may be implemented in at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), and Programmable Logic Array (PLA). The processor 21 may also include a main processor and a coprocessor, where the main processor is a processor for Processing data in an awake state, and is also called a Central Processing Unit (CPU); a coprocessor is a low power processor for processing data in a standby state. In some embodiments, the processor 21 may be integrated with a Graphics Processing Unit (GPU) which is responsible for rendering and drawing the content required to be displayed on the display screen.
The memory 20 may include one or more computer-readable storage media, which may be non-transitory. Memory 20 may also include high speed random access memory, as well as non-volatile memory, such as one or more magnetic disk storage devices, flash memory storage devices. In this embodiment, the memory 20 is at least used for storing the following computer program 201, wherein after being loaded and executed by the processor 21, the computer program can implement the relevant steps of the information flow model building method disclosed in any of the foregoing embodiments. In addition, the resources stored in the memory 20 may also include an operating system 202, data 203, and the like, and the storage manner may be a transient storage manner or a permanent storage manner. Operating system 202 may include, among others, Windows, Unix, Linux, and the like. Data 203 may include, but is not limited to, data involved in the information flow model construction method, and the like.
In some embodiments, the information flow modeling apparatus may further include a display 22, an input/output interface 23, a communication interface 24, a power supply 25, and a communication bus 26.
Those skilled in the art will appreciate that the architecture shown in FIG. 10 does not constitute a limitation of the information flow model building apparatus and may include more or fewer components than those shown.
The information flow model construction device provided by the embodiment of the application comprises a memory and a processor, wherein when the processor executes a program stored in the memory, the following method can be realized: and after the stay time and the interactive data of each sample are obtained, calculating the weight of each sample according to the corresponding weight of each stay time and the interactive data, and constructing an information flow model according to the corresponding weight of each sample. Because the weights corresponding to the stay time lengths and the interactive data are respectively calculated, the confidence coefficient of the stay time lengths of the samples is improved. If the original stay time is directly added into the information flow model, other core indexes on the line can obviously drop, so that the influence of the directly added original stay time on the prediction effect of the information flow model is reduced, the reliability of the simulation result of the information flow model is improved, the matching degree of recommended pictures and/or video contents is improved, and the stay time and the retention rate of a user are further improved.
Finally, the application also provides a corresponding embodiment of the computer readable storage medium. The computer-readable storage medium has stored thereon a computer program which, when being executed by a processor, carries out the steps as set forth in the above-mentioned method embodiments.
It is to be understood that if the method in the above embodiments is implemented in the form of software functional units and sold or used as a stand-alone product, it can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the embodiments of the present application may be embodied in the form of a software product, which is stored in a storage medium and executes all or part of the steps of the methods described in the embodiments of the present application, or all or part of the technical solutions. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
The computer readable storage medium provided by the embodiment of the application, the medium is stored with a computer program, and when the computer program is executed by a processor, the following method can be realized: and after the stay time and the interactive data of each sample are obtained, calculating the weight of each sample according to the corresponding weight of each stay time and the interactive data, and constructing an information flow model according to the corresponding weight of each sample. Because the weights corresponding to the stay time lengths and the interactive data are respectively calculated, the confidence coefficient of the stay time lengths of the samples is improved. If the original stay time is directly added into the information flow model, other core indexes on the line can obviously drop, so that the influence of the directly added original stay time on the prediction effect of the information flow model is reduced, the reliability of the simulation result of the information flow model is improved, the matching degree of recommended pictures and/or video contents is improved, and the stay time and the retention rate of a user are further improved.
The method, the device and the medium for constructing the information flow model provided by the application are described in detail above. The embodiments are described in a progressive manner in the specification, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. The device disclosed by the embodiment corresponds to the method disclosed by the embodiment, so that the description is simple, and the relevant points can be referred to the method part for description. It should be noted that, for those skilled in the art, it is possible to make several improvements and modifications to the present application without departing from the principle of the present application, and such improvements and modifications also fall within the scope of the claims of the present application.
It is further noted that, in the present specification, relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
Claims (10)
1. An information flow model construction method, characterized by comprising:
obtaining the stay time and interaction data of each sample;
determining a dwell time weight corresponding to the dwell time and an interaction weight corresponding to the interaction data to obtain a weight of each sample;
and constructing an information flow model according to the weight corresponding to each sample.
2. The method of claim 1, wherein determining the dwell time and the corresponding weight of the interaction data to obtain the weight of each sample comprises:
dividing the sample with at least one of the stay time and/or the interaction data into a positive sample, and dividing the sample without the stay time and the interaction data into a negative sample;
respectively calculating each interaction weight and each stay time length weight;
calculating the weight corresponding to each positive sample according to each stay time length weight and each interaction weight;
setting the weight corresponding to each negative sample as a first fixed value;
wherein the weight corresponding to the positive sample is greater than the first fixed value.
3. The method of claim 2, wherein calculating the dwell time weight comprises:
under the condition that the stay time length is not greater than a first threshold value, carrying out linear processing on the stay time length to obtain a first stay time length weight;
and under the condition that the stay time length is larger than the first threshold value, calculating a second stay time length weight according to a first logarithmic function.
4. The information flow model building method of claim 3, wherein calculating the dwell time duration weight comprises:
setting a third staying time length weight as a second fixed value under the condition that the staying time length is not more than a second threshold value;
under the condition that the stay time length is larger than the second threshold value, calculating a fourth stay time length weight according to a second logarithmic function;
wherein the second threshold is less than the first threshold.
5. The method of claim 3, wherein before determining the dwell time and the corresponding weight of the interaction data to obtain the weight of each sample, the method further comprises:
setting the stay time length as a third threshold value under the condition that the stay time length is greater than the third threshold value;
wherein the third threshold is greater than the first threshold.
6. The information flow model building method according to claims 1 to 5, wherein the interaction data comprises: like data, comment data, share data, concern data, and collection data.
7. The information flow model building method according to claims 3 to 5, characterized in that the first threshold is specifically 50 seconds, the second threshold is specifically 4 seconds, and the third threshold is specifically 300 seconds.
8. An information flow model building apparatus, comprising:
the acquisition module is used for acquiring the stay time and the interaction data of each sample;
the determining module is used for determining a dwell time weight corresponding to the dwell time and an interaction weight corresponding to the interaction data to obtain a weight of each sample;
and the construction module is used for constructing an information flow model according to the weight corresponding to each sample.
9. An information flow model building apparatus, comprising:
a memory for storing a computer program;
a processor for implementing the steps of the information flow model construction method according to any one of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium, characterized in that a computer program is stored thereon, which computer program, when being executed by a processor, carries out the steps of the information flow model building method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011189611.7A CN112200639A (en) | 2020-10-30 | 2020-10-30 | Information flow model construction method, device and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011189611.7A CN112200639A (en) | 2020-10-30 | 2020-10-30 | Information flow model construction method, device and medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112200639A true CN112200639A (en) | 2021-01-08 |
Family
ID=74011145
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011189611.7A Pending CN112200639A (en) | 2020-10-30 | 2020-10-30 | Information flow model construction method, device and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112200639A (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017164547A1 (en) * | 2016-03-24 | 2017-09-28 | (주)잼투고 | Immersion event-based content weight adjustment system and content weight adjustment method using same |
| CN108304441A (en) * | 2017-11-14 | 2018-07-20 | 腾讯科技（深圳）有限公司 | Network resource recommended method, device, electronic equipment, server and storage medium |
| CN109327739A (en) * | 2018-11-27 | 2019-02-12 | 广州虎牙信息科技有限公司 | A kind of method for processing video frequency, calculates equipment and storage medium at device |
| CN109902753A (en) * | 2019-03-06 | 2019-06-18 | 深圳市珍爱捷云信息技术有限公司 | User's recommended models training method, device, computer equipment and storage medium |
| CN110266745A (en) * | 2019-03-08 | 2019-09-20 | 腾讯科技（深圳）有限公司 | Information flow recommended method, device, equipment and storage medium based on depth network |
| CN111079061A (en) * | 2019-12-23 | 2020-04-28 | 北京奇艺世纪科技有限公司 | Data processing method and electronic equipment |
| CN111310053A (en) * | 2020-03-03 | 2020-06-19 | 上海喜马拉雅科技有限公司 | Information recommendation method, device, equipment and storage medium |
| CN111405325A (en) * | 2020-03-25 | 2020-07-10 | 北京达佳互联信息技术有限公司 | Video content distribution method and device and electronic equipment |
-
2020
- 2020-10-30 CN CN202011189611.7A patent/CN112200639A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017164547A1 (en) * | 2016-03-24 | 2017-09-28 | (주)잼투고 | Immersion event-based content weight adjustment system and content weight adjustment method using same |
| CN108304441A (en) * | 2017-11-14 | 2018-07-20 | 腾讯科技（深圳）有限公司 | Network resource recommended method, device, electronic equipment, server and storage medium |
| CN109327739A (en) * | 2018-11-27 | 2019-02-12 | 广州虎牙信息科技有限公司 | A kind of method for processing video frequency, calculates equipment and storage medium at device |
| CN109902753A (en) * | 2019-03-06 | 2019-06-18 | 深圳市珍爱捷云信息技术有限公司 | User's recommended models training method, device, computer equipment and storage medium |
| CN110266745A (en) * | 2019-03-08 | 2019-09-20 | 腾讯科技（深圳）有限公司 | Information flow recommended method, device, equipment and storage medium based on depth network |
| CN111079061A (en) * | 2019-12-23 | 2020-04-28 | 北京奇艺世纪科技有限公司 | Data processing method and electronic equipment |
| CN111310053A (en) * | 2020-03-03 | 2020-06-19 | 上海喜马拉雅科技有限公司 | Information recommendation method, device, equipment and storage medium |
| CN111405325A (en) * | 2020-03-25 | 2020-07-10 | 北京达佳互联信息技术有限公司 | Video content distribution method and device and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109300179B (en) | Animation production method, device, terminal and medium | |
| CN111090756B (en) | Artificial intelligence-based multi-target recommendation model training method and device | |
| US10537801B2 (en) | System and method for decision making in strategic environments | |
| US20200143000A1 (en) | Customized display of emotionally filtered social media content | |
| US11682415B2 (en) | Automatic video tagging | |
| CN111984821A (en) | Method and device for determining dynamic cover of video, storage medium and electronic equipment | |
| CN116363261A (en) | Training method of image editing model, image editing method and device | |
| CN113722588B (en) | Resource recommendation method and device and electronic equipment | |
| CN110909005B (en) | Model feature analysis method, device, equipment and medium | |
| CN112182281B (en) | Audio recommendation method, device and storage medium | |
| CN109377103B (en) | Evaluation method and device for course recommendation of learning platform | |
| CN110546678B (en) | Computationally derived assessment in a child education system | |
| Martens | Interactive statistics with illmo | |
| US20170262420A1 (en) | Persisting annotations applied to an electronic hosted whiteboard | |
| WO2023185125A1 (en) | Product resource data processing method and apparatus, electronic device and storage medium | |
| CN112200639A (en) | Information flow model construction method, device and medium | |
| CN113032251B (en) | Method, device and storage medium for determining service quality of application program | |
| CN114143571B (en) | User processing method, device, equipment and storage medium | |
| CN113849758A (en) | Webpage index generation method and device, electronic equipment and storage medium | |
| US11303683B2 (en) | Methods and systems for managing distribution of online content based on content maturity | |
| CN116150327A (en) | Text processing method and device | |
| CN109543557B (en) | Video frame processing method, device, equipment and storage medium | |
| CN109040164B (en) | Application recommendation method and device, storage medium and computer equipment | |
| US10884804B2 (en) | Information gathering command generator | |
| US20200279164A1 (en) | Discrete feature representation with class priority |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210108 |
|
| RJ01 | Rejection of invention patent application after publication |