CN117495459A - Man-machine interaction advertisement method, device, equipment and storage medium based on big data - Google Patents
Man-machine interaction advertisement method, device, equipment and storage medium based on big data Download PDFInfo
- Publication number
- CN117495459A CN117495459A CN202410001394.6A CN202410001394A CN117495459A CN 117495459 A CN117495459 A CN 117495459A CN 202410001394 A CN202410001394 A CN 202410001394A CN 117495459 A CN117495459 A CN 117495459A
- Authority
- CN
- China
- Prior art keywords
- advertisement
- application
- trigger time
- model
- time proportion
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 86
- 230000003993 interaction Effects 0.000 title claims abstract description 50
- 241000282461 Canis lupus Species 0.000 claims description 88
- 239000013598 vector Substances 0.000 claims description 53
- 238000004422 calculation algorithm Methods 0.000 claims description 38
- 230000008569 process Effects 0.000 claims description 37
- 238000005457 optimization Methods 0.000 claims description 24
- 241000282421 Canidae Species 0.000 claims description 22
- 238000004891 communication Methods 0.000 claims description 21
- 238000007637 random forest analysis Methods 0.000 claims description 18
- 230000001960 triggered effect Effects 0.000 claims description 18
- 238000003066 decision tree Methods 0.000 claims description 15
- 238000012549 training Methods 0.000 claims description 13
- 238000004590 computer program Methods 0.000 claims description 8
- 238000012795 verification Methods 0.000 claims description 8
- 238000004364 calculation method Methods 0.000 claims description 7
- 230000006870 function Effects 0.000 claims description 7
- 230000002195 synergetic effect Effects 0.000 claims description 3
- 230000000977 initiatory effect Effects 0.000 claims description 2
- 230000000694 effects Effects 0.000 abstract description 8
- 238000013461 design Methods 0.000 description 9
- 238000010586 diagram Methods 0.000 description 3
- 230000006872 improvement Effects 0.000 description 2
- 239000002245 particle Substances 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000002537 cosmetic Substances 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000006855 networking Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
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
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0251—Targeted advertisements
Abstract
The application discloses a man-machine interaction advertisement method, device and equipment based on big data and a storage medium, and relates to the technical field of intelligent advertisements. According to the method, firstly, a big data modeling is started according to historical application of a target user to obtain advertisement skip trigger time proportion prediction models, then the models are applied to obtain advertisement skip trigger time proportion prediction values of all advertisements to be pushed, then when a certain application is detected to be started, a certain advertisement to be pushed with a prediction maximum value is loaded onto an application starting page of the certain application, an advertisement skip trigger time proportion actual value is determined, finally, iteration optimizing is carried out on model parameters according to the actual value, and the proportion prediction values of all advertisements to be pushed are refreshed based on optimizing results, so that advertisement self-adaption pushing can be carried out according to user man-machine interaction results, the pushed advertisement content is not fixed any more, and accuracy of a final advertisement content pushing result and advertisement putting effect are ensured.
Description
Technical Field
The invention belongs to the technical field of intelligent advertising, and particularly relates to a man-machine interaction advertising method, device and equipment based on big data and a storage medium.
Background
With the continuous development of society and economy, people networking to acquire network data has become a main information acquisition mode, and has taken up most of the time of people's daily life, for example people often brush microblogs, learn how to brush train tickets and the like through smart phones. Under the background of the times, the traditional advertising mode based on media such as advertising machines or newspapers gradually declines, so that advertisers pay more attention to embedding advertisements into a starting page of an APP (Application program) Application program for popularization, for example, in the process of starting train ticket booking software at a smart phone end for 1-3 seconds, dynamic or static advertising content is loaded and displayed on the Application starting page, and then when a user clicks a skip button or reaches a preset display duration, the advertisement is automatically switched to a homepage of the train ticket booking software.
At present, the advertisement content recommendation scheme of the new advertisement mode mainly adopts an application service provider to appoint and push commercial advertisements which are cooperated and promoted, so that the advertisement content is relatively fixed, and personalized self-adaptive pushing cannot be carried out according to the man-machine interaction result of a user clicking a skip button, so that the advertisement content pushing accuracy and the advertisement putting effect are still to be further improved.
Disclosure of Invention
The invention aims to provide a man-machine interaction advertisement method, a man-machine interaction advertisement device, computer equipment and a computer readable storage medium based on big data, which are used for solving the problems that the advertisement content is relatively fixed and personalized self-adaptive pushing cannot be carried out according to the man-machine interaction result of a user clicking a skip button in the conventional advertisement content recommendation scheme.
In order to achieve the above purpose, the present invention adopts the following technical scheme:
in a first aspect, a man-machine interaction advertising method based on big data is provided, including:
acquiring historical application starting big data of a target user, wherein the historical application starting big data comprises a multidimensional advertisement characteristic value and an advertisement skipping trigger time proportion of a starting page pushing advertisement loaded and displayed in each application historical starting process, the starting page pushing advertisement is an advertisement loaded on an application starting page for pushing, the advertisement skipping trigger time proportion is a ratio of a time length from a loading time stamp of the starting page pushing advertisement to a trigger time stamp of an advertisement skipping event to a preset display time length of the starting page pushing advertisement, the advertisement skipping event is triggered on the application starting page by the target user in a man-machine interaction mode, or the advertisement skipping event is automatically triggered when the display time length of the starting page pushing advertisement reaches the preset display time length;
According to the historical application starting big data, taking the multidimensional advertisement characteristic values of the advertisement pushed by each starting page as input items and taking the advertisement skip trigger time proportion of the advertisement pushed by each starting page as output items, and carrying out rating verification modeling on an artificial intelligent model based on a random forest algorithm to obtain an advertisement skip trigger time proportion prediction model;
inputting the corresponding multidimensional advertisement characteristic values into the advertisement skipping triggering time proportion prediction model for each advertisement to be pushed, and outputting to obtain the corresponding advertisement skipping triggering time proportion prediction values;
when detecting that a certain application is started, loading a certain advertisement to be pushed, which currently has the maximum value of the advertisement skipping trigger time proportion prediction, onto an application starting page of the certain application, and determining the actual value of the advertisement skipping trigger time proportion according to the newly triggered advertisement skipping event;
according to the actual value of the advertisement skip trigger time proportion and the multidimensional advertisement characteristic value of the certain advertisement to be pushed, carrying out iterative optimization on the model super-parameters of the advertisement skip trigger time proportion prediction model by adopting a gray wolf optimization algorithm to obtain the current optimal model super-parameters;
Substituting the current optimal model super parameters into the advertisement skip trigger time proportion prediction model to obtain a new advertisement skip trigger time proportion prediction model;
and inputting the corresponding multidimensional advertisement characteristic values into the advertisement skip trigger time proportion prediction new model aiming at each advertisement to be pushed, and outputting to obtain the corresponding advertisement skip trigger time proportion prediction new values so as to select the pushing advertisement which is most suitable for loading when the next application is started.
Based on the above-mentioned invention content, a new scheme for accurately predicting advertisement skip trigger time proportion based on historical application start big data and random forest algorithm and applying push advertisement is provided, namely, firstly, according to the historical application start big data modeling of target users, advertisement skip trigger time proportion prediction model is obtained, then the model is applied to obtain advertisement skip trigger time proportion prediction value of each advertisement to be pushed, then when detecting that a certain application is started, a certain advertisement to be pushed with a prediction maximum value is loaded onto an application start page of the certain application, and the actual value of advertisement skip trigger time proportion is determined, finally, according to the actual value, iterative optimization is carried out on model parameters, and based on optimizing results, advertisement personalized self-adaptive push is carried out according to user man-machine interaction results, so that push advertisement content is not fixed any more, accuracy of the push result of the final advertisement content and advertisement release effect are ensured, and practical application and popularization are facilitated.
In one possible design, the multi-dimensional advertisement feature value includes an advertisement type number value, a target audience group type number value, and/or a keyword number value.
In one possible design, starting big data according to the historical application, taking multidimensional advertisement feature values of each starting page push advertisement as input items, including:
selecting a plurality of start page push advertisements in the history starting process of the last application according to the history application starting big data, wherein the plurality of start page push advertisements are in one-to-one correspondence with the history starting process of the last application;
and taking the multidimensional advertisement characteristic value of each start page pushing advertisement in the plurality of start page pushing advertisements as an input item.
In one possible design, selecting a plurality of start pages to push advertisements in a last plurality of application history starts according to the history application start big data, including:
determining occurrence in the latest unit period based on the historical application initiation big dataA secondary application history start-up procedure, wherein +.>Represents a positive integer of 10 or more;
will be in contact with theThe secondary application history starting process is one-to-one corresponding to +. >The start page push advertisement acts as a plurality of start page push advertisements during the last multiple application history starts.
In one possible design, when detecting that a certain application is started, loading a certain advertisement to be pushed, which currently has a maximum value of advertisement skip trigger time proportion prediction, onto an application starting page of the certain application, including:
when detecting that a certain application is started, loading a certain advertisement to be pushed in a plurality of advertisements to be pushed and currently having an advertisement skip trigger time proportion prediction maximum value onto an application starting page of the certain application, wherein the advertisements to be pushed are updated and downloaded from a server of the certain application in advance.
In one possible design, according to the actual value of the advertisement skip trigger time proportion and the multidimensional advertisement characteristic value of the advertisement to be pushed, performing iterative optimization on the model super-parameters of the advertisement skip trigger time proportion prediction model by adopting a gray wolf optimization algorithm to obtain the current optimal model super-parameters, wherein the method comprises the following steps of S51-S58:
s51, initializing a population: the number of the wolves is set asThe iteration number is set to +. >Next, and initializing search ranges of at least two model super parameters of the advertisement skip trigger time scale prediction model, and then performing step S52, wherein,represents a positive integer of 5 or more, < >>Represents a positive integer of 100 or more;
s52, initializing the gray wolves: at the position ofThree wolves were randomly selected as the initial +.>Wolf and jersey>Wolf and->Wolf, and is initially set at said ++within the search range of said at least two model hyper-parameters>Individual position vectors of each of the individual wolves, wherein the individual position vectors contain the search values of the at least two model hyper-parameters, are then performed in step S53;
s53, inputting the multidimensional advertisement characteristic value of the certain advertisement to be pushed into the advertisement skip trigger time proportion prediction model substituted into the corresponding current individual position vector, taking the absolute difference value of the actual advertisement skip trigger time proportion value and the corresponding output advertisement skip trigger time proportion prediction value as the corresponding individual fitness value, and then executing step S54;
s54, judging whether the current iteration number reachesStep S58 is executed if yes, otherwise the gray wolf with the minimum individual fitness is taken as new +. >Wolf's will have individual fitnessThe next largest value of wolf is taken as new +.>Wolf and also wolf with a further high individual fitness value as new +.>Wolf, then step S55 is performed;
s55, respectively calculating convergence factorsSynergistic vector->And synergistic vector->Step S56 is then performed, wherein the convergence factor +.>Said synergy vector->And said synergy vector->The calculation formulas of (a) are respectively as follows:
in the method, in the process of the invention,representing said current iteration number,/->Representing hyperbolic tangent function, ">And->Respectively represent [0,1 ]]Is a random vector of (a);
s56, for eachWolf, according to said new ++>Wolf and jersey>Wolf and->The current individual position vector of wolf is calculated to obtain the corresponding and +.>Individual position vectors in the multiple iterations->Step S57 is then performed, wherein the individual position vector +.>The method is calculated according to the following formula:
in the method, in the process of the invention,representing said new->Current individual position vector of wolf, +.>Representing said new->Current individual position vector of wolf, +.>Representing said new->Current individual position vector of wolf, +.>Is indicated at +.>Individual position vectors in the multiple iterations, +.>、/>And->Respectively represent the synergy vector calculated randomly +. >，/>、/>And->Respectively represent the synergy vector calculated randomly +.>；
S57, adding 1 to the iteration times, and then returning to execute the step S53;
s58, taking the current individual position vector of the gray wolf with the minimum individual fitness as the current optimal model super-parameter.
In one possible design, the at least two model hyper-parameters include a number of decision tree particles and a number of leaf nodes of the random forest algorithm.
The second aspect provides a man-machine interaction advertising device based on big data, which comprises a big data acquisition module, a model training module, a model application module, an advertisement loading module, a parameter optimizing module and a model updating module;
the big data acquisition module is used for acquiring historical application starting big data of a target user, wherein the historical application starting big data comprises a multidimensional advertisement characteristic value and an advertisement skip trigger time proportion of a starting page pushing advertisement loaded and displayed in each application historical starting process, the starting page pushing advertisement is an advertisement loaded on an application starting page for pushing, the advertisement skip trigger time proportion is a ratio of a time period from a loading time stamp of the starting page pushing advertisement to a trigger time stamp of an advertisement skip event to a preset display time period of the starting page pushing advertisement, and the advertisement skip event is triggered on the application starting page by the target user in a man-machine interaction mode or is automatically triggered when the display time period of the starting page pushing advertisement reaches the preset display time period;
The model training module is in communication connection with the big data acquisition module and is used for starting big data according to the historical application, taking the multidimensional advertisement characteristic values of the advertisements pushed by all the starting pages as input items and taking the advertisement skip trigger time proportion of the advertisements pushed by all the starting pages as output items, and carrying out rated verification modeling on the artificial intelligent model based on a random forest algorithm to obtain an advertisement skip trigger time proportion prediction model;
the model application module is in communication connection with the model training module and is used for inputting the corresponding multidimensional advertisement characteristic values into the advertisement skipping trigger time proportion prediction model for each advertisement to be pushed, and outputting the corresponding advertisement skipping trigger time proportion prediction values;
the advertisement loading module is in communication connection with the model application module and is used for loading a certain advertisement to be pushed, which currently has the maximum value of the advertisement skipping trigger time proportion prediction, onto an application starting page of the certain application when the fact that the certain application is started is detected, and determining the actual value of the advertisement skipping trigger time proportion according to the newly triggered advertisement skipping event;
the parameter optimizing module is in communication connection with the advertisement loading module and is used for carrying out iterative optimization on the model super-parameters of the advertisement skip trigger time proportion prediction model by adopting a gray wolf optimization algorithm according to the actual value of the advertisement skip trigger time proportion and the multidimensional advertisement characteristic value of the certain advertisement to be pushed so as to obtain the current optimal model super-parameters;
The model updating module is respectively in communication connection with the model training module and the parameter optimizing module and is used for substituting the current optimal model super parameters into the advertisement skip trigger time proportion prediction model to obtain a new advertisement skip trigger time proportion prediction model;
the model application module is also in communication connection with the model updating module and is further used for inputting the corresponding multidimensional advertisement characteristic values into the advertisement skip trigger time proportion prediction new model aiming at each advertisement to be pushed, and outputting the corresponding advertisement skip trigger time proportion prediction new values so as to select the pushing advertisement which is most suitable for loading when the next application is started.
In a third aspect, the present invention provides a computer device comprising a memory, a processor and a transceiver in communication connection in turn, wherein the memory is configured to store a computer program, the transceiver is configured to send and receive messages, and the processor is configured to read the computer program and perform the man-machine interaction advertising method according to the first aspect or any of the possible designs of the first aspect.
In a fourth aspect, the present invention provides a computer readable storage medium having instructions stored thereon which, when executed on a computer, perform the human-machine interaction advertising method as described in the first aspect or any of the possible designs of the first aspect.
In a fifth aspect, the present invention provides a computer program product comprising instructions which, when run on a computer, cause the computer to perform the human-machine interaction advertising method as described in the first aspect or any of the possible designs of the first aspect.
The beneficial effect of above-mentioned scheme:
the invention creatively provides a new scheme for accurately predicting advertisement skip trigger time proportion based on historical application starting big data and a random forest algorithm and applying push advertisements, namely, firstly, modeling according to the historical application starting big data of a target user to obtain advertisement skip trigger time proportion prediction models, then, applying the models to obtain advertisement skip trigger time proportion prediction values of all advertisements to be pushed, then, when detecting that a certain application is started, loading a certain advertisement to be pushed with a prediction maximum value onto an application starting page of the certain application, determining an actual value of advertisement skip trigger time proportion, finally, performing iterative optimization according to model parameters according to the actual value, refreshing the proportion prediction values of all advertisements to be pushed based on an optimizing result, so that advertisement personalized self-adaptive push can be performed according to a user man-machine interaction result, the push advertisement content is not immobilized any more, the accuracy of a final advertisement content push result and advertisement release effect are ensured, and practical application and popularization are facilitated.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a flow chart of a man-machine interaction advertising method based on big data according to an embodiment of the present application.
Fig. 2 is a diagram illustrating a structure of an artificial intelligence model based on a random forest algorithm according to an embodiment of the present application.
Fig. 3 is a schematic structural diagram of a man-machine interaction advertising device based on big data according to an embodiment of the present application.
Fig. 4 is a schematic structural diagram of a computer device according to an embodiment of the present application.
Detailed Description
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the present invention will be briefly described below with reference to the accompanying drawings and the description of the embodiments or the prior art, and it is obvious that the following description of the structure of the drawings is only some embodiments of the present invention, and other drawings can be obtained according to these drawings without inventive effort to a person skilled in the art. It should be noted that the description of these examples is for aiding in understanding the present invention, but is not intended to limit the present invention.
It should be understood that although the terms first and second, etc. may be used herein to describe various objects, these objects should not be limited by these terms. These terms are only used to distinguish one object from another. For example, a first object may be referred to as a second object, and similarly a second object may be referred to as a first object, without departing from the scope of example embodiments of the invention.
It should be understood that for the term "and/or" that may appear herein, it is merely one association relationship that describes an associated object, meaning that there may be three relationships, e.g., a and/or B, may represent: three cases of A alone, B alone or both A and B exist; as another example, A, B and/or C, can represent the presence of any one of A, B and C or any combination thereof; for the term "/and" that may appear herein, which is descriptive of another associative object relationship, it means that there may be two relationships, e.g., a/and B, it may be expressed that: the two cases of A and B exist independently or simultaneously; in addition, for the character "/" that may appear herein, it is generally indicated that the context associated object is an "or" relationship.
Examples
As shown in fig. 1, the man-machine interaction advertisement method based on big data provided in the first aspect of the present embodiment may be performed by, but not limited to, a computer device with a certain computing resource, for example, a platform server, a personal computer (Personal Computer, PC, refer to a multipurpose computer with a size, price and performance suitable for personal use, a desktop computer, a notebook computer, a small notebook computer, a tablet computer, an ultrabook, etc. all belong to a personal computer), a smart phone, a personal digital assistant (Personal Digital Assistant, PDA) or an electronic device such as a wearable device. As shown in FIG. 1, the man-machine interaction advertising method can include, but is not limited to, the following steps S1 to S7.
S1, acquiring historical application starting big data of a target user, wherein the historical application starting big data comprises, but is not limited to, multidimensional advertisement characteristic values of starting page pushing advertisements loaded and displayed in each application historical starting process and advertisement skipping trigger time proportion, the starting page pushing advertisements refer to advertisements loaded on an application starting page to be pushed, the advertisement skipping trigger time proportion refers to the ratio of the time from the loading time stamp of the starting page pushing advertisements to the trigger time stamp of advertisement skipping events to the preset display time of the starting page pushing advertisements, and the advertisement skipping events are triggered on the application starting page by the target user in a man-machine interaction mode or are automatically triggered when the display time of the starting page pushing advertisements reaches the preset display time.
In the step S1, the target user is the advertisement content pushing object. The historical application starting big data can be obtained from a user terminal side or an application service side in a conventional way but is not limited to the historical application starting big data. The multi-dimensional advertisement characteristic value is used for reflecting the characteristics of the pushed advertisement content from multiple dimensions, and specifically comprises, but is not limited to, advertisement type number value, target audience group type number value and/or keyword number value, wherein the advertisement type number value can be, but is not limited to, a number value corresponding to a static image-text advertisement type, a number value corresponding to a dynamic image-text advertisement type or a number value corresponding to an audio-video advertisement type, and the like; the target audience group type number value may be, but is not limited to, a number value corresponding to a teenager student, a number value corresponding to a young man, a number value corresponding to a middle-aged woman, a number value corresponding to an elderly person, or the like; the keyword number value may be, but not limited to, a number value corresponding to an automobile, a number value corresponding to cosmetics, a number value corresponding to travel, a number value corresponding to education, or the like. The advertisement skipping triggering time proportion is used for reflecting the attraction of the pushed advertisement content to the target user, and the advertisement skipping triggering time proportion is positively related to the advertisement attraction as the ratio is larger, namely the push advertisement content is more attracted to the target user, because the advertisement skipping triggering time proportion is the ratio of the time from the loading time stamp of the push advertisement of the starting page to the triggering time stamp of the advertisement skipping event to the preset display time of the push advertisement of the starting page; for example, if the preset presentation duration is 3 seconds and the target user triggers the advertisement skip event at the 2 nd second from the advertisement presentation, the advertisement skip trigger time ratio is about 0.67. In addition, the specific way of triggering the advertisement skipping event on the application starting page through a man-machine interaction way can be, but not limited to, a conventional way of clicking an advertisement skipping button on the application starting page.
S2, starting big data according to the historical application, taking multidimensional advertisement characteristic values of advertisements pushed by all starting pages as input items, taking advertisement skip trigger time proportion of the advertisements pushed by all starting pages as output items, and performing rated verification modeling on an artificial intelligent model based on a random forest algorithm to obtain an advertisement skip trigger time proportion prediction model.
In the step S2, the random forest algorithm is used as a representative of Bagging in the integrated algorithm, and by combining a plurality of decision tree models and processing the same in parallel, each decision tree is classified and calculated without interference, and the method has the characteristics of small calculation complexity, high calculation speed and the like. In the calculation process, the term "random" means that each decision tree randomly selects a part of samples from a training set to train, and simultaneously randomly selects a part of features from all input features to train, the decision tree is built according to the selected samples and features, and usually more than one decision tree is randomly built, then the steps are repeatedly performed to build a plurality of decision trees, for prediction samples, each decision tree is predicted, and a prediction result adopts a voting or average mode to give a final answer to a classification or regression problem, so that the finally built artificial intelligent model can be shown as a figure 2. The specific process of the rating verification modeling comprises a rating process and a checking process of the model, namely, the model parameters are adjusted according to the comparison result after the simulation result and the actual measured data of the model are compared, so that the simulation result is matched with the actual process, and the advertisement skip triggering time proportion prediction model can be obtained through a conventional rating verification modeling mode.
In the step S2, considering that the interest preference of the target user for different advertisements is dynamically changed, in order to obtain the advertisement skip trigger time proportion prediction model capable of accurately predicting the recent advertisement interest preference of the user, it is preferable to start big data according to the historical application, and take as input items the multidimensional advertisement feature values of the push advertisements of each start page, including but not limited to: firstly, according to the historical application starting big data, selecting a plurality of starting page pushing advertisements in the last multiple application historical starting process, wherein the plurality of starting page pushing advertisements are in one-to-one correspondence with the last multiple application historical starting process; and taking the multidimensional advertisement characteristic value of each start page pushing advertisement in the plurality of start page pushing advertisements as an input item. Because the time required by the application starting is shorter and is generally only a few seconds, only one push advertisement is loaded in one application history starting process, so that the push advertisements of the plurality of starting pages are in one-to-one correspondence with the latest multi-application history starting process; for example, 100 launch page push ads during the last 100 application history launches may be selected.
In the step S2, the specific number of the aforementioned plurality of start page push advertisements may be fixed or may be dynamically changed, specifically, according to the historical application start big data, a plurality of start page push advertisements in the last multiple application history start process are selected, including but not limited to: first, according to the history application, starting big data, determining the occurrence in the latest unit periodA secondary application history start-up procedure, wherein +.>Represents a positive integer of 10 or more; will be in contact with theThe secondary application history starting process is one-to-one corresponding to +.>The start page push advertisement acts as a plurality of start page push advertisements during the last multiple application history starts. The last unit period may be, but is not limited to, for example, the last 24 hours, the last week, the last month, or the like. Furthermore, if->Less than 10, the last unit period can also be extended appropriately (e.g. the last 24 hours to the last week) until +.>10 or more.
S3, inputting the corresponding multidimensional advertisement characteristic values into the advertisement skipping trigger time proportion prediction model aiming at each advertisement to be pushed, and outputting to obtain the corresponding advertisement skipping trigger time proportion prediction value.
In the step S3, the advertisement to be pushed may be stored locally in advance, and in particular, but not limited to, updated and downloaded from a server of each application.
S4, when detecting that a certain application is started, loading a certain advertisement to be pushed, which currently has the maximum value of the advertisement skipping trigger time proportion prediction, onto an application starting page of the certain application, and determining the actual value of the advertisement skipping trigger time proportion according to the newly triggered advertisement skipping event.
In the step S4, it is considered that the application server generally does not allow to recommend advertisements of other third parties on the application start page, so specifically, when detecting that a certain application is started, a certain advertisement to be pushed currently having a maximum value of advertisement skip trigger time proportion prediction is loaded on the application start page of the certain application, including but not limited to: when detecting that a certain application is started, loading a certain advertisement to be pushed in a plurality of advertisements to be pushed and currently having an advertisement skip trigger time proportion prediction maximum value onto an application starting page of the certain application, wherein the advertisements to be pushed are updated and downloaded from a server of the certain application in advance. That is, the application server may provide a plurality of advertisements to be pushed in advance, and then select the advertisement which is most suitable currently from the plurality of advertisements to be pushed for instant pushing when the application is started. In addition, the actual value of the advertisement skip trigger time proportion is used as a new human-computer interaction result of clicking a skip button by a user to reflect the attraction of the advertisement to the target user.
S5, according to the actual value of the advertisement skip trigger time proportion and the multidimensional advertisement characteristic value of the advertisement to be pushed, performing iterative optimization on the model super-parameters of the advertisement skip trigger time proportion prediction model by adopting a gray wolf optimization algorithm to obtain the current optimal model super-parameters.
In the step S5, the wolf optimization algorithm (Grey Wolf Optimizer, GWO) is an optimization algorithm (which mainly simulates the mechanism of hunting during wolf group hunting and the social status among wolves, which is respectively represented as hunting and level system) proposed by being inspired by the wolf hunting law in the natural environment, and has the following algorithm principles:
assume that the wolf group has four wolves with different status grades, respectively from top to bottomWolf and jersey>Wolf and jersey>Wolf and->The wolf with high position gives instructions to the wolf with low position, which first surrounds the prey, and this part of algorithm is expressed as follows:
in the method, in the process of the invention,indicating the distance between the individual gray wolf and the prey,/->Representing the current iteration number, +.>Position vector representing prey,/->Position vector representing individual gray wolves, +.>And->Respectively representing the synergy vectors, the calculation is as follows:
in the method, in the process of the invention,represents a convergence factor linearly decreasing from 2 to 0 in the course of an iteration, +. >And->Respectively represent [0,1 ]]Is a random vector of (a); then hunting is performed by +.>Wolf and jersey>Wolf and->Wolf leader->Wolves hunting, i.e. +.>Wolf and jersey>Wolf and->The position of wolves is not moving, and the patient is assy>Wolf iterates, the algorithm is as follows:
in the method, in the process of the invention,、/>and->Respectively represent +.>Wolf and jersey>Wolf and->Position vector of wolf,>、/>and->Respectively represent the synergy vector calculated randomly +.>，/>、/>And->Respectively represent the synergy vector calculated randomly +.>，/>、/>And->Respectively represent other individuals in the population and +.>Wolf and jersey>Wolf and->Distance between wolves, the->Is indicated at +.>Individual position vectors in the multiple iterations, +.>Is indicated at +.>Individual position vectors in the multiple iterations. In this way, considering that a reliable prediction model based on a random forest algorithm is desired, the numerical value of the super-parameter needs to be reasonably set, that is, specifically, according to the actual value of the advertisement skip trigger time proportion and the multidimensional advertisement characteristic value of the advertisement to be pushed, a gray wolf optimization algorithm is adopted to iteratively optimize the model super-parameter of the advertisement skip trigger time proportion prediction model, so as to obtain the current optimal model super-parameter, which includes but is not limited to the following steps S51-S58.
S51, initializing a population: the number of the wolves is set asThe iteration number is set to +.>Next, and initializing search ranges of at least two model super parameters of the advertisement skip trigger time scale prediction model, and then performing step S52, wherein,represents a positive integer of 5 or more, < >>Representing a positive integer of 100 or more.
In the step S51, the advertisement skip trigger time proportion prediction model is considered to be obtained based on a random forest algorithm, and the super parameters to be set by the random forest algorithm are two types, namely a decision tree number and a leaf node number, wherein the decision tree number is used for influencing the learning ability of the random forest algorithm, and the leaf node number is used for controlling the model to prevent over fitting. Thus in particular, the at least two modelsSuper parameters include, but are not limited to, the number of decision tree particles and the number of leaf nodes with the random forest algorithm. By way of example only, the process of the present invention,can be exemplified by 10, & gt>For example, 200, the search range of the decision tree number can be configured as [0,200 ]]A decision tree, the search range of the leaf node number can be configured as [0,20]And each leaf node.
S52, initializing the gray wolves: at the position ofThree wolves were randomly selected as the initial +. >Wolf and jersey>Wolf and->Wolf, and is initially set at said ++within the search range of said at least two model hyper-parameters>Individual position vectors of individual wolves of the individual wolves are then performed step S53, wherein the individual position vectors contain the search values of the at least two model hyper-parameters.
In the step S52, the individual position vector includes two values: the search value of the decision tree number and the search value of the leaf node number can be obtained from the random values in the corresponding search range during initialization.
S53, inputting the multidimensional advertisement characteristic value of the advertisement to be pushed into the advertisement skip trigger time proportion prediction model substituted into the corresponding current individual position vector according to the gray wolves, taking the absolute difference value of the actual advertisement skip trigger time proportion value and the corresponding output advertisement skip trigger time proportion prediction value as the corresponding individual fitness value, and then executing step S54.
S54, judging whether the current iteration number reachesStep S58 is executed if yes, otherwise the gray wolf with the minimum individual fitness is taken as new +.>Wolf, and gray wolf with a next highest individual fitness value as a new +. >Wolf and also wolf with a further high individual fitness value as new +.>Wolf, then step S55 is performed.
In the step S54, for example, if the fitness values of the first four bodies in the order of 0.01, 0.03, 0.06 and 0.08 are sequentially set, a wolf with 0.01 can be regarded as newWolf and the use of a wolf with 0.03 as a new +.>Wolf and the use of a wolf with 0.06 as a new +.>Wolf.
S55, respectively calculating convergence factorsSynergistic vector->And synergistic vector->Then execute the stepsStep S56, wherein the convergence factor +.>Said synergy vector->And said synergy vector->The calculation formulas of (a) are respectively as follows:
in the method, in the process of the invention,representing said current iteration number,/->Representing hyperbolic tangent function, ">And->Respectively represent [0,1 ]]Is a random vector of (c).
In the step S55, considering that the convergence factor of the conventional wolf algorithm has a pure linearity, in order to achieve the purposes of non-linearizing the convergence factor and facilitating the algorithm to achieve global optimization, the embodiment is affected by the tanh activation function image in the neural network (i.e. selecting the image with the function in the range of [ -3,3], and sequentially performing transformation operations such as expansion, symmetry, translation, etc.), and substituting the iteration times into the function, thereby improving the convergence factor expression.
S56, for eachWolf, according to said new ++>Wolf and jersey>Wolf and->The current individual position vector of wolf is calculated to obtain the corresponding and +.>Individual position vectors in the multiple iterations->Step S57 is then performed, wherein the individual position vector +.>The method is calculated according to the following formula:
in the method, in the process of the invention,representing said new->Current individual position vector of wolf, +.>Representing said new->Current individual position vector of wolf, +.>Representing said new->Current individual position vector of wolf, +.>Is indicated at +.>Individual position vectors in the multiple iterations, +.>、/>And->Respectively represent the synergy vector calculated randomly +.>，/>、/>And->Respectively represent the synergy vector calculated randomly +.>。
In the step S56, the present embodiment also performs weighting assignment on the gray wolf position update strategy, i.e. weight coefficients are calculated respectively、/>And->So as to form an improvement point of the gray wolf optimization algorithm together with the new convergence factor expression, and find good precision and convergence speed through performance under the commonly used 10 international standard test functions.
S57, adding 1 to the iteration number, and returning to the step S53.
S58, taking the current individual position vector of the gray wolf with the minimum individual fitness as the current optimal model super-parameter.
In said step S58, according to the principles of the Grey wolf algorithm, the currentThe current individual position vector of wolf is the current search value of the at least two model super parameters obtained by searching and can be used as the current optimal model super parameters. Based on the foregoing steps S51 to S58, the super parameter of the prediction model may be further adjusted by adopting an improved gray wolf algorithm according to the new man-machine interaction result, so that the advertisement skip trigger time proportion prediction model may more accurately predict the current advertisement interest preference of the user, so as to perform more accurate advertisement pushing when the next application is started.
S6, substituting the current optimal model super parameters into the advertisement skip trigger time proportion prediction model to obtain a new advertisement skip trigger time proportion prediction model.
S7, inputting the corresponding multidimensional advertisement characteristic values into the advertisement skip trigger time proportion prediction new model aiming at each advertisement to be pushed, and outputting to obtain the corresponding advertisement skip trigger time proportion prediction new value so as to select the pushing advertisement which is most suitable for loading when the next application is started (namely, executing the steps S4-S7 again).
According to the man-machine interaction advertising method based on big data, which is described in the steps S1-S7, a new scheme for accurately predicting advertisement skip trigger time proportion based on historical application starting big data and a random forest algorithm and applying push advertisements is provided, namely, firstly, big data modeling is started according to historical application of a target user to obtain advertisement skip trigger time proportion prediction models, then the models are applied to obtain advertisement skip trigger time proportion prediction values of all advertisements to be pushed, then when a certain application is detected to be started, a certain advertisement to be pushed with a prediction maximum value is loaded onto an application starting page of the certain application, an actual value of the advertisement skip trigger time proportion is determined, finally, iterative optimization is conducted on model parameters according to the actual value, and the proportion prediction values of all advertisements to be pushed are refreshed according to the optimizing result, so that advertisement personalized self-adaptive push can be conducted according to the man-machine interaction result of users, the push advertisement content is not immobilized any more, the accuracy of the pushing result of the final advertisement content and the advertising effect are ensured, and practical application and popularization are facilitated.
As shown in fig. 3, a second aspect of the present embodiment provides a virtual device for implementing the man-machine interaction advertising method according to the first aspect, where the virtual device includes a big data acquisition module, a model training module, a model application module, an advertisement loading module, a parameter optimizing module, and a model updating module;
the big data acquisition module is used for acquiring historical application starting big data of a target user, wherein the historical application starting big data comprises a multidimensional advertisement characteristic value and an advertisement skip trigger time proportion of a starting page pushing advertisement loaded and displayed in each application historical starting process, the starting page pushing advertisement is an advertisement loaded on an application starting page for pushing, the advertisement skip trigger time proportion is a ratio of a time period from a loading time stamp of the starting page pushing advertisement to a trigger time stamp of an advertisement skip event to a preset display time period of the starting page pushing advertisement, and the advertisement skip event is triggered on the application starting page by the target user in a man-machine interaction mode or is automatically triggered when the display time period of the starting page pushing advertisement reaches the preset display time period;
The model training module is in communication connection with the big data acquisition module and is used for starting big data according to the historical application, taking the multidimensional advertisement characteristic values of the advertisements pushed by all the starting pages as input items and taking the advertisement skip trigger time proportion of the advertisements pushed by all the starting pages as output items, and carrying out rated verification modeling on the artificial intelligent model based on a random forest algorithm to obtain an advertisement skip trigger time proportion prediction model;
the model application module is in communication connection with the model training module and is used for inputting the corresponding multidimensional advertisement characteristic values into the advertisement skipping trigger time proportion prediction model for each advertisement to be pushed, and outputting the corresponding advertisement skipping trigger time proportion prediction values;
the advertisement loading module is in communication connection with the model application module and is used for loading a certain advertisement to be pushed, which currently has the maximum value of the advertisement skipping trigger time proportion prediction, onto an application starting page of the certain application when the fact that the certain application is started is detected, and determining the actual value of the advertisement skipping trigger time proportion according to the newly triggered advertisement skipping event;
the parameter optimizing module is in communication connection with the advertisement loading module and is used for carrying out iterative optimization on the model super-parameters of the advertisement skip trigger time proportion prediction model by adopting a gray wolf optimization algorithm according to the actual value of the advertisement skip trigger time proportion and the multidimensional advertisement characteristic value of the certain advertisement to be pushed so as to obtain the current optimal model super-parameters;
The model updating module is respectively in communication connection with the model training module and the parameter optimizing module and is used for substituting the current optimal model super parameters into the advertisement skip trigger time proportion prediction model to obtain a new advertisement skip trigger time proportion prediction model;
the model application module is also in communication connection with the model updating module and is further used for inputting the corresponding multidimensional advertisement characteristic values into the advertisement skip trigger time proportion prediction new model aiming at each advertisement to be pushed, and outputting the corresponding advertisement skip trigger time proportion prediction new values so as to select the pushing advertisement which is most suitable for loading when the next application is started.
The working process, working details and technical effects of the foregoing device provided in the second aspect of the present embodiment may refer to the man-machine interaction advertising method described in the first aspect, which are not described herein again.
As shown in fig. 4, a third aspect of the present embodiment provides a computer device for executing the man-machine interaction advertisement method according to the first aspect, which includes a memory, a processor and a transceiver that are sequentially connected in communication, where the memory is used for storing a computer program, the transceiver is used for receiving and sending a message, and the processor is used for reading the computer program and executing the man-machine interaction advertisement method according to the first aspect. By way of specific example, the Memory may include, but is not limited to, random-Access Memory (RAM), read-Only Memory (ROM), flash Memory (Flash Memory), first-in first-out Memory (First Input First Output, FIFO), and/or first-in last-out Memory (First Input Last Output, FILO), etc.; the processor may be, but is not limited to, a microprocessor of the type STM32F105 family. In addition, the computer device may include, but is not limited to, a power module, a display screen, and other necessary components.
The working process, working details and technical effects of the foregoing computer device provided in the third aspect of the present embodiment may refer to the man-machine interaction advertisement method described in the first aspect, which are not described herein again.
A fourth aspect of the present embodiment provides a computer-readable storage medium storing instructions comprising the human-machine interaction advertising method according to the first aspect, i.e. the computer-readable storage medium has instructions stored thereon which, when executed on a computer, perform the human-machine interaction advertising method according to the first aspect. The computer readable storage medium refers to a carrier for storing data, and may include, but is not limited to, a floppy disk, an optical disk, a hard disk, a flash Memory, and/or a Memory Stick (Memory Stick), where the computer may be a general purpose computer, a special purpose computer, a computer network, or other programmable devices.
The working process, working details and technical effects of the foregoing computer readable storage medium provided in the fourth aspect of the present embodiment may refer to the man-machine interaction advertisement method as described in the first aspect, which are not described herein again.
A fifth aspect of the present embodiment provides a computer program product comprising instructions which, when run on a computer, cause the computer to perform the human-machine interaction advertising method of the first aspect. Wherein the computer may be a general purpose computer, a special purpose computer, a computer network, or other programmable apparatus.
Finally, it should be noted that: the foregoing description is only of the preferred embodiments of the invention and is not intended to limit the scope of the invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A man-machine interaction advertising method based on big data is characterized by comprising the following steps:
acquiring historical application starting big data of a target user, wherein the historical application starting big data comprises a multidimensional advertisement characteristic value and an advertisement skipping trigger time proportion of a starting page pushing advertisement loaded and displayed in each application historical starting process, the starting page pushing advertisement is an advertisement loaded on an application starting page for pushing, the advertisement skipping trigger time proportion is a ratio of a time length from a loading time stamp of the starting page pushing advertisement to a trigger time stamp of an advertisement skipping event to a preset display time length of the starting page pushing advertisement, the advertisement skipping event is triggered on the application starting page by the target user in a man-machine interaction mode, or the advertisement skipping event is automatically triggered when the display time length of the starting page pushing advertisement reaches the preset display time length;
According to the historical application starting big data, taking the multidimensional advertisement characteristic values of the advertisement pushed by each starting page as input items and taking the advertisement skip trigger time proportion of the advertisement pushed by each starting page as output items, and carrying out rating verification modeling on an artificial intelligent model based on a random forest algorithm to obtain an advertisement skip trigger time proportion prediction model;
inputting the corresponding multidimensional advertisement characteristic values into the advertisement skipping triggering time proportion prediction model for each advertisement to be pushed, and outputting to obtain the corresponding advertisement skipping triggering time proportion prediction values;
when detecting that a certain application is started, loading a certain advertisement to be pushed, which currently has the maximum value of the advertisement skipping trigger time proportion prediction, onto an application starting page of the certain application, and determining the actual value of the advertisement skipping trigger time proportion according to the newly triggered advertisement skipping event;
according to the actual value of the advertisement skip trigger time proportion and the multidimensional advertisement characteristic value of the certain advertisement to be pushed, carrying out iterative optimization on the model super-parameters of the advertisement skip trigger time proportion prediction model by adopting a gray wolf optimization algorithm to obtain the current optimal model super-parameters;
Substituting the current optimal model super parameters into the advertisement skip trigger time proportion prediction model to obtain a new advertisement skip trigger time proportion prediction model;
and inputting the corresponding multidimensional advertisement characteristic values into the advertisement skip trigger time proportion prediction new model aiming at each advertisement to be pushed, and outputting to obtain the corresponding advertisement skip trigger time proportion prediction new values so as to select the pushing advertisement which is most suitable for loading when the next application is started.
2. The human-machine interaction advertising method of claim 1, wherein the multi-dimensional advertising characteristic values comprise advertising type number values, target audience group type number values, and/or keyword number values.
3. The man-machine interaction advertising method according to claim 1, wherein starting big data according to the historical application, taking multidimensional advertisement feature values of each starting page push advertisement as input items comprises:
selecting a plurality of start page push advertisements in the history starting process of the last application according to the history application starting big data, wherein the plurality of start page push advertisements are in one-to-one correspondence with the history starting process of the last application;
And taking the multidimensional advertisement characteristic value of each start page pushing advertisement in the plurality of start page pushing advertisements as an input item.
4. The man-machine interaction advertisement method according to claim 3, wherein selecting a plurality of start page push advertisements in a last plurality of application history starts according to the history application start big data, comprises:
determining occurrence in the latest unit period based on the historical application initiation big dataA secondary application history start-up procedure, wherein +.>Represents a positive integer of 10 or more;
will be in contact with theThe secondary application history starting process is one-to-one corresponding to +.>The start page push advertisement acts as a plurality of start page push advertisements during the last multiple application history starts.
5. The method for man-machine interaction advertisement according to claim 1, wherein loading a certain advertisement to be pushed currently having a maximum value of advertisement skip trigger time proportion prediction onto an application start page of a certain application when it is detected that the certain application is started, comprises:
when detecting that a certain application is started, loading a certain advertisement to be pushed in a plurality of advertisements to be pushed and currently having an advertisement skip trigger time proportion prediction maximum value onto an application starting page of the certain application, wherein the advertisements to be pushed are updated and downloaded from a server of the certain application in advance.
6. The man-machine interaction advertisement method according to claim 1, wherein according to the actual value of the advertisement skip trigger time proportion and the multidimensional advertisement feature value of the advertisement to be pushed, iterative optimization is performed on the model super-parameters of the advertisement skip trigger time proportion prediction model by using a gray wolf optimization algorithm to obtain the current optimal model super-parameters, comprising the following steps of:
s51, initializing a population: the number of the wolves is set asThe iteration number is set to +.>Next, and initializing search ranges of at least two model super parameters of the advertisement skip trigger time scale prediction model, and then performing step S52, wherein +_>Represents a positive integer of 5 or more, < >>Represents a positive integer of 100 or more;
s52, initializing the gray wolves: at the position ofThree wolves were randomly selected as the initial +.>Wolf and jersey>Wolf and->Wolf, and is initially set at said ++within the search range of said at least two model hyper-parameters>Individual position vectors of individual wolves of the individual wolves, which individual position vectors contain the at least one of the at least one wolves, are then performed step S53Search values of the super parameters of the two models;
s53, inputting the multidimensional advertisement characteristic value of the certain advertisement to be pushed into the advertisement skip trigger time proportion prediction model substituted into the corresponding current individual position vector, taking the absolute difference value of the actual advertisement skip trigger time proportion value and the corresponding output advertisement skip trigger time proportion prediction value as the corresponding individual fitness value, and then executing step S54;
S54, judging whether the current iteration number reachesStep S58 is executed if yes, otherwise the gray wolf with the minimum individual fitness is taken as new +.>Wolf, and gray wolf with a next highest individual fitness value as a new +.>Wolf and also wolf with a further high individual fitness value as new +.>Wolf, then step S55 is performed;
s55, respectively calculating convergence factorsSynergistic vector->And synergistic vector->Step S56 is then performed, wherein the convergence factor +.>Said synergy vector->And said synergy vector->The calculation formulas of (a) are respectively as follows:
in the method, in the process of the invention,representing said current iteration number,/->Representing hyperbolic tangent function, ">And->Respectively represent [0,1 ]]Is a random vector of (a);
s56, for eachWolf, according to said new ++>Wolf and jersey>Wolf and->The current individual position vector of wolf is calculated to obtain the corresponding and +.>Individual position vectors in the multiple iterations->Step S57 is then performed, wherein the individual position vectorsThe method is calculated according to the following formula:
in the method, in the process of the invention,representing said new->Current individual position vector of wolf, +.>Representing said new->Current individual position vector of wolf, +.>Representing said new->Current individual position vector of wolf, +. >Is indicated at +.>Individual position vectors in the multiple iterations, +.>、/>And->Respectively represent the synergy vector calculated randomly +.>，/>、/>And->Respectively represent the synergy vector calculated randomly +.>；
S57, adding 1 to the iteration times, and then returning to execute the step S53;
s58, taking the current individual position vector of the gray wolf with the minimum individual fitness as the current optimal model super-parameter.
7. The human-machine interaction advertising method according to claim 6, wherein the at least two model hyper-parameters comprise a decision tree number and a leaf node number of the random forest algorithm.
8. The man-machine interaction advertising device based on big data is characterized by comprising a big data acquisition module, a model training module, a model application module, an advertisement loading module, a parameter optimizing module and a model updating module;
the big data acquisition module is used for acquiring historical application starting big data of a target user, wherein the historical application starting big data comprises a multidimensional advertisement characteristic value and an advertisement skip trigger time proportion of a starting page pushing advertisement loaded and displayed in each application historical starting process, the starting page pushing advertisement is an advertisement loaded on an application starting page for pushing, the advertisement skip trigger time proportion is a ratio of a time period from a loading time stamp of the starting page pushing advertisement to a trigger time stamp of an advertisement skip event to a preset display time period of the starting page pushing advertisement, and the advertisement skip event is triggered on the application starting page by the target user in a man-machine interaction mode or is automatically triggered when the display time period of the starting page pushing advertisement reaches the preset display time period;
The model training module is in communication connection with the big data acquisition module and is used for starting big data according to the historical application, taking the multidimensional advertisement characteristic values of the advertisements pushed by all the starting pages as input items and taking the advertisement skip trigger time proportion of the advertisements pushed by all the starting pages as output items, and carrying out rated verification modeling on the artificial intelligent model based on a random forest algorithm to obtain an advertisement skip trigger time proportion prediction model;
the model application module is in communication connection with the model training module and is used for inputting the corresponding multidimensional advertisement characteristic values into the advertisement skipping trigger time proportion prediction model for each advertisement to be pushed, and outputting the corresponding advertisement skipping trigger time proportion prediction values;
the advertisement loading module is in communication connection with the model application module and is used for loading a certain advertisement to be pushed, which currently has the maximum value of the advertisement skipping trigger time proportion prediction, onto an application starting page of the certain application when the fact that the certain application is started is detected, and determining the actual value of the advertisement skipping trigger time proportion according to the newly triggered advertisement skipping event;
the parameter optimizing module is in communication connection with the advertisement loading module and is used for carrying out iterative optimization on the model super-parameters of the advertisement skip trigger time proportion prediction model by adopting a gray wolf optimization algorithm according to the actual value of the advertisement skip trigger time proportion and the multidimensional advertisement characteristic value of the certain advertisement to be pushed so as to obtain the current optimal model super-parameters;
The model updating module is respectively in communication connection with the model training module and the parameter optimizing module and is used for substituting the current optimal model super parameters into the advertisement skip trigger time proportion prediction model to obtain a new advertisement skip trigger time proportion prediction model;
the model application module is also in communication connection with the model updating module and is further used for inputting the corresponding multidimensional advertisement characteristic values into the advertisement skip trigger time proportion prediction new model aiming at each advertisement to be pushed, and outputting the corresponding advertisement skip trigger time proportion prediction new values so as to select the pushing advertisement which is most suitable for loading when the next application is started.
9. A computer device comprising a memory, a processor and a transceiver in communication connection in sequence, wherein the memory is configured to store a computer program, the transceiver is configured to send and receive messages, and the processor is configured to read the computer program and perform the man-machine interaction advertising method according to any one of claims 1-7.
10. A computer readable storage medium having instructions stored thereon which, when executed on a computer, perform the human-machine interaction advertising method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410001394.6A CN117495459B (en) | 2024-01-02 | 2024-01-02 | Man-machine interaction advertisement method, device, equipment and storage medium based on big data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410001394.6A CN117495459B (en) | 2024-01-02 | 2024-01-02 | Man-machine interaction advertisement method, device, equipment and storage medium based on big data |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117495459A true CN117495459A (en) | 2024-02-02 |
| CN117495459B CN117495459B (en) | 2024-03-19 |
Family
ID=89683354
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410001394.6A Active CN117495459B (en) | 2024-01-02 | 2024-01-02 | Man-machine interaction advertisement method, device, equipment and storage medium based on big data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117495459B (en) |
Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120130805A1 (en) * | 2010-11-18 | 2012-05-24 | Google Inc. | Selecting media advertisements for presentation based on their predicted playtimes |
| US20130198013A1 (en) * | 2012-01-26 | 2013-08-01 | SPOTXCHANGE, INC. A Delaware Corporation | Consumer-initiated payment to skip electronic advertisements |
| US20140236708A1 (en) * | 2010-02-01 | 2014-08-21 | Nevallco, Llc | Methods and apparatus for a predictive advertising engine |
| WO2015178219A1 (en) * | 2014-05-22 | 2015-11-26 | ソニー株式会社 | Information-processing device, information processing method, and program |
| US20160191972A1 (en) * | 2014-12-31 | 2016-06-30 | Sanjeev Madhava Rao | Method and system for skipping optimum advertisements in content |
| US20170154356A1 (en) * | 2015-11-30 | 2017-06-01 | Yahoo! Inc. | Generating actionable suggestions for improving user engagement with online advertisements |
| CN107590689A (en) * | 2017-08-25 | 2018-01-16 | 北京奇艺世纪科技有限公司 | A kind of ad data recommends method and system |
| CN107613325A (en) * | 2017-09-01 | 2018-01-19 | 北京小米移动软件有限公司 | Advertisement delivery method and device |
| US20180165706A1 (en) * | 2016-12-13 | 2018-06-14 | Sprinklr, Inc. | Systems and methods for improving social media advertising efficiency |
| CN108965951A (en) * | 2018-07-12 | 2018-12-07 | 北京奇艺世纪科技有限公司 | The playback method and device of advertisement |
| CN112381290A (en) * | 2020-11-13 | 2021-02-19 | 辽宁工程技术大学 | Method for predicting coal gas content by random forest and gray wolf optimization |
| CN113869931A (en) * | 2021-08-26 | 2021-12-31 | 卓尔智联(武汉)研究院有限公司 | Advertisement putting strategy determining method and device, computer equipment and storage medium |
| CN115689642A (en) * | 2021-07-30 | 2023-02-03 | 腾讯科技（深圳）有限公司 | Media content recommendation method and device, electronic equipment and storage medium |
| CN117112381A (en) * | 2023-08-25 | 2023-11-24 | 上海泛微软件有限公司 | Form page opening duration prediction method, device, equipment and storage medium |
| CN117132330A (en) * | 2023-10-23 | 2023-11-28 | 蓝色火焰科技成都有限公司 | Intelligent advertisement method, device, equipment and storage medium based on big data |
-
2024
- 2024-01-02 CN CN202410001394.6A patent/CN117495459B/en active Active
Patent Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140236708A1 (en) * | 2010-02-01 | 2014-08-21 | Nevallco, Llc | Methods and apparatus for a predictive advertising engine |
| US20120130805A1 (en) * | 2010-11-18 | 2012-05-24 | Google Inc. | Selecting media advertisements for presentation based on their predicted playtimes |
| US20130198013A1 (en) * | 2012-01-26 | 2013-08-01 | SPOTXCHANGE, INC. A Delaware Corporation | Consumer-initiated payment to skip electronic advertisements |
| WO2015178219A1 (en) * | 2014-05-22 | 2015-11-26 | ソニー株式会社 | Information-processing device, information processing method, and program |
| US20160191972A1 (en) * | 2014-12-31 | 2016-06-30 | Sanjeev Madhava Rao | Method and system for skipping optimum advertisements in content |
| US20170154356A1 (en) * | 2015-11-30 | 2017-06-01 | Yahoo! Inc. | Generating actionable suggestions for improving user engagement with online advertisements |
| US20180165706A1 (en) * | 2016-12-13 | 2018-06-14 | Sprinklr, Inc. | Systems and methods for improving social media advertising efficiency |
| CN107590689A (en) * | 2017-08-25 | 2018-01-16 | 北京奇艺世纪科技有限公司 | A kind of ad data recommends method and system |
| CN107613325A (en) * | 2017-09-01 | 2018-01-19 | 北京小米移动软件有限公司 | Advertisement delivery method and device |
| CN108965951A (en) * | 2018-07-12 | 2018-12-07 | 北京奇艺世纪科技有限公司 | The playback method and device of advertisement |
| CN112381290A (en) * | 2020-11-13 | 2021-02-19 | 辽宁工程技术大学 | Method for predicting coal gas content by random forest and gray wolf optimization |
| CN115689642A (en) * | 2021-07-30 | 2023-02-03 | 腾讯科技（深圳）有限公司 | Media content recommendation method and device, electronic equipment and storage medium |
| CN113869931A (en) * | 2021-08-26 | 2021-12-31 | 卓尔智联(武汉)研究院有限公司 | Advertisement putting strategy determining method and device, computer equipment and storage medium |
| CN117112381A (en) * | 2023-08-25 | 2023-11-24 | 上海泛微软件有限公司 | Form page opening duration prediction method, device, equipment and storage medium |
| CN117132330A (en) * | 2023-10-23 | 2023-11-28 | 蓝色火焰科技成都有限公司 | Intelligent advertisement method, device, equipment and storage medium based on big data |
Non-Patent Citations (3)
| Title |
|---|
| SU JUNG KIM 等: "How mobile users respond to pre-roll skippable advertising differently than personal computer(PC) users: and analysis of individual-level clickstream data", INTERNATIONAL JOURNAL OF ADVERTISING, vol. 42, no. 4, 5 July 2022 (2022-07-05), pages 713 - 739 * |
| ZHABIZ GHARIBSHAH 等: "User response prediction in online advertising", ACM COMPUTING SURVEYS, vol. 54, no. 3, 8 May 2021 (2021-05-08), pages 1 - 43, XP058679022, DOI: 10.1145/3446662 * |
| 李元慧: "手机开屏广告的跳过设计效应及其影响因素的眼动研究", 中国优秀硕士学位论文全文数据库 哲学与人文科学辑, no. 04, 15 April 2023 (2023-04-15), pages 102 - 21 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117495459B (en) | 2024-03-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2020107806A1 (en) | Recommendation method and device | |
| US10509837B2 (en) | Modeling actions for entity-centric search | |
| CN103064945B (en) | Based on the Situational searching method of body | |
| EP3623956A1 (en) | Method and apparatus for obtaining recommendation information, and electronic device | |
| CN111784455A (en) | Article recommendation method and recommendation equipment | |
| WO2017197806A1 (en) | Method for providing intelligent service, intelligent service system and intelligent terminal based on artificial intelligence | |
| US20210271970A1 (en) | Neural network optimizer search | |
| US20230153857A1 (en) | Recommendation model training method, recommendation method, apparatus, and computer-readable medium | |
| US20100017391A1 (en) | Polarity estimation system, information delivery system, polarity estimation method, polarity estimation program and evaluation polarity estimatiom program | |
| CN108897823B (en) | Personalized commodity retrieval method and device based on deep learning attention mechanism | |
| CN108629608A (en) | User data processing method and processing device | |
| CN117132330B (en) | Intelligent advertisement method, device, equipment and storage medium based on big data | |
| US10552863B1 (en) | Machine learning approach for causal effect estimation | |
| CN110390052A (en) | Search for recommended method, the training method of CTR prediction model, device and equipment | |
| CN111930366B (en) | Rule engine implementation method and system based on JIT real-time compilation | |
| US20230342797A1 (en) | Object processing method based on time and value factors | |
| CN112435064A (en) | Method, device and equipment for evaluating recommendation information and computer readable storage medium | |
| CN111667024A (en) | Content pushing method and device, computer equipment and storage medium | |
| CN108154288A (en) | New user subsidizes computational methods, device, electronic equipment and the storage medium of strategy | |
| CN112100511B (en) | Preference degree data obtaining method and device and electronic equipment | |
| CN117495459B (en) | Man-machine interaction advertisement method, device, equipment and storage medium based on big data | |
| CN108764489B (en) | Model training method and device based on virtual sample | |
| CA3062986C (en) | Information pushing method and device based on search content | |
| CN113255231B (en) | Data processing method, device, equipment and storage medium | |
| CN111177657B (en) | Demand determining method, system, electronic device and storage medium |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |