CN113888238A - Advertisement click rate prediction method and device and computer equipment - Google Patents
Advertisement click rate prediction method and device and computer equipment Download PDFInfo
- Publication number
- CN113888238A CN113888238A CN202111243596.4A CN202111243596A CN113888238A CN 113888238 A CN113888238 A CN 113888238A CN 202111243596 A CN202111243596 A CN 202111243596A CN 113888238 A CN113888238 A CN 113888238A
- Authority
- CN
- China
- Prior art keywords
- user
- interest
- advertisement
- time
- data
- 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 46
- 239000013598 vector Substances 0.000 claims abstract description 55
- 238000012549 training Methods 0.000 claims abstract description 25
- 239000013604 expression vector Substances 0.000 claims abstract description 23
- 238000013528 artificial neural network Methods 0.000 claims abstract description 14
- 238000007781 pre-processing Methods 0.000 claims abstract description 8
- 230000000694 effects Effects 0.000 claims abstract description 6
- 230000006399 behavior Effects 0.000 claims description 98
- 230000006870 function Effects 0.000 claims description 25
- 230000008569 process Effects 0.000 claims description 14
- 230000003068 static effect Effects 0.000 claims description 12
- 238000000605 extraction Methods 0.000 claims description 11
- 238000012545 processing Methods 0.000 claims description 11
- 238000004590 computer program Methods 0.000 claims description 8
- 239000011159 matrix material Substances 0.000 claims description 6
- 230000004913 activation Effects 0.000 claims description 5
- 238000009499 grossing Methods 0.000 claims description 2
- 238000012163 sequencing technique Methods 0.000 claims description 2
- 230000007246 mechanism Effects 0.000 abstract description 8
- 239000010410 layer Substances 0.000 description 23
- 238000010586 diagram Methods 0.000 description 9
- 230000009471 action Effects 0.000 description 5
- 238000004088 simulation Methods 0.000 description 5
- 230000003993 interaction Effects 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000004891 communication Methods 0.000 description 2
- 230000006835 compression Effects 0.000 description 2
- 238000007906 compression Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 241000288105 Grus Species 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000006397 emotional response Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 239000004576 sand Substances 0.000 description 1
- 238000011524 similarity measure Methods 0.000 description 1
- 239000002356 single layer Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000001502 supplementing 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
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0242—Determining effectiveness of advertisements
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
- G06Q30/0255—Targeted advertisements based on user history
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Finance (AREA)
- Strategic Management (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Biophysics (AREA)
- General Business, Economics & Management (AREA)
- Economics (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Marketing (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention belongs to the field of E-commerce big data recommendation, and relates to an advertisement click rate prediction method, an advertisement click rate prediction device and computer equipment; the method comprises the steps of obtaining user behavior data, user portrait data and advertisement data of an e-commerce platform; preprocessing user behavior data to form a user behavior sequence; respectively encoding and representing the user behavior sequence, the user portrait data and the advertisement data to obtain embedded vectors of corresponding characteristics; extracting an interest expression vector of a user by adopting a deep neural network based on an attention mechanism; extracting an invisible relation vector between the user portrait characteristic and the advertisement characteristic by adopting a stack type automatic coding machine; inputting the interest expression vector and the invisible relation vector of the user into a multilayer perceptron to carry out joint training to obtain a prediction result of the advertisement click rate; the invention can effectively improve the click rate of the E-commerce platform advertisement and realize the effects of accurate marketing and recommendation.
Description
Technical Field
The invention belongs to the field of E-commerce big data recommendation, and particularly relates to an advertisement click rate prediction method and device based on user interest and time sequence behaviors, and computer equipment.
Background
With the development of information technology, many internet e-commerce platforms at home and abroad increasingly pay attention to the profit effect of the online advertising system, and pay attention to the realization of personalized and accurate marketing strategies. The Click Through Rate (CTR) of the advertisement is one of the most core indexes in the E-commerce platform system, and is of great importance in the fields of advertisement recommendation, web page search, sponsorship recommendation and the like. The accuracy of click rate prediction can not only affect the revenue of the e-commerce platform, but also affect the satisfaction of the user and the consumption experience.
In current e-commerce platforms, although marketers want to know the response of web visitors, it is almost impossible to quantify the emotional response to a web site and the impact of the web site on the company brand using current technology. However, click-through rates are readily available. Click-through rate measures the number of page visitors as a proportion of visitors to the page after the commercial advertisement has been clicked on and redirected to another page where they can purchase goods or learn more about a product or service. Generally, a higher click through rate indicates that the advertised item is more commercially valuable or that the marketing campaign is more appealing. Most E-commerce websites aim at adjusting the display of homepage commodity advertisements through click rate and making personalized recommendations.
At present, many domestic and foreign scholars carry out intensive research on the CTR model, and research results are mainly reflected in the following aspects: on one hand, with the development of deep learning technology, a deep CTR model gradually replaces an LR-based CTR model requiring artificial feature engineering and the like. On the other hand, some deep CTR models focus on feature compression and interaction. In addition, the extraction of the user behavior sequence features is also focused by the model. But the advertisement click rate at the present stage still has the following defects:
1. the timeliness of the user's historical behavior sequence. The traditional time sequence model ignores the influence of time intervals among sequence behaviors on user interest expression, and the traditional RNN can well capture the dependency relationship among the sequence relationships in a behavior sequence, but the user behaviors are not only sequence relationships, and the time intervals of the behaviors, the characteristics of the behaviors and the like contain more prior information which is important for representing the user interest.
2. Generalization and complexity of user interests. The interests of the users have diverse and variable trends, the preferences of the users in a certain period of time have concentration, each interest has an own evolution trend, and the interests of different types have little mutual influence.
3. The dimensionality of data features is high, and the amount of hidden information is large. Besides the user behavior sequence characteristics, the relationship between the contextual characteristics, the advertising characteristics and other characteristics of the e-commerce platform advertising data input characteristics also influences the accuracy of click rate estimation. The features have high dimensionality and large hidden information, and the relationship between the features is difficult to acquire.
Disclosure of Invention
Aiming at the defects, the invention provides a method and a device for predicting the click rate of an advertisement and computer equipment for solving the problem of predicting the click rate of the advertisement.
In a first aspect of the present invention, the present invention provides an advertisement click-through rate prediction method, including:
acquiring user behavior data, user portrait data and advertisement data of the e-commerce platform;
preprocessing user behavior data of the e-commerce platform and forming a user behavior sequence;
respectively encoding and representing the user behavior sequence, the user portrait data and the advertisement data to obtain embedded vectors of corresponding characteristics;
inputting user behavior sequence characteristics, and outputting an interest expression vector of a user by adopting a Time-GRU deep neural network based on a Time factor;
inputting an interest expression vector of a user, simulating an interest updating process by adopting an attention-based AT-GRU deep neural network, and outputting an interest updating vector of the user;
inputting user portrait characteristics and advertisement characteristics, and extracting an invisible relation vector between the user portrait characteristics and the advertisement characteristics by adopting a stack type automatic coding machine;
and inputting the interest expression vector of the user and the invisible relation vector between the user portrait characteristics and the advertisement characteristics into a multilayer perceptron for joint training to obtain a prediction result of the advertisement click rate.
In a second aspect of the present invention, the present invention also provides an advertisement click-through rate prediction apparatus, comprising:
the acquisition module is used for acquiring user behavior data, user portrait data and advertisement data of the e-commerce platform;
the processing module is used for preprocessing the user behavior data of the e-commerce platform and forming a user behavior sequence;
the embedding module is used for respectively encoding and representing the user behavior sequence, the user portrait data and the advertisement data to obtain an embedded vector of the corresponding characteristics;
the first feature extraction module is used for inputting user behavior sequence features and outputting an interest expression vector of a user by adopting a Time-GRU deep neural network based on a Time factor;
the second feature extraction module is used for inputting an interest expression vector of a user, simulating an interest updating process by adopting an attention-based AT-GRU deep neural network, and outputting an interest updating vector of the user;
the third characteristic extraction module is used for inputting the user portrait characteristics and the advertisement characteristics, and extracting a recessive relation vector between the user portrait characteristics and the advertisement characteristics by adopting a stack type automatic coding machine;
and the advertisement click rate prediction module is used for inputting the interest expression vector of the user and the invisible relation vector between the user portrait characteristics and the advertisement characteristics into the multilayer perceptron to carry out joint training to obtain a prediction result of the advertisement click rate.
In a third aspect of the invention, the invention also provides a computer device comprising a memory and a processor, the memory having stored therein a computer program which, when executed by the processor, causes the processor to carry out the steps of the method according to the first aspect of the invention.
The invention has the beneficial effects that:
according to the method, the user behavior data and the target advertisement data of the e-commerce platform are utilized, the interest of the user hidden behind the user time sequence behavior sequence is obtained by updating and modeling the interest updating process, and the advertisement click rate is predicted by combining the invisible association among other non-time sequence characteristics.
Drawings
FIG. 1 is a block diagram of an advertisement click-through rate prediction framework according to an embodiment of the present invention;
FIG. 2 is a flowchart illustrating a method for predicting advertisement click-through rate according to an embodiment of the present invention;
FIG. 3 is a diagram of the present invention for constructing a timing model to simulate user interest characteristics;
FIG. 4 is a diagram of an interest update model for constructing an attention mechanism according to the present invention;
FIG. 5 is a diagram of an unsupervised feature extraction method introduced by the present invention;
FIG. 6 is a diagram of a private advertisement hit rate predicting apparatus according to the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Fig. 1 is a frame diagram of an advertisement click rate prediction in an embodiment of the present invention, as shown in fig. 1, the prediction frame in the embodiment mainly includes four parts, first, user behavior data, user portrait data, and advertisement data are collected; processing the data to obtain a user behavior sequence, a user portrait characteristic and an advertisement characteristic; secondly, an interest updating model is constructed by using a user behavior sequence in a data feature processing mode, an interest expression vector of a user is obtained by using an interest simulation model, and meanwhile, user portrait data and advertisement data are processed by using a feature interaction model, and a recessive relation vector is extracted; then, a click rate prediction model is constructed by combining the final interest representation of the user and the non-time sequence characteristic invisible relation; the click rate prediction model can be used for predicting the click rate of the advertisement and finishing accurate advertisement pushing according to the click rate of the advertisement.
Fig. 2 is a flowchart of an advertisement click-through rate prediction method according to an embodiment of the present invention, as shown in fig. 2, the method includes:
s1, acquiring user behavior data, user portrait data and advertisement data of the E-commerce platform;
in the embodiment of the invention, some basic data of the commercial platform can be obtained, including user historical behavior data, user portrait data and advertisement data. Taking the online advertisement display data set of the Taobao e-commerce platform as an example, the data set mainly records browsing/clicking records of Taobao users on the e-commerce platform, and comprises three parts, namely user behavior history, user portraits and advertisement basic information. The user behavior history data comprises fields such as user ID, advertisement ID, time, click and the like, is time sequence characteristic data reflecting the implicit interest of the user, and a user behavior sequence can be obtained by processing the user behavior data. The user profile data includes characteristic information such as a user ID, age, sex, and shopping depth, which reflects the basic characteristic information of the user. The advertisement basic data comprises characteristic information such as advertisement ID, commodity category ID, commodity brand ID and price, and the advertisement data shows some basic characteristics of the advertisement to be recommended and is an important component of non-time sequence data.
In the embodiment of the present invention, as for the method for acquiring data, the original data may be obtained from a data source provided by an e-commerce platform or by directly downloading an existing public data source, which is not limited in the present invention.
The raw data that is typically acquired is unstructured and cannot be used directly for data analysis. Most unstructured data can be structured by simple data cleansing. For example, duplicate data may be deleted, invalid nodes may be cleaned up, such as portions of guest data, etc.
S2, preprocessing the user behavior data of the E-commerce platform and forming a user behavior sequence;
in the embodiment of the invention, invalid user behavior data can be removed; for example, some user behaviors with too short a browsing time may affect the effectiveness of the data, and the browsing threshold for the effectiveness of the advertisement click is defined to be 25 seconds. When the user browsing time on an advertisement page is higher than the threshold value, the data is considered to be valid, otherwise, the data is invalid; these invalid data are deleted.
In the embodiment of the invention, data statistics can be carried out according to the number of users, and original user behavior data can be spliced according to the ID, the advertisement browsing and the timestamp information of each user; in this way a series of user behaviour data is formed.
In the embodiment of the invention, the counted user behavior data can be supplemented by adopting a multiple interpolation method; for example, when the missing rate exceeds 15%, the data is directly removed, and the data value not exceeding 15% is supplemented, wherein the supplementing mode can be set by a person skilled in the art according to the actual requirement.
In the embodiment of the invention, a user behavior sequence based on time difference is constructed; grouping the user behavior data according to the user ID, and sequencing the user behavior data in time sequence to form a user behavior sequence; for each sequence of behaviors therein, the difference between the timestamp of the next behavior and the timestamp of the current behavior is used as a time factor characteristic.
S3, respectively encoding and representing the user behavior sequence, the user portrait data and the advertisement data to obtain embedded vectors of corresponding characteristics;
in the embodiment of the invention, the user behavior sequence, the advertisement data and the user portrait data which are obtained by processing are subjected to one-hot coding, and then the data characteristics can be respectively subjected to normalization processing. And then, converting the input high-dimensional sparse feature vector into a low-dimensional dense vector by using a feature embedding method to obtain the embedded representation of feature data, namely outputting low-dimensional dense user behavior sequence features, advertisement features and user portrait features.
In the preferred embodiment of the present invention, the user behavior sequence feature adopts time sequence modeling, that is, it is assumed that the user behavior sequence feature is a user behavior-time sequence feature binary set U (B, Δ T), specifically expressed as;
U(B,ΔT)＝{(b1,δt1),(b2,δt2),…,(bn,δtn)}
the user behavior-time series characteristic binary set U (B, delta T) is defined as a binary set formed by B and delta T and is represented as:
B＝{b1,b2,...,bn}
ΔT＝{δt1,δt2,...,δtn|δt1＝0，δti＝time(bi)-time(bi-1)i>1}
b represents a historical behavior sequence feature set of a user, delta T represents the time difference corresponding to two adjacent user behaviors in B, and delta T represents the time difference corresponding to two adjacent user behaviors in B 10 means that the time difference for the first sequence set is 0.
S4, inputting user behavior sequence characteristics, and outputting an interest expression vector of a user by adopting a Time-GRU deep neural network based on a Time factor;
in the embodiment of the present invention, a Time-Gate recovery Unit (Time-GRU for short) based on a Time factor adopts a mode of combining the Time factor with the GRU; the process of simulating the interest expression vector of the user mainly adopts a time gating circulation unit to learn the static state of the user according to the user behavior sequenceSet of interesting groups states, denoted Intress＝Time-GRU(eu)＝{h'1,h'2,...,h'n}。
Wherein the static interest group state set IntressThe method is defined as a hidden state set of each moment output after a user behavior-time sequence feature binary set is subjected to feature embedding processing and time sequence modeling, wherein euRepresenting the embedded representation of the user behavior-time series characteristic binary set, wherein each hidden state in the set reflects the user interest extracted from the user behavior sequence at the moment; Time-GRU (e)u) Representation pair embedding vector euAdopting a time-gated circulation unit to obtain a result; h'nRepresenting the nth hidden interest state.
FIG. 3 is a diagram of modeling user interest characteristics by constructing a Time sequence model according to an embodiment of the present invention, and as shown in FIG. 3, the structure represents a Time-GRU kernel structure in the Time sequence model, and the structure combines a hidden interest state h of a last behavior sequencet-1With the input user's behavior characteristics i in the current behavior sequencetAnd a time factor Δ t for outputting a user interest h represented by a next hidden interest state through the actions of the time gate, the update gate and the reset gatet. The time gating cycle unit strengthens the influence of time factors in a user behavior sequence on user interest, and each intermediate hidden interest state highlights the user static interest which embodies the short-term and high-frequency behaviors of the user at a certain moment.
In a preferred embodiment of the present invention, the step S4 specifically includes the following steps:
s41, calculating time gate weight according to the input time interval and input characteristics, wherein the specific formula is as follows, Tg＝σ(Wt[Δt,it]): where Δ t is a time factor, i.e., the difference between the timestamp of the current activity and the timestamp of the last activity. Considering that the decay of the preference behaviors such as browsing and clicking of the user along with the time in the serialization model accords with the long tail distribution, the invention adds the logarithm processing to the time factor delta T and improves the time gate weight TtThe formula of (1) is as follows:
Tt＝σ(Wt[log(Δt+ζ),it])
the time gate weight T is smaller when the input time factor Δ T is smallergThe smaller, logarithmically processed time gate weight TtThe smaller the size, the less information the current step retains and the more information the previous step retains. Namely, the time interval between two adjacent behaviors of the user is shorter, and the dependency relationship between the two behaviors is higher;
s42, respectively updating the reset gate r according to the input characteristics, the time interval and the state of the last statetUpdate gate ztAnd intermediate hidden interest states
The specific formula is as follows:
zt＝σ(Wzit+Uzht-1+bz)，
rt＝σ(Writ+Urht-1+br)，
where σ is the sigmoid activation function, which is element-by-element multiplication, wz,wr,
Uz,Ur,Uh∈nH×nH，nHIs the size of the hidden layer, nIIs the size of the input layer. i.e. itAn input vector representing Time-GRU,
temporary state representing the t-th hidden interest state, ztIs the update gate, rtIs a reset gate. z is a radical oftAnd rtThe value range is 0 to 1 under the mapping action of the sigmoid function;
s43, adding the time gate weight into the updating strategy of the updating gate, wherein the specific formula is as follows:
in the embodiment of the invention, a time factor is added in the gate structure as an input, and a time weight is introduced in the gate structure as an auxiliary participation in an updating strategy of the updating gate, so that the time factor can be used as an important factor to participate in the interest simulation process.
S5, inputting an interest expression vector of a user, simulating an interest updating process by adopting an attention-based AT-GRU deep neural network, and outputting an interest updating vector of the user;
in the embodiment of the present invention, an Attention-based gated-loop Unit (AT-GRU for short) combines an Attention-based mechanism with a GRU, and the network will be described in detail below.
FIG. 4 is a diagram of an interest update model for constructing an attention mechanism according to the present invention, as shown in FIG. 4, the structure is a core structure of a gated loop unit AT-GRU based on the attention mechanism, and the input part of the structure is a core structure representing the user static interest h simulated by a timing model in the current step ttThe attention score a associated with this interest and targeted advertisingtThe improved updating gate is used for simulating the user interest to update along the process related to the target advertisement, and the final unit outputs the simulated final interest state. In an embodiment of the present invention, the step S5 may include the following steps:
s51, calculating a set of attention scores of each interest state and the target advertisement according to the static interest group state set, wherein the set is expressed as Atns ═ { alpha ═ alphai|i＝1,2,...T}；
In the examples of the present invention, αiRepresents the ith attention score, T represents the number of states of interest; alpha is alphaiIs defined as each state of interest h 'in SI'iA weighting parameter distribution mechanism with the target advertisement qThe calculated similarity measure can capture important features in the model.
Wherein, s (h'iQ) represents an interest state h'iSimilarity weight after similarity function calculation with the target advertisement q through a bilinear model,
s52, learning an interest final updating state represented as H-AT-GRU (Intres) by adopting a gating cycle unit based on an attention mechanism according to the static interest group state set and the attention score sets，Atns)＝{h i1,2.. T }, wherein T is the size of a hidden layer in the AT-GRU;
considering the dynamics and generalization of user interests, the interest update final state is defined as the static interest group state IntressAnd under the attention-based mechanism updating strategy, extracting a final interest expression vector through the interest updating model.
In a preferred embodiment of the present invention, the step S52 specifically includes the following steps:
s521, according to the formula of the step S51, the relevance weight of each interest and the candidate advertisement, namely the attention score a, can be obtainediThe method comprises the following steps:
wherein e isadIs a concatenation of embedded vectors from different categories of advertisement fields,
is a parameter matrix, nHIs the dimension of the hidden interest state vector, nAIs the dimension of the embedded vector of the advertisement. Attention score aiReflects the target advertisement and the input interest statusThe relevance between states, the more relevant the interest state is to the targeted advertisement, the greater the attention score.
S522, according to the attention score calculated in the above steps, introducing an attention-based gating loop unit AT-GRU, where the structure can determine the update strength of the hidden interest state according to the magnitude of the attention score, that is, the interest state related to the target advertisement can participate in the update process of the final interest state with a greater strength, and the interest unrelated to the target advertisement can participate in the update process with a smaller or even no strength, and the specific update strategy is as follows:
r’t＝σ(Wrii't+Urih't-1+bri)，
wherein h't、h't-1And
are all hidden states of AT-GRU, wri，
Uri,Uhi∈nHi×nHi，nHiIs the size of the AT-GRU hidden layer, nIiIs the size of the AT-GRU input layer. i'tAn input vector representing AT-GRU, i.e., the user's static interest learned through a time-gated cyclic unit, atAttention scores, and compared with an original GRU structure, the AT-GRU structure replaces an original update gate with the attention scores, can effectively avoid the interest drift problem caused by the generalization and the isomerism of user interest, realizes the process of simulating interest development and update from the continuously changing and developing interest of the user, and promotes the final interest to follow the target wide rangeAnd (5) updating the related direction.
S6, inputting user portrait characteristics and advertisement characteristics, and extracting an invisible relation vector between the user portrait characteristics and the advertisement characteristics by adopting a stack type automatic coding machine;
in the embodiment of the invention, according to the user portrait characteristic data and the target advertisement characteristic data, the implicit relation between the characteristics is calculated and is expressed as Rimplicit＝SAE(I,P)。
Because a recessive relation also exists between non-time sequence characteristics such as target advertisements and user portraits, the embodiment of the invention defines RimplicitThe method is an implicit relation among the features extracted after feature compression by a stack type coding machine, wherein I is a target advertisement feature embedding vector set, and P is a user portrait embedding vector set.
In the embodiment of the present invention, a stacked automatic coding machine structure is designed to obtain further relationships of other non-time-series characteristics, where the structure of a single-layer automatic coding machine is shown in fig. 5, and the structure is composed of an encoder and a decoder, and is divided into three parts, i.e., an input layer, a hidden layer, and an output layer. The encoder layer converts the features of the input layer into the features of the hidden layer under the action of the encoder function and then converts the hidden layer features into the output layer under the action of the decoder function. The stack type automatic coding machine initializes the parameters of the deep network through the pre-training of the layer-by-layer unsupervised learning, and can learn the high-dimensional nonlinear element interaction relationship between the non-time sequence characteristics by using the training parameters after the pre-training is finished.
In a preferred embodiment of the present invention, the step S6 may specifically include the following steps:
s61, calculating an encoding layer according to the input user portrait characteristics and the target advertisement characteristics, wherein the encoding layer is responsible for converting input data X of the input layer into a hidden layer state H, and the specific formula is as follows:
Z＝sigmoid(W1X+b1)，
wherein, W1Is a weight matrix, b1Is the first training offset and the second training offset,
is an activation function;
s62, similarly, performing a decoding layer calculation, where the decoding layer converts the state H of the hidden layer into an output layer Y, defined as:
Y＝sigmoid(W2Z+b2)，
wherein, W2Is a weight matrix, b2Is the second training bias.
S63, calculating the reconstruction error so that the error between the output Y and the original X is small enough, the specific formula is,
wherein W is W1And W2The combination of (1) and (b), where λ is a regularization coefficient, a penalty factor λ may be added to control the magnitude of the weights to prevent overfitting.
And S64, repeating the training process of the steps S61-S63, obtaining the training parameters of the whole stack type automatic coding machine in a layer-by-layer superposition training mode, and then learning the implicit relation between the non-time sequence characteristics according to the parameters.
And S7, inputting the interest expression vector of the user and the invisible relation vector between the user portrait characteristics and the advertisement characteristics into a multilayer perceptron to carry out joint training to obtain a prediction result of the advertisement click rate.
In the embodiment of the invention, the interest updating vector of the user and the invisible relation vector are connected, and the connected vectors are subjected to smoothing treatment; and respectively carrying out combined training on an auxiliary loss function of a time gating circulation unit and a predicted target loss function of the multilayer perceptron in the interest simulation and update model, and obtaining a prediction result of the advertisement click rate after the training is finished.
When the prediction model is trained, joint training is carried out on the Time-GRU part and the target loss function of the prediction model respectively in a joint training mode, and the global loss function of the model is expressed as follows:
L＝Ltarget+λ*Laux
where λ is the hyper-parameter, the simulation of user equilibrium interest and the prediction of advertisement click-through rate, LauxIndicating the loss of assistance for Time-GRU. In the embodiment of the invention, the target loss function L is subjected totargetWhen improving the loss function of the MLP, setting the MLP loss function as weighted mean square error, and setting the coefficient of the loss function according to the proportion of positive and negative samples in the existing data, wherein the improved target loss function is expressed as:
wherein L istargetThe target loss function of the improved multilayer perceptron is obtained; n1 denotes the number of positive samples; n2 denotes the number of negative samples; y is an indicator variable, 1 if the class is the same as the class of the sample, and 0 otherwise; p (Y ═ 0| X) and p (Y ═ 1| X) are different prediction probabilities that the network output belongs to the label, respectively. Meanwhile, the invention also introduces auxiliary loss LauxThe auxiliary loss is expressed as:
wherein,
the t-th embedded vector representing the user's single click, G being the entire set of items;
representing the embedding of samples outside the item clicked by the user i at the t step;
is a sigmoid activation function that is,
indicating the t-th hidden interest state of the user i in the Time-GRU. Loss of assistance uses next positive and negative click sample behavior to supervise learning of the current state of interest. The design of auxiliary loss introduces the feedback information of the whole network behavior of the user, and simultaneously, the click deviation among multiple scenes is not introduced and the multi-scene coupling is not caused; the overhead penalty from an optimization perspective can reduce the difficulty of gradient back propagation in long sequence modeling of GRUs. Under the action of auxiliary loss, the Time-GRU can correlate the hidden interest state output by each unit with the next click behavior, so that the interest of a user can be better simulated according to a user behavior sequence; meanwhile, under the updating strategy of the Time gate, the Time-GRU strengthens the influence of the Time interval in the user behavior sequence on the interest simulation, so that the influence of behaviors with shorter Time interval and more frequent clicks on the user interest is larger.
Fig. 6 is a block diagram of an advertisement click-through rate prediction apparatus according to an embodiment of the present invention, as shown in fig. 6, the apparatus includes:
201. the acquisition module is used for acquiring user behavior data, user portrait data and advertisement data of the e-commerce platform;
202. the processing module is used for preprocessing the user behavior data of the e-commerce platform and forming a user behavior sequence;
203. the embedding module is used for respectively encoding and representing the user behavior sequence, the user portrait data and the advertisement data to obtain an embedded vector of the corresponding characteristics;
204. the first feature extraction module is used for inputting user behavior sequence features and outputting an interest expression vector of a user by adopting a Time-GRU deep neural network based on a Time factor;
205. the second feature extraction module is used for inputting an interest expression vector of a user, simulating an interest updating process by adopting an attention-based AT-GRU deep neural network, and outputting an interest updating vector of the user;
206. the third characteristic extraction module is used for inputting the user portrait characteristics and the advertisement characteristics, and extracting a recessive relation vector between the user portrait characteristics and the advertisement characteristics by adopting a stack type automatic coding machine;
207. and the advertisement click rate prediction module is used for inputting the interest expression vector of the user and the implicit relation vector between the user portrait characteristics and the advertisement characteristics into the multilayer perceptron to carry out joint training to obtain a prediction result of the advertisement click rate.
In a preferred embodiment of the present invention, a computer device of the present invention may comprise a memory having stored therein a computer program, which, when executed by the processor, causes the processor to perform the steps of the advertisement click rate prediction method according to the present invention. The computer device may be a terminal or a server. The computer device includes a processor, a memory, and a network interface connected by a system bus. Wherein the memory includes a non-volatile storage medium and an internal memory. The non-volatile storage medium of the computer device may store an operating system and a computer program. The computer program, when executed, may cause a processor to perform a method of advertisement click rate prediction. The processor of the computer device is used for providing calculation and control capability and supporting the operation of the whole computer device. The internal memory may have stored therein a computer program that, when executed by the processor, causes the processor to perform a preference prediction method. The network interface of the computer device is used for network communication.
In one embodiment, the advertisement click-through rate prediction apparatus provided by the present application may be implemented in the form of a computer program that is executable on the computer device, and the non-volatile storage medium of the computer device may store the program modules constituting the advertisement click-through rate prediction apparatus. For example, the acquiring module, the processing module, the embedding module, the first feature extracting module, the second feature extracting module and the advertisement click-through rate predicting module shown in fig. 6. The computer program comprised of the respective program modules is for causing the computer device to execute the steps of the advertisement click rate prediction method of the embodiments of the present application described in the present specification.
In the description of the present invention, it is to be understood that the terms "coaxial", "bottom", "one end", "top", "middle", "other end", "upper", "one side", "top", "inner", "outer", "front", "center", "both ends", and the like, indicate orientations or positional relationships based on those shown in the drawings, and are only for convenience of description and simplicity of description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and thus, are not to be construed as limiting the present invention.
In the present invention, unless otherwise expressly stated or limited, the terms "mounted," "disposed," "connected," "fixed," "rotated," and the like are to be construed broadly, e.g., as meaning fixedly connected, detachably connected, or integrally formed; can be mechanically or electrically connected; the terms may be directly connected or indirectly connected through an intermediate, and may be communication between two elements or interaction relationship between two elements, unless otherwise specifically limited, and the specific meaning of the terms in the present invention will be understood by those skilled in the art according to specific situations.
Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (10)
1. The method for predicting the advertisement click rate is characterized by comprising the following steps of:
acquiring user behavior data, user portrait data and advertisement data of the e-commerce platform;
preprocessing user behavior data of the e-commerce platform and forming a user behavior sequence;
respectively encoding and representing the user behavior sequence, the user portrait data and the advertisement data to obtain embedded vectors of corresponding characteristics;
inputting user behavior sequence characteristics, and outputting an interest expression vector of a user by adopting a Time-GRU deep neural network based on a Time factor;
inputting an interest expression vector of a user, simulating an interest updating process by adopting an attention-based AT-GRU deep neural network, and outputting an interest updating vector of the user;
inputting user portrait characteristics and advertisement characteristics, and extracting an invisible relation vector between the user portrait characteristics and the advertisement characteristics by adopting a stack type automatic coding machine;
and respectively inputting the interest updating vector of the user and the invisible relation vector between the portrait characteristic and the advertisement characteristic of the user into a multilayer perceptron to carry out joint training to obtain a prediction result of the advertisement click rate.
2. The method of claim 1, wherein the pre-processing of the user behavior data of the e-commerce platform comprises removing invalid user behavior data; counting data according to the number of users, and splicing original user behavior data according to the ID, the advertisement browsed and the timestamp information of each user; completing the counted user behavior data by adopting a multiple interpolation method; constructing a user behavior sequence based on the time difference; grouping the user behavior data according to the user ID, and sequencing the user behavior data in time sequence to form a user behavior sequence; for each sequence of behaviors therein, the difference between the timestamp of the next behavior and the timestamp of the current behavior is used as the time.
3. The method of claim 1, wherein outputting the interest expression vector of the user comprises learning a set of static interest group states of the user using a Time-GRU (Time-gated loop unit) based on a Time factor according to the user behavior sequence; namely, calculating the weight of the time gate according to the input user behavior sequence characteristics; adding the time gate weight to a first update policy for updating a gate; the set of static interest group states is selected by time-gating reset and update gates in the loop unit.
4. The method of claim 3, wherein the first update strategy is expressed as
Wherein h istRepresenting the t-th hidden interest state in the Time-GRU; z is a radical oftIs the update gate in the Time-GRU;
is element-by-element multiplication, TtRepresents the time gate weight, ht-1Representing the t-1 hidden interest state in the Time-GRU; t ist＝σ(Wt[log(Δt+ζ),it]) (ii) a σ is a sigmoid activation function, WtA tth implicit state matrix representing a time gate; Δ t represents a time factor, i.e., the difference between the timestamp of the current activity and the timestamp of the last activity; ζ represents time gate bias; i.e. itAn input vector representing Time-GRU;
represents the temporary state of the tth hidden interest state in the Time-GRU.
5. The method of claim 3 or 4, wherein outputting the interest update vector comprises calculating an attention score of each interest state and the target advertisement according to the static interest group state set; and calculating an interest final updating state by adopting an attention-based AT-GRU (AT-ground-group unit) based on the attention system, namely a gating cycle unit based on the attention system according to the static interest group state set and the attention score, namely taking the attention score as an updating gate, adopting the size of the attention score as a second updating strategy of the updating gate, and selecting the interest final updating state through the updating gate and a reset gate in the time gating cycle unit.
6. The method of claim 5, wherein the second update strategy is expressed as
Wherein, h'tRepresenting the tth hidden interest state of the AT-GRU; a istIndicating the attention score;
are multiplied element by element, h't-1Representing t-1 hidden interest state of AT-GRU;
a temporary state representing the tth hidden interest state in the AT-GRU.
7. The method according to claim 1, wherein the interest update vector of the user, the invisible relationship vector between the user portrait feature and the advertisement feature are respectively input into a multi-layer perceptron for joint training, and obtaining the prediction result of the advertisement click rate comprises connecting the interest update vector of the user and the invisible relationship vector, and smoothing the connected vectors; respectively carrying out combined training on a local loss function of the deep neural network part of the Time-GRU based on the Time factor and a global loss function of the multilayer perceptron, and obtaining a prediction result of the advertisement click rate after the training is finished; wherein,
the global loss function is represented as:
L＝Ltarget+λ*Laux
wherein L represents a global loss function of the multi-layer perceptron; λ is a hyperparameter, LauxAn auxiliary loss function representing a time-gated loop unit;
n representsThe number of users;
the t-th embedding vector representing the single click of the user i, and G is the whole behavior sequence embedding set;
an embedded vector representing samples outside the item clicked on by user i at step t; sigma denotes a sigmoid activation function,
representing the t-th hidden interest state of the user i in the Time-GRU; l istargetRepresenting an objective loss function improved according to the positive and negative sample ratios;
n1 denotes the number of positive samples; n2 denotes the number of negative samples; y is an indicator variable, 1 if the class of the sample X is the same as the class of the sample Y, and 0 otherwise; p represents the predicted probability that the multi-layer perceptron network output belongs to a tag.
8. The method of claim 1 or 7, wherein the advertisement click-through rate is predicted as follows:
y＝sigmoid(WL(...(Wc(Rc)+bc)...)+bL)
wherein R iscA connection vector representing a vector for updating the user's interest and a vector for the implicit relationship; wLIs a first weight parameter matrix; wcIs a second weight parameter matrix, bcRepresenting a first training bias; bLIs the second training bias.
9. An advertisement click-through rate prediction apparatus, comprising:
the acquisition module is used for acquiring user behavior data, user portrait data and advertisement data of the e-commerce platform;
the processing module is used for preprocessing the user behavior data of the e-commerce platform and forming a user behavior sequence;
the embedding module is used for respectively encoding and representing the user behavior sequence, the user portrait data and the advertisement data to obtain an embedded vector of the corresponding characteristics;
the first feature extraction module is used for inputting user behavior sequence features and outputting an interest expression vector of a user by adopting a Time-GRU deep neural network based on a Time factor;
the second feature extraction module is used for inputting an interest expression vector of a user, simulating an interest updating process by adopting an attention-based AT-GRU deep neural network, and outputting an interest updating vector of the user;
the third characteristic extraction module is used for inputting the user portrait characteristics and the advertisement characteristics, and extracting a recessive relation vector between the user portrait characteristics and the advertisement characteristics by adopting a stack type automatic coding machine;
and the advertisement click rate prediction module is used for inputting the interest expression vector of the user and the invisible relation vector between the user portrait characteristics and the advertisement characteristics into the multilayer perceptron to carry out joint training to obtain a prediction result of the advertisement click rate.
10. A computer arrangement comprising a memory and a processor, the memory having stored therein a computer program which, when executed by the processor, causes the processor to carry out the steps of the method of any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111243596.4A CN113888238B (en) | 2021-10-25 | 2021-10-25 | Advertisement click rate prediction method and device and computer equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111243596.4A CN113888238B (en) | 2021-10-25 | 2021-10-25 | Advertisement click rate prediction method and device and computer equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113888238A true CN113888238A (en) | 2022-01-04 |
| CN113888238B CN113888238B (en) | 2024-04-09 |
Family
ID=79014053
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111243596.4A Active CN113888238B (en) | 2021-10-25 | 2021-10-25 | Advertisement click rate prediction method and device and computer equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113888238B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114565436A (en) * | 2022-02-28 | 2022-05-31 | 重庆长安汽车股份有限公司 | Vehicle model recommendation system, method, device and storage medium based on time sequence modeling |
| WO2024140829A1 (en) * | 2022-12-27 | 2024-07-04 | 华为技术有限公司 | Data processing method and apparatus, computing device, system and readable storage medium |
| CN114565436B (en) * | 2022-02-28 | 2024-10-29 | 重庆长安汽车股份有限公司 | Vehicle model recommendation system, method, equipment and storage medium based on time sequence modeling |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109960759A (en) * | 2019-03-22 | 2019-07-02 | 中山大学 | Recommender system clicking rate prediction technique based on deep neural network |
| CN112307257A (en) * | 2020-11-25 | 2021-02-02 | 中国计量大学 | Short video click rate prediction method based on multi-information node graph network |
| CN112381581A (en) * | 2020-11-17 | 2021-02-19 | 东华理工大学 | Advertisement click rate estimation method based on improved Transformer |
| CN112395505A (en) * | 2020-12-01 | 2021-02-23 | 中国计量大学 | Short video click rate prediction method based on cooperative attention mechanism |
| CN112700274A (en) * | 2020-12-29 | 2021-04-23 | 华南理工大学 | Advertisement click rate estimation method based on user preference |
-
2021
- 2021-10-25 CN CN202111243596.4A patent/CN113888238B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109960759A (en) * | 2019-03-22 | 2019-07-02 | 中山大学 | Recommender system clicking rate prediction technique based on deep neural network |
| CN112381581A (en) * | 2020-11-17 | 2021-02-19 | 东华理工大学 | Advertisement click rate estimation method based on improved Transformer |
| CN112307257A (en) * | 2020-11-25 | 2021-02-02 | 中国计量大学 | Short video click rate prediction method based on multi-information node graph network |
| CN112395505A (en) * | 2020-12-01 | 2021-02-23 | 中国计量大学 | Short video click rate prediction method based on cooperative attention mechanism |
| CN112700274A (en) * | 2020-12-29 | 2021-04-23 | 华南理工大学 | Advertisement click rate estimation method based on user preference |
Non-Patent Citations (1)
| Title |
|---|
| 朱江湖: "新型电商平台广告点击率算法研究与实现", 《》, 27 May 2022 (2022-05-27) * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114565436A (en) * | 2022-02-28 | 2022-05-31 | 重庆长安汽车股份有限公司 | Vehicle model recommendation system, method, device and storage medium based on time sequence modeling |
| CN114565436B (en) * | 2022-02-28 | 2024-10-29 | 重庆长安汽车股份有限公司 | Vehicle model recommendation system, method, equipment and storage medium based on time sequence modeling |
| WO2024140829A1 (en) * | 2022-12-27 | 2024-07-04 | 华为技术有限公司 | Data processing method and apparatus, computing device, system and readable storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113888238B (en) | 2024-04-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111222332B (en) | Commodity recommendation method combining attention network and user emotion | |
| CN115082147B (en) | Sequence recommendation method and device based on hypergraph neural network | |
| CN111339415B (en) | Click rate prediction method and device based on multi-interactive attention network | |
| CN111737578B (en) | Recommendation method and system | |
| CN111080400B (en) | Commodity recommendation method and system based on gate control graph convolution network and storage medium | |
| CN110955826B (en) | Recommendation system based on improved cyclic neural network unit | |
| CN111177575A (en) | Content recommendation method and device, electronic equipment and storage medium | |
| CN112288471B (en) | Advertisement click rate prediction method based on user history behavior sequence | |
| CN111325579A (en) | Advertisement click rate prediction method | |
| CN109657156A (en) | A kind of personalized recommendation method generating confrontation network based on circulation | |
| CN111581520A (en) | Item recommendation method and system based on item importance in session | |
| CN109034960B (en) | Multi-attribute inference method based on user node embedding | |
| US20220171760A1 (en) | Data processing method and apparatus, computer-readable storage medium, and electronic device | |
| Li et al. | A CTR prediction model based on user interest via attention mechanism | |
| CN112527966A (en) | Network text emotion analysis method based on Bi-GRU neural network and self-attention mechanism | |
| CN110619540A (en) | Click stream estimation method of neural network | |
| CN115186097A (en) | Knowledge graph and reinforcement learning based interactive recommendation method | |
| CN110175857A (en) | It is preferred that business determines method and device | |
| CN109189922B (en) | Comment evaluation model training method and device | |
| CN115760271A (en) | Electromechanical commodity personalized recommendation method and system based on graph neural network | |
| CN114925270A (en) | Session recommendation method and model | |
| Wang et al. | Session-based recommendation with time-aware neural attention network | |
| Ren et al. | A co-attention based multi-modal fusion network for review helpfulness prediction | |
| CN113888238B (en) | Advertisement click rate prediction method and device and computer equipment | |
| CN112270570B (en) | Click conversion rate prediction method based on feature combination and representation learning |
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 |