CN114386688B - User intention prediction method and system based on multi-data fusion - Google Patents
User intention prediction method and system based on multi-data fusion Download PDFInfo
- Publication number
- CN114386688B CN114386688B CN202210012281.7A CN202210012281A CN114386688B CN 114386688 B CN114386688 B CN 114386688B CN 202210012281 A CN202210012281 A CN 202210012281A CN 114386688 B CN114386688 B CN 114386688B
- Authority
- CN
- China
- Prior art keywords
- user
- sequence
- intention
- prediction
- neural network
- 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.)
- Active
Links
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
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield 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
- G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
- G06N3/061—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using biological neurons, e.g. biological neurons connected to an integrated circuit
-
- 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
Abstract
The invention provides a user intention prediction method and a system based on multi-data fusion, which comprises the following steps: collecting historical operation data of a certain user, and training a user intention prediction neural network aiming at the user according to the historical operation data; collecting current multiple items of operation data of the user, and fusing the multiple items of operation data into an operation behavior sequence according to time; according to the operation behavior sequence, obtaining a behavior prediction result of the user through a user intention prediction neural network; and determining the corresponding intention according to the behavior prediction result. The invention realizes that the operation behavior which the user wants to carry out and the intention corresponding to the operation behavior are determined based on the historical operation habit of the user according to the current multiple items of operation data of the user. And then, connection preparation is made in advance for realizing the next operation of the user, and the access efficiency of the user is improved. And the user can be provided with corresponding operation behaviors and recommendation services of commodity products according to the predicted intention.
Description
Technical Field
The invention relates to the technical field of network information, in particular to a user intention prediction method and a user intention prediction system based on multi-data fusion.
Background
With the rapid development of modern information technology represented by the internet in China, online shopping, online marketing services, video software of each, other entertainment service platforms and learning platforms begin to emerge on a large scale. The users inevitably spend a lot of time on the service software in daily life, so how to provide better operation experience for the users is a very important problem for many internet enterprises.
Today, many internet enterprises still provide service recommendations to users by acquiring personal data (such as phone numbers, locations, operating states, consumption frequencies and capabilities) of users simply from collecting the personal operation data of the users in order to provide good use experiences for the users, thereby improving the income benefits of the companies. However, a user intention prediction method and a user intention prediction system which predict the intention of a user through the historical access behavior data and the access operation steps of the user so as to make a connection preparation in advance for the next operation of the user according to the prediction result to improve the access efficiency of the user, provide corresponding operation behaviors and recommendation service of commodity products for the user and achieve the purpose of improving the income benefit of a company are lacked.
Disclosure of Invention
The invention provides a user intention prediction method and a user intention prediction system based on multi-data fusion, which are used for determining an operation behavior which a user wants to carry out and an intention corresponding to the operation behavior based on historical operation habits of the user according to current multi-item operation data of the user. And then, connection preparation is made in advance for realizing the next operation of the user, and the access efficiency of the user is improved. And the user can be provided with corresponding operation behaviors and recommendation services of commodity products according to the predicted intention.
The invention provides a user intention prediction method based on multi-data fusion, which comprises the following steps:
collecting historical operation data of a certain user, and training a user intention prediction neural network aiming at the user according to the historical operation data;
collecting current multiple items of operation data of the user, and fusing the multiple items of operation data into an operation behavior sequence according to time;
according to the operation behavior sequence, obtaining a behavior prediction result of the user through the user intention prediction neural network;
and determining the corresponding intention according to the behavior prediction result.
Preferably, the collecting historical operation data of a certain user and training the user intention prediction neural network for the user according to the historical operation data comprises:
collecting historical operation data of a certain user, and determining a plurality of operation items in the historical operation data;
determining a plurality of operation starting items and a plurality of operation final items in the plurality of operation items based on the sequence of the operation items;
determining a plurality of operation items from an operation starting item to an operation final item as an operation step sequence based on a corresponding relation between a preset operation final item and an intention judgment result, thereby obtaining a plurality of operation step sequences;
and training the pre-established RNN neural network by utilizing a plurality of operation step sequences to obtain the user intention prediction neural network.
Preferably, the training the pre-created RNN neural network by using a plurality of operation step sequences to obtain the user intention prediction neural network includes:
for a certain operation step sequence, determining an intention judgment result corresponding to an operation final item in the operation step sequence;
determining a necessary operation step sequence preset by the intention corresponding to the intention judgment result;
vectorizing the operation step sequence based on the necessary operation step sequence to obtain an operation step sequence vector;
taking the operation step sequence vector as an input item of the RNN neural network, taking an intention judgment result corresponding to an operation final item in the operation step sequence as an output item of the RNN neural network, and training the RNN neural network;
and training the RNN neural network by using the plurality of operation step sequence vectors to obtain a user intention prediction neural network.
Preferably, when the RNN neural network is trained, the training steps are as follows:
inputting a first value in a certain operation step sequence vector into a first neuron of the RNN neural network to output to obtain a first intermediate state vector;
inputting the first intermediate state vector into a second neuron, and simultaneously inputting a second value in the operation step sequence vector into the second neuron;
the second neuron outputs a second intermediate state vector, and the second intermediate state vector and a third value in the operation step sequence vector are used as the input of a third neuron to obtain a third intermediate vector;
all values in the operation step sequence vector are sequentially and correspondingly input into a plurality of neurons, and finally an output result is bound as an intention judgment result corresponding to an operation final item in the operation step sequence;
and training the RNN neural network by using a plurality of operation step sequences so as to obtain a user intention prediction neural network capable of outputting a plurality of intention judgment results.
Preferably, the acquiring a plurality of items of current operation data of the user, and integrating the plurality of items of operation data into the operation behavior sequence according to time includes:
acquiring a plurality of items of operation data generated from the beginning of executing the intention prediction work to the current time of a user side, and determining the operation time corresponding to each of the plurality of items of operation data in the operation data sequence;
based on the operation time of a plurality of items of operation data in the operation data sequence, sequencing the generated plurality of items of operation data to generate an operation data sequence;
determining the front-back relation among a plurality of items of operation data, eliminating the operation data of a round-trip repeated type in an operation data sequence based on the operation time of the plurality of items of operation data in the operation data sequence, and reserving a unidirectional operation data sequence;
and fusing a plurality of items of the operation data in the unidirectional operation data sequence to generate an operation behavior sequence.
Preferably, the obtaining of the behavior prediction result of the user through the user intention prediction neural network according to the operation behavior sequence includes:
inputting the operation behavior sequence as a first sequence into the user intention prediction neural network to obtain a first prediction result; the first prediction result is an intention prediction result which is obtained by screening a plurality of tendency probabilities of the user on all intentions according to the tendency probability of the user predicted by the current first sequence, wherein the tendency probabilities are not zero;
acquiring a new operation behavior sequence as a second sequence in real time, and determining a sequence length difference value between the second sequence and the first sequence;
if the sequence length difference is larger than a preset difference threshold value, inputting the second sequence into the user intention prediction neural network to obtain a second prediction result;
matching probability distribution results of tendency probabilities of the users on the intention prediction results in the first prediction result and the second prediction result to obtain matching values;
when the matching value is larger than a preset matching threshold value, extracting the intention prediction result of which the tendency probability is larger than a preset probability threshold value in the first prediction result to obtain a first column of prediction intents, and extracting the intention prediction result of which the tendency probability is larger than a preset probability threshold value in the second prediction result to obtain a second column of prediction intents;
and extracting a prediction intention group of a progressive relation existing in the first row of prediction intentions corresponding to the second row of prediction intentions according to a preset dependency relation among the prediction intentions, and taking the prediction intention group as a behavior prediction result of the user.
Preferably, the method further comprises predicting a neural network according to the user intention of the user obtained by training, determining the commonly used functions or operation behaviors of the user and solidifying the functions or operation behaviors, and the specific steps comprise:
determining a weight average value of connection weights among a plurality of neurons for obtaining one intention judgment result according to process data generated when the user intention of the user is trained to predict the neural network;
the multiple intention judgment results are arranged in a descending order according to the corresponding weight mean value, and N intention judgment results with the largest weight mean value are determined;
solidifying the functions or operation behaviors corresponding to the N intention judgment results, binding a trigger button and displaying the trigger button;
when a function or an operation behavior corresponding to a certain intention judgment result is solidified, feature extraction needs to be performed on the user intention prediction neural network in advance to obtain a behavior sequence feature group corresponding to each of multiple intention judgment results of the user;
determining a behavior sequence characteristic group corresponding to the intention judgment result, and determining all intermediate operation steps required to be executed for realizing the intention corresponding to the intention judgment result according to the behavior sequence characteristic group;
sequencing all the intermediate operation steps in sequence to obtain a step curing sequence, and determining fixed behavior information when the user executes a certain intermediate operation step in the step curing sequence according to historical operation data of the user; wherein the fixed behavior information comprises an icon, a window and input character information which need to be clicked when the user executes the intermediate operation step;
determining a user behavior sequence corresponding to the step curing sequence according to the fixed behavior information of each intermediate operation step in the step curing sequence;
and determining the automatic execution steps of the computer according to the user behavior sequence, packaging the automatic execution steps of the computer, binding a trigger button, and naming the trigger button.
Preferably, the method further includes detecting an operation link in which an incorrect operation is likely to occur in a process of implementing a certain intention of the user according to the historical operation data of the user, and specifically includes:
pre-extracting the characteristics of weighted values connected between all neurons in the user intention prediction neural network to obtain behavior sequence characteristic groups corresponding to various intention judgment results of the user;
screening out an operation behavior sequence which occurs in the process that the user realizes the intention A each time according to the historical operation data of the user;
determining a necessary operation behavior sequence of the user for realizing the intention A according to the behavior sequence characteristic group corresponding to the various intention judgment results of the user;
comparing the operation behavior sequence which appears in the process that the user realizes the intention A for one time with the necessary operation behavior sequence, and determining the position of the step node which appears in the process and operates back and forth;
counting the total times of the round-trip operation of each position for the positions of a plurality of step nodes of the round-trip operation of the user in the process of realizing the intention A for a plurality of times, determining an operation link corresponding to the position of the step node with the highest total times, and taking the operation link as an operation link of the user which is easy to have misoperation in the process of realizing the intention A.
In order to achieve the above object, the present invention further provides a system for predicting user intention based on multi-data fusion, including:
the neural network training module is used for acquiring historical operation data of a certain user and training a user intention prediction neural network aiming at the user according to the historical operation data;
the behavior data fusion module is used for collecting the current multiple items of operation data of the user and fusing the multiple items of operation data into an operation behavior sequence according to time;
the behavior prediction module is used for obtaining a behavior prediction result of the user through the user intention prediction neural network according to the operation behavior sequence;
and the intention determining module is used for determining the corresponding intention according to the behavior prediction result.
Preferably, the neural network training module includes:
the operation item identification unit is used for acquiring historical operation data of a certain user and determining a plurality of operation items in the historical operation data;
the operation item dividing unit is used for determining a plurality of operation starting items and a plurality of operation final items in a plurality of operation items based on the sequence of the operation items;
an operation step sequence determination unit configured to determine, based on a correspondence between a preset operation final item and a certain intention judgment result, a plurality of operation items including an operation start item and an operation final item as one operation step sequence, thereby obtaining a plurality of operation step sequences;
and the network training unit is used for training the pre-established RNN neural network by utilizing a plurality of operation step sequences to obtain the user intention prediction neural network.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by the practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.
The technical solution of the present invention is further described in detail by the accompanying drawings and embodiments.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the principles of the invention and not to limit the invention. In the drawings:
FIG. 1 is a flowchart illustrating steps of a method for predicting user intent based on multi-data fusion according to an embodiment of the present invention;
FIG. 2 is a schematic structural diagram of a system for predicting user intention based on multiple data fusion according to an embodiment of the present invention;
fig. 3 is a schematic structural diagram of a neural network training module in an embodiment of the present invention.
Detailed Description
The preferred embodiments of the present invention will be described in conjunction with the accompanying drawings, and it will be understood that they are described herein for the purpose of illustration and explanation and not limitation.
The invention provides a user intention prediction method based on multi-data fusion, as shown in fig. 1, comprising the following steps:
s1, collecting historical operation data of a certain user, and training a user intention prediction neural network aiming at the user according to the historical operation data;
s2, collecting current multiple items of operation data of the user, and fusing the multiple items of operation data into an operation behavior sequence according to time;
s3, according to the operation behavior sequence, obtaining a behavior prediction result of the user through a user intention prediction neural network;
and S4, determining the corresponding intention according to the behavior prediction result.
The working principle and the beneficial effects of the technical scheme are as follows: the method comprises the steps that historical operation data of a certain user are collected, and a user intention prediction neural network for the user is trained according to the historical operation data; collecting current multiple items of operation data of the user, and fusing the multiple items of operation data into an operation behavior sequence according to time; according to the operation behavior sequence, inputting a user intention prediction neural network to obtain a behavior prediction result of the user; and finally, determining the corresponding intention according to the behavior prediction result. Therefore, the operation behavior which the user wants to perform and the intention corresponding to the operation behavior are determined based on the historical operation habit of the user according to the current multiple items of operation data of the user. And then, connection preparation is made in advance for realizing the next operation of the user, and the access efficiency of the user is improved. And the corresponding operation behavior and the recommendation service of the commodity product can be provided for the user according to the predicted intention.
In a preferred embodiment, collecting historical operational data of a user and training a user intent prediction neural network for the user based on the historical operational data comprises:
collecting historical operation data of a certain user, and determining a plurality of operation items in the historical operation data;
determining a plurality of operation starting items and a plurality of operation final items in the plurality of operation items based on the sequence of the plurality of operation items;
determining a plurality of operation items from an operation starting item to an operation final item as an operation step sequence based on a corresponding relation between a preset operation final item and an intention judgment result, so as to obtain a plurality of operation step sequences;
and training the pre-established RNN neural network by utilizing a plurality of operation step sequences to obtain the user intention prediction neural network.
The working principle and the beneficial effects of the technical scheme are as follows: when training the user intention prediction neural network of the user, acquiring historical operation data of the user, and determining a plurality of operation items in the historical operation data; determining a plurality of operation starting items and a plurality of operation final items in the plurality of operation items according to a preset operation item table based on the sequence of the plurality of operation items; based on the corresponding relationship between the preset operation final item and a certain intention judgment result, determining a plurality of operation items from the operation starting item to the operation final item as an operation step sequence, thereby obtaining a plurality of operation step sequences, for example, for historical operation data of a user, the operation items comprise a plurality of operation items A-C-D-A-B-F-A which occur successively according to a preset operation item table such as [ operation starting item: A. g and H; operation final item: D. f, T ] determines an operation start item a therein and operation final items D, F therein, D being predetermined to correspond to the intention judgment result 1 and F corresponding to the intention judgment result 2, thereby determining an operation step sequence a-C-D corresponding to the intention judgment result 1 and an operation step sequence a-B-F corresponding to the intention judgment result 2; and training the pre-established RNN neural network by utilizing a plurality of operation step sequences to obtain the user intention prediction neural network. The method and the device realize quick reading and division of a plurality of operation step sequences in historical operation data, facilitate training of the pre-established RNN neural network by using the divided operation step sequences, and accelerate the training speed of the neural network.
In a preferred embodiment, training a pre-created RNN neural network with a plurality of sequences of operational steps to obtain a user-intention predicting neural network comprises:
for a certain operation step sequence, determining an intention judgment result corresponding to an operation final item in the operation step sequence;
determining a necessary operation step sequence preset by the intention corresponding to the intention judgment result;
vectorizing the operation step sequence based on the necessary operation step sequence to obtain an operation step sequence vector;
taking the operation step sequence vector as an input item of the RNN neural network, taking an intention judgment result corresponding to an operation final item in the operation step sequence as an output item of the RNN neural network, and training the RNN neural network;
and training the RNN neural network by using a plurality of operation step sequence vectors to obtain the user intention prediction neural network.
The working principle and the beneficial effects of the technical scheme are as follows: when a plurality of operation step sequences are used for training a pre-established RNN neural network, an intention judgment result corresponding to an operation final item in a certain operation step sequence needs to be determined; determining a necessary operation step sequence preset by the intention corresponding to the intention judgment result; vectorizing the operation step sequence based on the necessary operation step sequence to obtain an operation step sequence vector, for example, for the operation step sequence A-B-C-B-E-F, vectorizing the operation step sequence A-B-C-B-E-F to obtain an operation step sequence vector [1, 0, 1] based on an intention preset by an intention corresponding to the intention judgment result 3 corresponding to the operation final item D; taking the operation step sequence vector as an input item of the RNN neural network, taking an intention judgment result corresponding to an operation final item in the operation step sequence as an output item of the RNN neural network, and training the RNN neural network; and training the RNN neural network by using a plurality of operation step sequence vectors to obtain the user intention prediction neural network. And training the RNN neural network by using a plurality of operation step sequence vectors based on the historical operation data of the user person to obtain the personal intention prediction neural network of the user.
In a preferred embodiment, when the RNN neural network is trained, the training steps are as follows:
inputting a first value in a certain operation step sequence vector into a first neuron of an RNN neural network to output to obtain a first intermediate state vector;
inputting the first intermediate state vector into a second neuron, and simultaneously inputting a second value in the operation step sequence vector into the second neuron;
the second neuron outputs a second intermediate state vector, and the second intermediate state vector and a third value in the operation step sequence vector are used as the input of a third neuron to obtain a third intermediate vector;
all values in the operation step sequence vector are sequentially and correspondingly input into a plurality of neurons, and finally an output result is bound as an intention judgment result corresponding to an operation final item in the operation step sequence;
the RNN neural network is trained by using a plurality of operation step sequences, so that a user intention prediction neural network capable of outputting a plurality of intention judgment results is obtained.
The working principle and the beneficial effects of the technical scheme are as follows: when the RNN neural network is trained, inputting a first value in a sequence vector of a certain operation step into a first neuron of the RNN neural network to output to obtain a first intermediate state vector; inputting the first intermediate state vector into a second neuron, and simultaneously inputting a second value in the operation step sequence vector into the second neuron; the second neuron outputs a second intermediate state vector, and the second intermediate state vector and a third value in the operation step sequence vector are used as the input of a third neuron to obtain a third intermediate vector; all values in the operation step sequence vector are sequentially and correspondingly input into a plurality of neurons, and finally an output result is bound as an intention judgment result corresponding to an operation final item in the operation step sequence; the RNN neural network is trained by using a plurality of operation step sequences, so that a user intention prediction neural network capable of outputting a plurality of intention judgment results is obtained. The user intention prediction neural network is trained through a plurality of different operation step sequences corresponding to different intentions by the method, so that the user intention prediction neural network can output a plurality of intention judgment results corresponding to a plurality of intention judgment results, and the user intention prediction neural network trained by the method has strong universality.
In a preferred embodiment, collecting a plurality of items of operation data of the user at present, and integrating the plurality of items of operation data into an operation behavior sequence according to time includes:
acquiring a plurality of items of operation data generated by a user terminal at the current moment from the beginning of executing the intention prediction work, and determining operation time corresponding to the plurality of items of operation data in an operation data sequence;
based on the operation time of a plurality of items of operation data in the operation data sequence, sequencing the generated plurality of items of operation data to generate an operation data sequence;
determining the front-back relation among a plurality of items of operation data, eliminating the operation data of a round-trip repeated type in the operation data sequence based on the operation time of the plurality of items of operation data in the operation data sequence, and reserving a one-way operation data sequence;
and fusing multiple pieces of operation data in the unidirectional operation data sequence to generate an operation behavior sequence.
The working principle and the beneficial effects of the technical scheme are as follows: when multiple items of operation data are fused into an operation behavior sequence according to time, multiple items of operation data generated from the beginning of executing the intention prediction work to the current time of a user side need to be obtained, and operation time corresponding to the multiple items of operation data in the operation data sequence is determined; based on the operation time of a plurality of items of operation data in the operation data sequence, sequencing the generated plurality of items of operation data to generate an operation data sequence; determining the front-back relation among a plurality of items of operation data, eliminating the operation data of a round-trip repeated type in the operation data sequence based on the operation time of the plurality of items of operation data in the operation data sequence, and reserving a unidirectional operation data sequence; and fusing a plurality of operation data in the unidirectional operation data sequence to generate an operation behavior sequence. The operation data of the round trip repeat class is eliminated, and the influence of the misoperation data on the judgment result is prevented. The fusion of a plurality of items of operation data generated by the user side is realized.
In a preferred embodiment, obtaining the behavior prediction result of the user through the user intention prediction neural network according to the operation behavior sequence comprises:
inputting the operation behavior sequence as a first sequence into a user intention prediction neural network to obtain a first prediction result; the first prediction result is an intention prediction result which is obtained by screening a plurality of tendency probabilities which are not zero according to tendency probabilities of the user on all intentions predicted by the current first sequence;
acquiring a new operation behavior sequence as a second sequence in real time, and determining a sequence length difference value between the second sequence and the first sequence;
if the sequence length difference is larger than a preset difference threshold value, inputting a second sequence into the user intention prediction neural network to obtain a second prediction result;
matching probability distribution results of tendency probabilities of the users on all the intention prediction results in the first prediction result and the second prediction result to obtain matching values, wherein the calculation formula is as follows:
wherein S is the matching degree, n is the total number of the intention prediction results in the first prediction result and the second prediction result, and x i Is the probability of inclination, y, of the first prediction to the intended prediction i i Is the tendency probability of the corresponding intention prediction result i in the second prediction result;
when the matching value is larger than a preset matching threshold value, extracting the intention prediction result of which the tendency probability is larger than a preset probability threshold value in the first prediction result to obtain a first row of prediction intents, and extracting the intention prediction result of which the tendency probability is larger than the preset probability threshold value in the second prediction result to obtain a second row of prediction intents;
and extracting a prediction intention group of a progressive relation existing in the first column of prediction intentions corresponding to the second column of prediction intentions according to the dependency relation among the preset prediction intentions, and taking the prediction intention group as a behavior prediction result of the user.
The working principle and the beneficial effects of the technical scheme are as follows: when the behavior of the user is predicted, the operation behavior sequence is used as a first sequence to be input into a user intention prediction neural network, and a first prediction result is obtained, wherein the first prediction result is [ intention A-80%, intention B-10%, intention C-2% and intention D-8% ]; the first prediction result is an intention prediction result which is obtained by screening a plurality of tendency probabilities which are not zero according to tendency probabilities of the user on all intentions predicted by the current first sequence; acquiring a new operation behavior sequence as a second sequence in real time, and determining a sequence length difference value between the second sequence and the first sequence; if the difference value of the sequence lengths is larger than a preset difference value threshold value, inputting the second sequence into the user intention prediction neural network to obtain a second prediction result, such as intention A-60%, intention B-20%, intention C-12% and intention D-8%; matching probability distribution results of tendency probabilities of the users on all intention prediction results in the first prediction result and the second prediction result to obtain matching values; when the matching value is larger than a preset matching threshold value, extracting the intention prediction result of which the tendency probability is larger than a preset probability threshold value in the first prediction result to obtain a first row of prediction intents, and extracting the intention prediction result of which the tendency probability is larger than the preset probability threshold value in the second prediction result to obtain a second row of prediction intents; and extracting a prediction intention group of a progressive relation existing in the first row of prediction intentions corresponding to the second row of prediction intentions according to the preset dependency relation among the prediction intentions, and taking the prediction intention group as a behavior prediction result of the user. Therefore, the user behaviors can be more accurately predicted according to the operation behavior sequences acquired twice.
In a preferred embodiment, the method further comprises the steps of predicting a neural network according to the user intention of the user obtained by training, determining the commonly used functions or operation behaviors of the user and solidifying the functions or operation behaviors, wherein the specific steps comprise:
determining a weight average value of connection weights among a plurality of neurons for obtaining one intention judgment result according to process data generated when the user intention of the user is trained and the neural network is predicted;
the multiple intention judgment results are arranged in a descending order according to the size of the corresponding weight mean value, and N intention judgment results with the largest weight mean value are determined;
solidifying the functions or operation behaviors corresponding to the N intention judgment results, binding a trigger button and displaying the trigger button;
when a function or an operation behavior corresponding to a certain intention judgment result is solidified, feature extraction needs to be performed on a user intention prediction neural network in advance to obtain behavior sequence feature groups corresponding to various intention judgment results of a user;
determining a behavior sequence characteristic group corresponding to the intention judgment result, and determining all intermediate operation steps required to be executed for realizing the intention corresponding to the intention judgment result according to the behavior sequence characteristic group;
sequencing all the intermediate operation steps in sequence to obtain a step curing sequence, and determining fixed behavior information when the user executes a certain intermediate operation step in the step curing sequence according to historical operation data of the user; wherein the fixed behavior information comprises an icon, a window and input character information which need to be clicked when the user executes the intermediate operation step;
determining a user behavior sequence corresponding to the step curing sequence according to the fixed behavior information of each intermediate operation step in the step curing sequence;
determining the automatic execution steps of the computer according to the user behavior sequence, packaging the automatic execution steps of the computer, binding a trigger button, and naming the trigger button.
The working principle and the beneficial effects of the technical scheme are as follows: determining and solidifying functions or operation behaviors commonly used by the user according to the user intention prediction neural network of the user obtained by training, wherein the specific steps comprise determining and obtaining a weight mean value of connection weights among a plurality of neurons of one intention judgment result according to process data generated when the user intention prediction neural network of the user is trained; the multiple intention judgment results are arranged in a descending order according to the size of the corresponding weight mean value, and N intention judgment results with the largest weight mean value are determined; after the functions or operation behaviors corresponding to the N intention judgment results are solidified, binding a trigger button and displaying the trigger button, realizing locking of an operation behavior sequence of the repeated operation of the user, binding the trigger button after programming the series of behaviors, and automatically executing the subsequent series of steps when the user clicks the trigger button, thereby shortening the operation time of the user and improving the experience of the user; when a function or an operation behavior corresponding to a certain intention judgment result is solidified, feature extraction needs to be carried out on a user intention prediction neural network in advance to obtain a behavior sequence feature group corresponding to various intention judgment results of the user; determining a behavior sequence characteristic group corresponding to the intention judgment result, and determining all intermediate operation steps required to be executed for realizing the intention corresponding to the intention judgment result according to the behavior sequence characteristic group; sequencing all the intermediate operation steps in sequence to obtain a step curing sequence, and determining fixed behavior information when the user executes a certain intermediate operation step in the step curing sequence according to historical operation data of the user; wherein the fixed behavior information comprises an icon, a window and input character information which need to be clicked when the user executes the intermediate operation step; determining a user behavior sequence corresponding to the step curing sequence according to the fixed behavior information of each intermediate operation step in the step curing sequence; determining the automatic execution steps of the computer according to the user behavior sequence, packaging the automatic execution steps of the computer, binding a trigger button, and naming the trigger button. The method realizes the locking of the frequently operated operation behaviors or the frequently used functional step sequences of the user, programs the series of behaviors or steps, binds the trigger button, and automatically executes the subsequent series of steps when the user clicks the trigger button, thereby shortening the operation time of the user and improving the experience of the user.
In a preferred embodiment, the method further includes detecting, according to the historical operation data of the user, an operation link in which an incorrect operation is likely to occur in a process of implementing a certain intention by the user, and specifically includes:
pre-extracting the characteristics of weighted values connected among all neurons in the user intention prediction neural network to obtain behavior sequence characteristic groups corresponding to various intention judgment results of the user;
screening out an operation behavior sequence which appears in the process of realizing the intention A of the user each time according to the historical operation data of the user;
determining a necessary operation behavior sequence of the user for realizing the intention A according to behavior sequence feature groups corresponding to various intention judgment results of the user;
comparing an operation behavior sequence which appears in the process that the user realizes the intention A at a certain time with a necessary operation behavior sequence, and determining the position of a step node which appears in the process and carries out the back-and-forth operation;
counting the total times of the round-trip operation of each position for the positions of a plurality of step nodes of the round-trip operation of the user in the process of realizing the intention A for a plurality of times, determining an operation link corresponding to the position of the step node with the highest total times, and taking the operation link as an operation link of the user which is easy to have misoperation in the process of realizing the intention A.
The working principle and the beneficial effects of the technical scheme are as follows: according to the historical operation data of the user, an operation link for detecting that the user is easy to have misoperation in the process of realizing a certain intention comprises the steps of extracting features of a user intention prediction neural network in advance to obtain behavior sequence feature groups corresponding to various intention judgment results of the user, and extracting weighted values connected among all neurons as features in the output process of a specific result intention A of the user intention prediction neural network (RNN recurrent neural network) when the features of the user intention prediction neural network are extracted to obtain the behavior sequence feature group corresponding to the intention A; screening out an operation behavior sequence which appears in the process of realizing the intention A of the user each time according to the historical operation data of the user; determining the behavior input by the neurons at the two ends of the linkage with the largest weight value among the hidden layer neurons corresponding to the behavior sequence feature group as the necessary operation behavior according to the behavior sequence feature group corresponding to each of the multiple intention judgment results of the user, thereby determining the necessary operation behavior sequence of the user for realizing the intention A; comparing an operation behavior sequence which appears in the process that the user realizes the intention A for one time with a necessary operation behavior sequence, and determining the position of a step node which appears in the process and performs back-and-forth operation; counting the total times of the round-trip operation of each position for the positions of a plurality of step nodes of the round-trip operation of the user in the process of realizing the intention A for a plurality of times, determining an operation link corresponding to the position of the step node with the highest total times, and taking the operation link as an operation link of the user which is easy to have misoperation in the process of realizing the intention A. Therefore, the link of repeated back-and-forth operation of the user in the operation process is identified, the step is determined to be a nonsense step, and the corresponding function or operation behavior solidification function is further suggested for the client.
To achieve the above object, the present invention further provides a system for predicting user intention based on multi-data fusion, as shown in fig. 2, including:
the neural network training module 1 is used for acquiring historical operation data of a certain user and training a user intention prediction neural network aiming at the user according to the historical operation data;
the behavior data fusion module 2 is used for collecting the current multiple items of operation data of the user and fusing the multiple items of operation data into an operation behavior sequence according to time;
the behavior prediction module 3 is used for predicting a neural network through the intention of the user according to the operation behavior sequence to obtain a behavior prediction result of the user;
and the intention determining module 4 is used for determining the corresponding intention according to the behavior prediction result.
The working principle and the beneficial effects of the technical scheme are as follows: the method comprises the steps that historical operation data of a user are collected through a neural network training module 1, and a neural network is predicted according to the user intention of the user through training of the historical operation data; collecting current multiple items of operation data of the user through a behavior data fusion module 2, and fusing the multiple items of operation data into an operation behavior sequence according to time; inputting a user intention prediction neural network through a behavior prediction module 3 according to the operation behavior sequence to obtain a behavior prediction result of the user; finally, the corresponding intention is determined by the intention determining module 4 according to the behavior prediction result. Therefore, the operation behavior which the user wants to carry out and the intention corresponding to the operation behavior are determined based on the historical operation habit of the user according to the current multiple items of operation data of the user. And then, connection preparation is made in advance for realizing the next operation of the user, and the access efficiency of the user is improved. And the user can be provided with corresponding operation behaviors and recommendation services of commodity products according to the predicted intention.
In a preferred embodiment, as shown in fig. 3, the neural network training module comprises:
the operation item identification unit 11 is configured to collect historical operation data of a certain user, and determine a plurality of operation items in the historical operation data;
the operation item dividing unit 12 is configured to determine a plurality of operation start items and a plurality of operation final items in the plurality of operation items based on a sequence of the plurality of operation items;
an operation step sequence determining unit 13, configured to determine, based on a corresponding relationship between a preset operation final item and a certain intention judgment result, that a plurality of operation items, from an operation start item to the operation final item, are included as one operation step sequence, so as to obtain a plurality of operation step sequences;
and the network training unit 14 is configured to train the pre-created RNN neural network by using a plurality of operation step sequences to obtain a user intention prediction neural network.
The working principle and the beneficial effects of the technical scheme are as follows: when training the user intention prediction neural network of the user, acquiring historical operation data of the user through an operation item identification unit 11, and determining a plurality of operation items in the historical operation data; determining a plurality of operation starting items and a plurality of operation final items in the plurality of operation items by the operation item dividing unit 12 based on the sequence of the plurality of operation items; determining, by the operation step sequence determining unit 13, a plurality of operation items from the operation start item to the operation end item as one operation step sequence based on a correspondence between a preset operation end item and a certain intention judgment result, thereby obtaining a plurality of operation step sequences, for example, determining, for the historical operation data of the user, the operation start item a and the operation end items D and F existing therein, wherein D is predetermined to correspond to the intention judgment result 1 and F is predetermined to correspond to the intention judgment result 2, and thereby determining the operation step sequences a-C-D corresponding to the intention judgment result 1 and the operation step sequences a-B-F corresponding to the intention judgment result 2, and the plurality of operation items from the operation start item to the operation end item are included according to a preset operation item characteristic; finally, the network training unit 14 trains the pre-created RNN neural network by using a plurality of operation step sequences, so as to obtain the user intention prediction neural network. The method and the device realize quick reading and division of a plurality of operation step sequences in historical operation data, and facilitate training of the pre-established RNN neural network by using the divided operation step sequences, thereby accelerating the training speed of the neural network.
It will be apparent to those skilled in the art that various changes and modifications may be made in the present invention without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.
Claims (5)
1. A user intention prediction method based on multi-data fusion is characterized by comprising the following steps:
collecting historical operation data of a certain user, and training a user intention prediction neural network aiming at the user according to the historical operation data;
collecting current multiple items of operation data of the user, and fusing the multiple items of operation data into an operation behavior sequence according to time;
according to the operation behavior sequence, obtaining a behavior prediction result of the user through the user intention prediction neural network;
determining a corresponding intention according to the behavior prediction result;
the collecting historical operation data of a certain user and training the user intention prediction neural network aiming at the user according to the historical operation data comprises the following steps:
collecting historical operation data of a certain user, and determining a plurality of operation items in the historical operation data;
determining a plurality of operation starting items and a plurality of operation final items in the plurality of operation items based on the sequence of the operation items;
determining a plurality of operation items from an operation starting item to an operation final item as an operation step sequence based on a corresponding relation between a preset operation final item and an intention judgment result, thereby obtaining a plurality of operation step sequences;
training a pre-established RNN neural network by utilizing a plurality of operation step sequences to obtain a user intention prediction neural network;
the training of the pre-established RNN neural network by using a plurality of operation step sequences to obtain the user intention prediction neural network comprises the following steps:
for a certain operation step sequence, determining an intention judgment result corresponding to an operation final item in the operation step sequence;
determining a necessary operation step sequence preset by the intention corresponding to the intention judgment result;
vectorizing the operation step sequence based on the necessary operation step sequence to obtain an operation step sequence vector;
taking the operation step sequence vector as an input item of the RNN neural network, taking an intention judgment result corresponding to an operation final item in the operation step sequence as an output item of the RNN neural network, and training the RNN neural network;
training the RNN neural network by using a plurality of operation step sequence vectors to obtain a user intention prediction neural network;
when the RNN neural network is trained, the training steps are as follows:
inputting a first value in a certain operation step sequence vector into a first neuron of the RNN neural network to output to obtain a first intermediate state vector;
inputting the first intermediate state vector into a second neuron, and simultaneously inputting a second value in the operation step sequence vector into the second neuron;
the second neuron outputs a second intermediate state vector, and the second intermediate state vector and a third value in the operation step sequence vector are used as the input of a third neuron to obtain a third intermediate vector;
all values in the operation step sequence vector are sequentially and correspondingly input into a plurality of neurons, and finally an output result is bound as an intention judgment result corresponding to an operation final item in the operation step sequence;
training the RNN neural network by using a plurality of operation step sequences so as to obtain a user intention prediction neural network capable of outputting a plurality of intention judgment results;
the obtaining of the behavior prediction result of the user through the user intention prediction neural network according to the operation behavior sequence comprises:
inputting the operation behavior sequence serving as a first sequence into the user intention prediction neural network to obtain a first prediction result; the first prediction result is an intention prediction result which is obtained by screening a plurality of tendency probabilities which are not zero according to tendency probabilities of the user on all intentions predicted by the current first sequence;
acquiring a new operation behavior sequence in real time as a second sequence, and determining a sequence length difference value between the second sequence and the first sequence;
if the sequence length difference is larger than a preset difference threshold value, inputting the second sequence into the user intention prediction neural network to obtain a second prediction result;
matching probability distribution results of tendency probabilities of the users on the intention prediction results in the first prediction result and the second prediction result to obtain matching values;
when the matching value is larger than a preset matching threshold value, extracting the intention prediction result of which the tendency probability is larger than a preset probability threshold value in the first prediction result to obtain a first row of prediction intents, and extracting the intention prediction result of which the tendency probability is larger than the preset probability threshold value in the second prediction result to obtain a second row of prediction intents;
and extracting a prediction intention group of a progressive relation existing in the first row of prediction intentions corresponding to the second row of prediction intentions according to a preset dependency relation among the prediction intentions, and taking the prediction intention group as a behavior prediction result of the user.
2. The method according to claim 1, wherein the collecting operation data of a plurality of current items of the user and fusing the operation data into the operation behavior sequence according to time comprises:
acquiring a plurality of items of operation data generated from the beginning of executing the intention prediction work to the current time of a user terminal, generating an operation data sequence, and determining operation time corresponding to the plurality of items of operation data in the operation data sequence;
based on the operation time of a plurality of items of operation data in the operation data sequence, sequencing the generated plurality of items of operation data to generate an operation data sequence;
determining the front-back relation among a plurality of items of operation data, eliminating the operation data of a round-trip repeated type in an operation data sequence based on the operation time of the plurality of items of operation data in the operation data sequence, and reserving a unidirectional operation data sequence;
and fusing a plurality of items of the operation data in the unidirectional operation data sequence to generate an operation behavior sequence.
3. The method for predicting user intention based on multiple data fusion as claimed in claim 1, further comprising determining and consolidating commonly used functions or operational behaviors of the user according to the trained user intention prediction neural network of the user, the specific steps comprising:
determining a weight average value of connection weights among a plurality of neurons for obtaining one intention judgment result according to process data generated when the user intention of the user is trained to predict the neural network;
the multiple intention judgment results are arranged in a descending order according to the size of the corresponding weight mean value, and N intention judgment results with the largest weight mean value are determined;
solidifying the functions or operation behaviors corresponding to the N intention judgment results, binding a trigger button and displaying the trigger button;
when a function or an operation behavior corresponding to a certain intention judgment result is solidified, feature extraction needs to be performed on the user intention prediction neural network in advance to obtain a behavior sequence feature group corresponding to each of multiple intention judgment results of the user;
determining a behavior sequence characteristic group corresponding to the intention judgment result, and determining all intermediate operation steps required to be executed for realizing the intention corresponding to the intention judgment result according to the behavior sequence characteristic group;
sequencing all the intermediate operation steps in sequence to obtain a step curing sequence, and determining fixed behavior information when the user executes a certain intermediate operation step in the step curing sequence according to historical operation data of the user; wherein the fixed behavior information comprises an icon, a window and input character information which need to be clicked when the user executes the intermediate operation step;
determining a user behavior sequence corresponding to the step curing sequence according to the fixed behavior information of each intermediate operation step in the step curing sequence;
and determining the automatic execution steps of the computer according to the user behavior sequence, packaging the automatic execution steps of the computer, binding a trigger button, and naming the trigger button.
4. The method for predicting user intention based on multiple data fusion as claimed in claim 1, further comprising detecting an operation link that is likely to cause a misoperation in a process of implementing a certain intention of the user according to historical operation data of the user, specifically comprising:
pre-extracting the characteristics of weighted values connected between all neurons in the user intention prediction neural network to obtain behavior sequence characteristic groups corresponding to various intention judgment results of the user;
screening out an operation behavior sequence which appears in the process of realizing the intention A of the user each time according to the historical operation data of the user;
determining a necessary operation behavior sequence of the user for realizing the intention A according to the behavior sequence characteristic group corresponding to the various intention judgment results of the user;
comparing the operation behavior sequence which appears in the process that the user realizes the intention A for one time with the necessary operation behavior sequence, and determining the position of the step node which appears in the process and operates back and forth;
counting the total times of the round-trip operation of each position for the positions of a plurality of step nodes of the round-trip operation of the user in the process of realizing the intention A for a plurality of times, determining an operation link corresponding to the position of the step node with the highest total times, and taking the operation link as an operation link of the user which is easy to have misoperation in the process of realizing the intention A.
5. A system for predicting user intent based on multidata fusion, comprising:
the neural network training module is used for acquiring historical operation data of a certain user and training a user intention prediction neural network aiming at the user according to the historical operation data;
the behavior data fusion module is used for collecting the current multiple items of operation data of the user and fusing the multiple items of operation data into an operation behavior sequence according to time;
the behavior prediction module is used for obtaining a behavior prediction result of the user through the user intention prediction neural network according to the operation behavior sequence;
the intention determining module is used for determining the corresponding intention according to the behavior prediction result;
the neural network training module comprises:
the operation item identification unit is used for acquiring historical operation data of a certain user and determining a plurality of operation items in the historical operation data;
the operation item dividing unit is used for determining a plurality of operation starting items and a plurality of operation final items in the plurality of operation items based on the sequence of the plurality of operation items;
an operation step sequence determination unit configured to determine, based on a correspondence between a preset operation final item and a certain intention judgment result, a plurality of operation items including an operation start item and an operation final item as one operation step sequence, thereby obtaining a plurality of operation step sequences;
the network training unit is used for training a pre-established RNN neural network by utilizing a plurality of operation step sequences to obtain a user intention prediction neural network;
the training of the pre-created RNN neural network by using the plurality of operation step sequences to obtain the user intention prediction neural network includes:
for a certain operation step sequence, determining an intention judgment result corresponding to an operation final item in the operation step sequence;
determining a necessary operation step sequence preset by the intention corresponding to the intention judgment result;
vectorizing the operation step sequence based on the necessary operation step sequence to obtain an operation step sequence vector;
taking the operation step sequence vector as an input item of the RNN neural network, taking an intention judgment result corresponding to an operation final item in the operation step sequence as an output item of the RNN neural network, and training the RNN neural network;
training the RNN neural network by using a plurality of operation step sequence vectors to obtain a user intention prediction neural network;
when the RNN neural network is trained, the training steps are as follows:
inputting a first value in a certain operation step sequence vector into a first neuron of the RNN neural network to output to obtain a first intermediate state vector;
inputting the first intermediate state vector into a second neuron, and simultaneously inputting a second value in the operation step sequence vector into the second neuron;
the second neuron outputs a second intermediate state vector, and the second intermediate state vector and a third value in the operation step sequence vector are used as the input of a third neuron to obtain a third intermediate vector;
all values in the operation step sequence vector are sequentially and correspondingly input into a plurality of neurons, and finally an output result is bound as an intention judgment result corresponding to an operation final item in the operation step sequence;
training the RNN neural network by using a plurality of operation step sequences so as to obtain a user intention prediction neural network capable of outputting a plurality of intention judgment results;
the obtaining of the behavior prediction result of the user through the user intention prediction neural network according to the operation behavior sequence comprises:
inputting the operation behavior sequence serving as a first sequence into the user intention prediction neural network to obtain a first prediction result; the first prediction result is an intention prediction result which is obtained by screening a plurality of tendency probabilities of the user on all intentions according to the tendency probability of the user predicted by the current first sequence, wherein the tendency probabilities are not zero;
acquiring a new operation behavior sequence as a second sequence in real time, and determining a sequence length difference value between the second sequence and the first sequence;
if the sequence length difference is larger than a preset difference threshold value, inputting the second sequence into the user intention prediction neural network to obtain a second prediction result;
matching probability distribution results of tendency probabilities of the users on the intention prediction results in the first prediction result and the second prediction result to obtain matching values;
when the matching value is larger than a preset matching threshold value, extracting the intention prediction result of which the tendency probability is larger than a preset probability threshold value in the first prediction result to obtain a first row of prediction intents, and extracting the intention prediction result of which the tendency probability is larger than the preset probability threshold value in the second prediction result to obtain a second row of prediction intents;
and extracting a prediction intention group of a progressive relation existing in the first row of prediction intentions corresponding to the second row of prediction intentions according to a preset dependency relation among the prediction intentions, and taking the prediction intention group as a behavior prediction result of the user.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210012281.7A CN114386688B (en) | 2022-01-07 | 2022-01-07 | User intention prediction method and system based on multi-data fusion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210012281.7A CN114386688B (en) | 2022-01-07 | 2022-01-07 | User intention prediction method and system based on multi-data fusion |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114386688A CN114386688A (en) | 2022-04-22 |
| CN114386688B true CN114386688B (en) | 2022-12-06 |
Family
ID=81200018
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210012281.7A Active CN114386688B (en) | 2022-01-07 | 2022-01-07 | User intention prediction method and system based on multi-data fusion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114386688B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11436501B1 (en) * | 2019-08-09 | 2022-09-06 | Bottomline Technologies, Inc. | Personalization of a user interface using machine learning |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102130732A (en) * | 2011-04-01 | 2011-07-20 | 北京邮电大学 | Cooperative spectrum detection method for cognitive radio based on neural network |
| CN108363690A (en) * | 2018-02-08 | 2018-08-03 | 北京十三科技有限公司 | Dialog semantics Intention Anticipation method based on neural network and learning training method |
| CN113780684A (en) * | 2021-10-15 | 2021-12-10 | 国网福建省电力有限公司龙岩供电公司 | Intelligent building user energy consumption behavior prediction method based on LSTM neural network |
Family Cites Families (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108446374B (en) * | 2018-03-16 | 2019-09-20 | 北京三快在线科技有限公司 | User's Intention Anticipation method, apparatus, electronic equipment, storage medium |
| CN109558032B (en) * | 2018-12-05 | 2020-09-04 | 北京三快在线科技有限公司 | Operation processing method and device and computer equipment |
| CN111292105B (en) * | 2018-12-06 | 2023-12-08 | 北京嘀嘀无限科技发展有限公司 | Service demand determining method and device |
| CN109583659A (en) * | 2018-12-07 | 2019-04-05 | 南京富士通南大软件技术有限公司 | User's operation behavior prediction method and system based on deep learning |
| CN109814952A (en) * | 2019-01-30 | 2019-05-28 | 维沃移动通信有限公司 | A kind of application interface quickly starting control processing method, device and mobile terminal |
| US11176330B2 (en) * | 2019-07-22 | 2021-11-16 | Advanced New Technologies Co., Ltd. | Generating recommendation information |
| CN110471735A (en) * | 2019-08-21 | 2019-11-19 | 中国工商银行股份有限公司 | Operational view interface method for pushing and device |
| CN111833115B (en) * | 2020-09-16 | 2020-12-11 | 腾讯科技（深圳）有限公司 | Operation identification method and device, storage medium and server |
| CN113239272B (en) * | 2021-05-12 | 2022-11-29 | 烽火通信科技股份有限公司 | Intention prediction method and intention prediction device of network management and control system |
| CN113298321B (en) * | 2021-06-22 | 2022-03-11 | 深圳市查策网络信息技术有限公司 | User intention prediction method based on multi-data fusion |
-
2022
- 2022-01-07 CN CN202210012281.7A patent/CN114386688B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102130732A (en) * | 2011-04-01 | 2011-07-20 | 北京邮电大学 | Cooperative spectrum detection method for cognitive radio based on neural network |
| CN108363690A (en) * | 2018-02-08 | 2018-08-03 | 北京十三科技有限公司 | Dialog semantics Intention Anticipation method based on neural network and learning training method |
| CN113780684A (en) * | 2021-10-15 | 2021-12-10 | 国网福建省电力有限公司龙岩供电公司 | Intelligent building user energy consumption behavior prediction method based on LSTM neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114386688A (en) | 2022-04-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110598037B (en) | Image searching method, device and storage medium | |
| CN110503531B (en) | Dynamic social scene recommendation method based on time sequence perception | |
| CN107093084A (en) | Potential user predicts method for transformation and device | |
| CN109299344A (en) | The generation method of order models, the sort method of search result, device and equipment | |
| CN105183772A (en) | Release information click rate estimation method and apparatus | |
| CN110163647A (en) | A kind of data processing method and device | |
| CN109583468A (en) | Training sample acquisition methods, sample predictions method and corresponding intrument | |
| CN102789449B (en) | The method and apparatus that comment text is evaluated | |
| CN103761254A (en) | Method for matching and recommending service themes in various fields | |
| CN109903103A (en) | A kind of method and apparatus for recommending article | |
| CN111210072B (en) | Prediction model training and user resource limit determining method and device | |
| CN114386688B (en) | User intention prediction method and system based on multi-data fusion | |
| CN112559923A (en) | Website resource recommendation method and device, electronic equipment and computer storage medium | |
| CN110490683A (en) | A kind of method and system of the offline upper collaboration multi-model mixed recommendation of line | |
| CN111984873A (en) | Service recommendation system and method | |
| CN110083774A (en) | Using determination method, apparatus, computer equipment and the storage medium of recommendation list | |
| CN110910201A (en) | Information recommendation control method and device, computer equipment and storage medium | |
| GB2367402A (en) | Method and system for recognizing end-user transactions | |
| CN113570398A (en) | Promotion data processing method, model training method, system and storage medium | |
| CN109308564A (en) | The recognition methods of crowd's performance ratings, device, storage medium and computer equipment | |
| CN111831918A (en) | Mixed recommendation system in online education scene | |
| CN115025497A (en) | Abnormal data detection method and device, electronic equipment and storage medium | |
| CN115130542A (en) | Model training method, text processing device and electronic equipment | |
| CN112749850B (en) | Information updating method based on cloud computing and big data requirements and big data server | |
| CN112541806A (en) | Recommendation method and device based on heterogeneous information network |
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 | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20221114 Address after: Room 908, Building 2, No. 968 Jinzhong Road, Changning District, Shanghai, 200000 (the actual floor is 8 floors) Applicant after: CUMMY TECHNOLOGY (SHANGHAI) Co.,Ltd. Address before: 518109 floor 12, building h, gangzhilong science and Technology Park, No. 6, Qinglong Road, Qinghua community, Longhua street, Longhua District, Shenzhen, Guangdong Province Applicant before: Shenzhen jushang Dingli Network Technology Co.,Ltd. |
|
| TA01 | Transfer of patent application right | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |