CN109584020B - Information processing method and electronic equipment - Google Patents
Information processing method and electronic equipment Download PDFInfo
- Publication number
- CN109584020B CN109584020B CN201811475767.4A CN201811475767A CN109584020B CN 109584020 B CN109584020 B CN 109584020B CN 201811475767 A CN201811475767 A CN 201811475767A CN 109584020 B CN109584020 B CN 109584020B
- Authority
- CN
- China
- Prior art keywords
- input data
- processing model
- weight
- preset
- weight matrix
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—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
- 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
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
- G06Q30/0202—Market predictions or forecasting for commercial activities
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Strategic Management (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Human Resources & Organizations (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Data Mining & Analysis (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application provides an information processing method, which comprises the following steps: acquiring input data of at least two dimensions, wherein the input data of any dimension represents one characteristic of an object to be predicted; processing the input data according to a first processing model which is trained in advance to obtain a prediction result, and updating a first weight matrix in the first processing model based on the input data, wherein the first processing model comprises at least two groups of weight matrices. In the scheme, in the process of processing the input data based on the first processing model, the first matrix of the input data is also adjusted according to the received input data, and in the data processing process, the input data received each time is combined to adjust the input data, so that the residual matrix is kept unchanged. Therefore, when the first processing model is adopted for processing input data, factors of historical input data are considered, and factors of the input data are referred to, so that the data calculation burden is reduced under the condition that the stability of a predicted value is ensured.
Description
Technical Field
The present invention relates to the field of electronic devices, and more particularly, to an information processing method and an electronic device.
Background
With the development of information technology, a function of predicting existing data by using a prediction algorithm has been applied to various applications.
In the conventional prediction algorithm (such as K-Means or naive bayes), only the diversity of the data is considered, but the time sequence of the data is ignored, if only the incremental data is considered, the situation that the predicted value is frequently changed may occur, and if the full-scale data is used each time, a large amount of calculation resources are consumed, so that the stability of the predicted value cannot be guaranteed and the calculation load of the data cannot be reduced.
Disclosure of Invention
In view of the above, the present invention provides an information processing method, which solves the problem that in the prior art, only the diversity of data is considered, so that the stability of the predicted value cannot be guaranteed and the data calculation burden is reduced.
In order to achieve the above purpose, the present invention provides the following technical solutions:
an information processing method, comprising:
acquiring input data of at least two dimensions, wherein the input data of any dimension represents one feature of an object to be predicted;
and processing the input data according to a first processing model which is trained in advance to obtain a prediction result, and updating a first weight matrix in the first processing model based on the input data, wherein the first processing model comprises at least two groups of weight matrices.
Preferably, the method further comprises: pre-training a first process model;
the method specifically comprises the following steps:
acquiring sample input characteristics;
extracting the sample input features as at least one group of feature pairs as sample input data according to a preset data processing method;
constructing a first processing model, wherein the first processing model comprises at least T groups of weight matrixes, the weights in the memory matrixes are random assignment, and the T values are natural numbers larger than 1;
inputting the sample input data into the first processing model to obtain a prediction result and a prediction group weight matrix;
based on the fact that the predicted result is not matched with a preset result corresponding to the sample input data, adjusting the weight of a first processing model according to a preset adjusting mode and the preset result, and returning to the cyclic execution to input the sample input data into the first processing model;
and updating the first processing model according to a preset updating mode and the prediction group weight matrix based on the matching of the prediction result and a preset result corresponding to the sample input data.
Preferably, in the above method, the updating the first processing model according to a preset updating manner and the prediction group weight matrix includes:
Comparing the deviation in the at least T groups of weight matrixes and the deviation between the at least T groups of weight matrixes and the prediction group of weight matrixes in the first processing model to obtain a comparison result;
based on the comparison result, representing that the deviation between the at least T groups of weight matrixes and the prediction group of weight matrixes is smaller than the deviation in the at least T groups of weight matrixes, selecting a first weight matrix with the largest intra-group variance in the first processing model, and replacing the first weight matrix with the prediction group of weight matrixes;
according to a preset mean algorithm and a first processing model comprising a first weight matrix, calculating to obtain a second weight matrix;
and replacing the first weight matrix in the first processing model with the second weight matrix to obtain an updated first processing model.
Preferably, in the above method, the extracting the sample input data of the at least two dimensions as at least one set of feature pairs according to a preset data processing method includes:
extracting the sample input data by adopting a preset first rule to obtain at least two arrays, wherein any one array represents one characteristic of an object to be predicted;
And processing the at least two groups according to a preset second rule, and extracting at least one group meeting preset conditions from the at least two groups as sample input data.
Preferably, the method processes the input data according to a first processing model that is trained in advance, updates a first weight matrix in the first processing model based on the input data, and includes:
inputting the input data into the first processing model to obtain a third weight matrix;
and updating the first processing model according to a preset updating mode and the third weight matrix.
An electronic device, comprising:
a body;
the processor is used for acquiring input data of at least two dimensions, and the input data of any dimension characterizes one characteristic of an object to be predicted; and processing the input data according to a first processing model which is trained in advance to obtain a prediction result, and updating a first weight matrix in the first processing model based on the input data, wherein the first processing model comprises at least two groups of weight matrices.
Preferably, in the electronic device, the processor is further configured to: pre-training a first process model;
The method specifically comprises the following steps:
acquiring sample input characteristics;
extracting the sample input features as at least one group of feature pairs as sample input data according to a preset data processing method;
constructing a first processing model, wherein the first processing model comprises at least T groups of weight matrixes, the weights in the memory matrixes are random assignment, and the T values are natural numbers larger than 1;
inputting the sample input data into the first processing model to obtain a prediction result and a prediction group weight matrix;
based on the fact that the predicted result is not matched with a preset result corresponding to the sample input data, adjusting the weight of a first processing model according to a preset adjusting mode and the preset result, and returning to the cyclic execution to input the sample input data into the first processing model;
and updating the first processing model according to a preset updating mode and the prediction group weight matrix based on the matching of the prediction result and a preset result corresponding to the sample input data.
Preferably, in the electronic device, the processor is specifically configured to:
comparing the deviation between the at least T groups of weight matrixes and the prediction group of weight matrixes in the first processing model, and enabling the first weight matrix with the smallest deviation to be the same;
Based on the fact that the first weight matrix with the smallest deviation is the prediction group weight matrix, selecting a second group of matrix weight matrices with the largest intra-group variance in the first processing model, and replacing the first weight matrix with the first weight matrix;
according to a preset mean algorithm and a first processing model comprising a first weight matrix, calculating to obtain a second weight matrix;
and replacing the first weight matrix in the first processing model with the second weight matrix to obtain an updated first processing model.
Preferably, in the electronic device, the processor is specifically configured to:
extracting the sample input data by adopting a preset first rule to obtain at least two arrays, wherein any one array represents one characteristic of an object to be predicted;
and processing the at least two groups according to a preset second rule, and extracting at least one group meeting preset conditions from the at least two groups as sample input data.
An electronic device, comprising:
the system comprises an acquisition module, a prediction module and a prediction module, wherein the acquisition module is used for acquiring input data of at least two dimensions, and the input data of any dimension represents one feature of an object to be predicted;
the processing module is used for processing the input data according to a first processing model which is trained in advance to obtain a prediction result, and updating a first weight matrix in the first processing model based on the input data, wherein the first processing model comprises at least two groups of weight matrices.
As can be seen from the above technical solution, compared with the prior art, the present invention provides an information processing method, including: acquiring input data of at least two dimensions, wherein the input data of any dimension represents one feature of an object to be predicted; and processing the input data according to a first processing model which is trained in advance to obtain a prediction result, and updating a first weight matrix in the first processing model based on the input data, wherein the first processing model comprises at least two groups of weight matrices. In the scheme, in the process of processing the input data based on the first processing model, the first matrix of the input data is adjusted according to the received input data, and in the process of processing the data, the input data received each time is combined to adjust the input data, so that the residual matrix is kept unchanged. Therefore, when the first processing model is adopted for processing input data, factors of historical input data are considered, and factors of the input data are referred to, so that the data calculation burden is reduced under the condition that the stability of a predicted value is ensured.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are required to be used in the embodiments or the description of the prior art will be briefly described below, and it is obvious that the drawings in the following description are only embodiments of the present invention, and that other drawings can be obtained according to the provided drawings without inventive effort for a person skilled in the art.
Fig. 1 is a flowchart of an embodiment 1 of an information processing method provided in the present application;
FIG. 2 is a flowchart of an embodiment 2 of an information processing method provided in the present application;
FIG. 3 is a flowchart of training a first processing model in embodiment 2 of an information processing method provided in the present application;
fig. 4 is a schematic operation flow diagram of a first processing model in embodiment 2 of an information processing method provided in the present application;
FIG. 5 is a schematic diagram of a first processing model in embodiment 2 of an information processing method provided in the present application;
fig. 6 is a flowchart of an embodiment 3 of an information processing method provided in the present application;
fig. 7 is a schematic diagram of a processing flow in embodiment 3 of an information processing method provided in the present application;
FIG. 8 is a flowchart of training a first processing model in embodiment 4 of an information processing method provided in the present application;
fig. 9 is a flowchart of an embodiment 4 of an information processing method provided in the present application;
fig. 10 is a schematic structural diagram of an embodiment 1 of an electronic device provided in the present application;
fig. 11 is a schematic structural diagram of an embodiment 2 of an electronic device provided in the present application.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
As shown in fig. 1, a flowchart of an embodiment 1 of an information processing method provided in the present application is applied to an electronic device, where the electronic device has a first processing model, and the method includes the following steps:
step S101: acquiring input data of at least two dimensions;
wherein the input data of any dimension characterizes a feature of the object to be predicted;
the dimension specifically refers to different parameters, such as information related to various operations performed by a user.
In particular, the information related to the operation can be obtained on the network, such as various information of purchasing goods, such as price of shopping mall purchasing mobile phone/PC (personal computer ), number of shopping mall purchasing mobile phone/PC, color of shopping mall purchasing mobile phone/PC, etc.; and information of the used equipment, such as the number of Applications (APP) used by the mobile phone terminal, APP class used by the mobile phone terminal, and the like.
It should be noted that, the input parameters of the multiple dimensions characterize the preference of the user, and correspondingly, based on the input parameters of the multiple dimensions, the prediction of the user can be realized.
In particular, the predictions may include various aspects of the user's preferences, gender, etc.
Step S102: and processing the input data according to a first processing model which is trained in advance to obtain a prediction result, and updating a first weight matrix in the first processing model based on the input data.
Wherein the first processing model comprises at least two groups of weight matrixes.
The first processing model may be provided with a rule related to a prediction object, where the preset object specifically refers to content to be predicted of the user, such as preference, gender, and the like of the user, but is not limited thereto.
The first processing model is trained in advance, can process input data, and obtains a prediction result.
In particular implementations, the prediction may be used as a tag for a predicted user, such as predicting the gender, preferences, etc. of the user to facilitate processing of the user in other ways based on the user's tag at a later time.
For example, merchandise, applications, etc. are recommended for the gender or preference of the user.
And, during the processing of the input data by the first processing model, the input data is also updated, such as updating part of the weight matrix, based on the input data.
Specifically, the first processing model includes a plurality of groups of weight matrices, and in the process of processing the input data to obtain the prediction result, a part of weight matrices in the first processing model are updated, while the weight matrices of the rest part are kept unchanged.
It should be noted that, because the first processing model processes the input data, the prediction result obtained based on the input data obtains the prediction result according to the content of the current prediction; in the process of processing the input data, the partial weight matrix in the first processing model is updated, the weight matrix of the rest part is kept unchanged, the partial weight matrix is updated and adjusted by referring to the input data received each time, and the previous partial weight matrix is kept, so that factors of historical input data are considered, factors of the input data are also referred to, the stability of a predicted value is ensured due to the fact that the considered input data penetrates through the whole time span, and only the partial weight matrix in the first processing model is adjusted each time when the input data is processed, and the data calculation load is reduced.
In this application, the prediction result obtained by processing the input data according to the first processing model is related to the setting of the first processing model, and different prediction results may set the first processing model of different processing rules.
In summary, the information processing method provided in this embodiment includes: acquiring input data of at least two dimensions, wherein the input data of any dimension represents one feature of an object to be predicted; and processing the input data according to a first processing model which is trained in advance to obtain a prediction result, and updating a first weight matrix in the first processing model based on the input data, wherein the first processing model comprises at least two groups of weight matrices. In the scheme, in the process of processing the input features based on the first processing model, the first matrix of the input features is also adjusted according to the received input data, and in the data processing process, the input features received each time are combined to adjust the input features, so that the remaining matrix is kept unchanged. Therefore, when the first processing model is adopted for processing input data, factors of historical input data are considered, and factors of the input data are referred to, so that the data calculation burden is reduced under the condition that the stability of a predicted value is ensured.
As shown in fig. 2, a flowchart of an embodiment 2 of an information processing method provided in the present application includes the following steps:
step S201: pre-training a first process model;
it should be noted that, in this embodiment, explanation is mainly made with respect to the process of training the first processing model.
In particular, the first process model may be pre-trained based on known user tags and their corresponding user characteristics.
The training process may refer to the following process.
A flowchart for training the first process model is shown in fig. 3.
Wherein, the process comprises the following steps:
step S301: acquiring sample input characteristics;
the sample input features correspond to preset user labels, the user can be defined as a training user, and the input features and the labels of the user are known.
Specifically, the sample input feature includes a plurality of dimensions, one dimension corresponding to each feature of the training user.
The dimension specifically refers to different parameters, such as information related to various operations performed by a training user.
In specific implementation, the operation related information refers to operation related information of the training user, which can be acquired on a network, for example, various information of purchasing goods, such as price of a shopping mall purchasing mobile phone/PC, number of the shopping mall purchasing mobile phone/PC, color of the shopping mall purchasing mobile phone/PC, and the like; and information of the used equipment, such as the number of APP used by the mobile phone end, APP class used by the mobile phone end and the like.
It should be noted that, the input parameters of the multiple dimensions characterize the preference of the training user, and correspondingly, based on the input parameters of the multiple dimensions, the prediction of the training user can be realized.
In particular, the predictions may include various aspects of the user's preferences, gender, etc.
In specific implementation, the first processing model can be trained aiming at a plurality of groups of input parameters with multiple dimensions, and one training of the first processing model can be realized by a plurality of groups of input parameters with multiple dimensions.
Step S302: extracting the sample input features as at least one group of feature pairs as sample input data according to a preset data processing method;
the method comprises the steps of presetting a data processing method, and processing and extracting at least part of characteristics of the sample input characteristics based on the data processing method to serve as sample input data.
Specifically, in the scheme, a large number of sample input features are processed, part of the features are extracted as sample input data, part of the features are removed, and the speed of training the first processing model is improved.
It should be noted that, the following embodiments will be described in detail with respect to the extraction process, which is not described in detail in this embodiment.
In particular, the sample input data may be divided into two groups, one group for training the first process model and the other group for testing the training results of the first process model.
For example, the ratio of the sample input data for training and testing may be 8:2, although the ratio may be other, and the specific value of the ratio is not limited in the application.
Step S303: constructing a first processing model;
the first processing model comprises at least T groups of weight matrixes, the weights in the memory matrixes are randomly assigned, and the T values are natural numbers larger than 1.
The weight matrix is set according to actual conditions, and the larger the number of groups of the weight matrix is, the higher the precision value is.
Specifically, the initial value of the weight matrix in the first processing model is a random assignment.
In specific implementation, when the first processing model is built, firstly, a weight matrix in the first processing model is initialized, and random assignment is carried out on the weight matrix.
It should be noted that, the process of obtaining the sample input data in the steps S301 to 302 and the process of constructing the first processing model in the step S303 are not limited in this application, and the sequence of the two processes may be interchanged or may be performed simultaneously.
Step S304: inputting the sample input data into the first processing model to obtain a prediction result and a prediction group weight matrix;
the sample input data is input into the first processing model, and the first processing model is used for processing based on the weight matrix in the first processing model to obtain a prediction result and a weight matrix of a prediction group.
The sample input data is input into the first processing model, and the data dimension of each weight matrix in the first processing model corresponds to the dimension of the sample input data.
Wherein the dimension correspondingly comprises the same and different, and if the data dimension of the weight matrix is different from the sample input data, zero is added in the weight matrix of the obtained prediction group.
In a specific implementation, the predicted result obtained by calculation may be represented by a numerical value, if the numerical value of the predicted result and a preset value (preset result) meet a preset proximity condition, the predicted result is matched with the preset result, otherwise, the predicted result and the preset result are not matched.
Step S305: based on the fact that the predicted result is not matched with a preset result corresponding to the sample input data, adjusting the weight of a first processing model according to a preset adjusting mode and the preset result, and returning to the cyclic execution to input the sample input data into the first processing model;
And when the predicted result is not matched with the preset result, adjusting the weight of the weight matrix in the first processing model.
Specifically, a random adjustment manner may be adopted, so that the predicted result approaches the preset result.
In specific implementation, the weight in the weight matrix can be increased, prediction can be performed, if the deviation between the predicted result and the preset result is larger, the weight in the weight matrix is decreased, and prediction is performed again, so that the predicted result approaches the preset result in a gradually attenuated approach process until the predicted result is matched with the preset result.
Step S306: and updating the first processing model according to a preset updating mode and the prediction group weight matrix based on the matching of the prediction result and a preset result corresponding to the sample input data.
And updating a set of weight matrix in the first processing model based on the predicted set of weight matrix obtained according to the sample input data when the predicted result is matched with the preset result.
It should be noted that, each time the first processing model predicts, a prediction set of weight matrices is generated, and a set of weight matrices in the first processing model is updated based on the prediction set of weight matrices.
The updating process will be described in detail in the following embodiments, which will not be described in detail in this embodiment.
After the training of the first processing model is completed, the process of processing the input data to obtain the prediction result and the process of updating the first weight matrix in the first processing model may refer to the training process.
FIG. 4 is a schematic diagram of the operation flow of the first process model, which includes:
step S401: initializing data;
wherein, a first processing model is constructed and the weight value is initialized.
Step S402: using batch data for prediction;
and processing the first processing model by adopting a plurality of groups of sample input data respectively, so that the sample input data is predicted.
Step S403: adjusting the weight value to make the preset value and the predicted value consistent;
and adjusting the weight value based on the predicted value obtained by prediction and the preset value so that the predicted value is consistent with the preset value.
Step S404: obtaining a weight value;
and if the predicted value is consistent with the preset value, the weight value of the weight matrix obtained by the sample input data is the optimal weight value corresponding to the sample input data.
Step S405: and (5) weight feedback.
And feeding the obtained optimal weight value corresponding to the sample input data back to the first processing model, updating the first processing model, and completing training, so that the first processing model which is completed based on the training processes the input data, and the prediction of the user is realized, and the label of the user is obtained.
FIG. 5 is a schematic diagram of the first process model, wherein X1-4 represents sample input data and H1-4 represents the first process model.
At time T1, the sample input data X1 is input into the first processing model H1, and updated; at the time of T2, the sample input data X2 is input into the first processing model H2, and the first processing model H2 is updated; at time T3, the sample input data X3 is input into the first processing model H3, and is updated; at time T4, the sample input data X4 is input to the first process model H4, updated, and so on.
Step S202: acquiring input data of at least two dimensions;
step S203: and processing the input data according to a first processing model which is trained in advance to obtain a prediction result, and updating a first weight matrix in the first processing model based on the input data, wherein the first processing model comprises at least two groups of weight matrices.
Steps S202-203 are identical to steps S101-102 in embodiment 1, and are not described in detail in this embodiment.
In summary, the information processing method provided in this embodiment further includes pre-training a first processing model, which specifically includes: acquiring sample input characteristics; extracting the sample input features as at least one group of feature pairs as sample input data according to a preset data processing method; constructing a first processing model, wherein the first processing model comprises at least T groups of weight matrixes, the weights in the memory matrixes are random assignment, and the T values are natural numbers larger than 1; inputting the sample input data into the first processing model to obtain a prediction result and a prediction group weight matrix; based on the fact that the predicted result is not matched with a preset result corresponding to the sample input data, adjusting the weight of a first processing model according to a preset adjusting mode and the preset result, and returning to the cyclic execution to input the sample input data into the first processing model; and updating the first processing model according to a preset updating mode and the prediction group weight matrix based on the matching of the prediction result and a preset result corresponding to the sample input data. In the scheme, sample input data is processed based on a first processing model to obtain a prediction result and a prediction group weight matrix, the first processing model is updated according to the prediction group weight matrix based on the matching of the prediction result and a preset result; and if the two are not matched, adjusting the weight in the first processing model and processing the sample input data again until the predicted result is matched with the preset result. The process realizes the training of the first processing model so as to ensure that the first processing model after the training processes the input data to obtain a prediction result, and the prediction result is stable, thereby ensuring the label stability of a predicted user.
As shown in fig. 6, a flowchart for training a first processing model in embodiment 3 of an information processing method provided in the present application includes the following steps:
step S601: acquiring sample input characteristics;
step S602: extracting the sample input features as at least one group of feature pairs as sample input data according to a preset data processing method;
step S603: constructing a first processing model;
step S604: inputting the sample input data into the first processing model to obtain a prediction result and a prediction group weight matrix;
step S605: based on the fact that the predicted result is not matched with a preset result corresponding to the sample input data, adjusting the weight of a first processing model according to a preset adjusting mode and the preset result, and returning to the cyclic execution to input the sample input data into the first processing model;
steps S601-605 are identical to steps S301-305 in embodiment 2, and detailed description is omitted in this embodiment.
Step S606: based on the matching of the prediction result and a preset result corresponding to the sample input data, comparing the deviation in the at least T groups of weight matrixes in the first processing model with the deviation between the at least T groups of weight matrixes and the prediction group of weight matrixes to obtain a comparison result;
And if the predicted result is matched with the preset result, the accuracy of the processing result of the version input data of the first processing model is higher.
Specifically, the deviation is calculated in the weight matrix set formed by the T group weight matrix and the prediction group weight matrix, specifically, the deviation is made in the T group weight matrix and the prediction value weight matrix respectively, and the obtained deviation is compared.
Specifically, the comparison result may include one of the following: 1. the deviation in the T group weight matrix is larger than the deviation between the T group weight matrix and the predicted value weight matrix respectively; 2. the deviation in the T group weight matrix is partially larger than the deviation between the T group weight matrix and the predicted value weight matrix respectively; 3. the deviation in the T groups of weight matrixes is smaller than the deviation between the T groups of weight matrixes and the predicted value weight matrixes respectively.
Step S607: based on the comparison result, representing that the deviation between the at least T groups of weight matrixes and the prediction group of weight matrixes is smaller than the deviation in the at least T groups of weight matrixes, selecting a first weight matrix with the largest variance in the first processing model, and replacing the first weight matrix with the prediction group of weight matrixes;
Specifically, based on that the deviation between the at least T sets of weight matrixes and the predicted set of weight matrixes is smaller than the deviation in the at least T sets of weight matrixes, the variance of each weight matrix in the first processing model is calculated, the variances of the weight matrixes are compared to obtain a weight matrix with the largest variance, the weight matrix is recorded as a first weight matrix, and the predicted set of weight matrixes replace the first weight matrix.
Step S608: calculating to obtain a second weight matrix according to a preset mean algorithm and a first processing model containing a predicted group weight matrix;
and calculating to obtain a second weight matrix according to the first processing model containing the predicted group weight matrix.
Specifically, the weight of the predicted group weight matrix is increased (for example, the weight value in the predicted group weight matrix is increased by an equal amount/equal ratio), and a weight average value is calculated based on the T group weight matrix in the first processing model containing the predicted group weight matrix, so as to obtain a second weight matrix.
For example, the first processing model has 7 sets of weight matrices, after replacing one set of weight matrices with the predicted set of weight matrices, increasing the weight value of the predicted set of weight matrices, and then performing mean calculation on the rest 6 sets of weight matrices and the predicted set of weight matrices to obtain a set of weight matrices, which are marked as second weight matrices.
Step S609: and replacing the first weight matrix in the first processing model with the second weight matrix to obtain an updated first processing model.
Specifically, the second weight matrix is used for replacing the first weight matrix in the first processing model, and the second weight matrix related to the preset weight matrix is used for replacing the first weight matrix due to the maximum variance of the first weight matrix, so that the first processing model is updated.
Fig. 7 is a schematic view of a process flow provided in this embodiment, where the process flow includes a first process model 001, and the first process model includes: core matrix, forget gate and select gate.
Specifically, the core matrix is a T-group weight matrix provided in this embodiment.
Specifically, the selection gate is configured to compare the deviation between the original weight matrix and the predicted group weight matrix in the first processing model, so as to select a weight matrix with small deviation.
Specifically, the forgetting gate is used for calculating the variance of each weight matrix in the core matrix, removing the matrix (the first weight matrix) with the maximum variance, improving the weight coefficient of the weight matrix screened by the selection gate, then solving the weight average value of the rest matrix in the core matrix to generate a new weight matrix, and updating the memory matrix.
Specifically, the process comprises the following steps:
step S701: initializing a core matrix and node information;
wherein, a random number mode is adopted to initialize the core matrix.
Specifically, the node information refers to an instance that the first processing model runs on a specific machine device, and corresponds to a master (master device) and a slave (slave device) in the large-scale cluster software, and the node refers to all the slave devices running the model.
Step S702: using the sample data to predict;
step S703: obtaining a weight value;
step S704: iterative processing, namely adjusting weight parameters of the core matrix according to a preset value;
specifically, until the predicted value is consistent with the preset value.
Step S705: the selection gate and the forget gate process the core matrix based on the weight value;
the obtained core matrix is fed back to step S702.
In summary, in the information processing method provided in this embodiment, updating the first processing model according to the preset updating manner and the prediction group weight matrix includes: comparing the deviation in the at least T groups of weight matrixes and the deviation between the at least T groups of weight matrixes and the prediction group of weight matrixes in the first processing model to obtain a comparison result; based on the comparison result, representing that the deviation between the at least T groups of weight matrixes and the prediction group of weight matrixes is smaller than the deviation in the at least T groups of weight matrixes, selecting a first weight matrix with the largest intra-group variance in the first processing model, and replacing the first weight matrix with the prediction group of weight matrixes; according to a preset mean algorithm and a first processing model comprising a first weight matrix, calculating to obtain a second weight matrix; and replacing the first weight matrix in the first processing model with the second weight matrix to obtain an updated first processing model. According to the method, through mathematical calculation, the first weight matrix in the first processing model is replaced by the second weight matrix related to the predicted group weight matrix, so that the first processing model is updated for each training, historical sample input data is considered, the sample input data at the time is considered, and the data calculation load is reduced under the condition that the stability of a predicted value is ensured.
As shown in fig. 8, a flowchart for training a first processing model in embodiment 4 of an information processing method provided in the present application includes the following steps:
step S801: acquiring sample input characteristics;
step S801 is identical to step S301 in embodiment 2, and is not described in detail in this embodiment.
Step S802: extracting the sample input data by adopting a preset first rule to obtain at least two arrays, wherein any one array represents one characteristic of an object to be predicted;
in specific implementation, word2vec (word vector) and one-hot (one-hot encoding) are used to process the sample input features, so as to extract the sample input features into a multidimensional array.
Step S803: processing the at least two groups according to a preset second rule, and extracting at least one group meeting preset conditions from the at least two groups as sample input data;
specifically, the Bayesian algorithm is adopted to combine the extracted different feature data in the form of multi-dimensional array two by two, the prediction is carried out on the user based on the feature of the combination of the two by two, the initial prediction result is obtained, a plurality of groups of features with the highest preset user tag matching precision corresponding to the sample input data of the initial prediction result are used as sample input features, and the feature data corresponding to the sample input features are recorded as sample input data.
In this step, the sample input features are processed, and data corresponding to several groups of features with the highest matching precision of the user tag are extracted from the sample input features and are input into the subsequent first processing model as sample input data, so that features with lower precision are removed, and the speed of training the first processing model is improved.
Step S804: constructing a first processing model;
step S805: inputting the sample input data into the first processing model to obtain a prediction result and a prediction group weight matrix;
step S806: based on the fact that the predicted result is not matched with a preset result corresponding to the sample input data, adjusting the weight of a first processing model according to a preset adjusting mode and the preset result, and returning to the cyclic execution to input the sample input data into the first processing model;
step S807: and updating the first processing model according to a preset updating mode and the prediction group weight matrix based on the matching of the prediction result and a preset result corresponding to the sample input data.
Steps S804-807 are identical to steps S304-306 in embodiment 2, and are not described in detail in this embodiment.
In summary, in the information processing method provided in the present embodiment, according to a preset data processing method, the extracting the sample input data of the at least two dimensions as at least one set of feature pairs as the sample input data includes: extracting the sample input data by adopting a preset first rule to obtain at least two arrays, wherein any one array represents one characteristic of an object to be predicted; and processing the at least two groups according to a preset second rule, and extracting at least one group meeting preset conditions from the at least two groups as sample input data. By adopting the method, a large amount of sample input data is processed through calculation rules, so that the sample input data with higher correlation with the training preset result is obtained, the characteristic with lower precision is eliminated, and the speed of training the first processing model is improved.
As shown in fig. 9, a flowchart of an embodiment 4 of an information processing method provided in the present application includes the following steps:
step S901: acquiring input data of at least two dimensions;
step S902: processing the input data according to a first processing model which is trained in advance, so as to obtain a prediction result;
steps S901-902 are identical to steps S101-102 in embodiment 1, and are not described in detail in this embodiment.
Step S903: inputting the input data into the first processing model to obtain a third weight matrix;
the input data is input into the first processing model, and a third weight matrix can be obtained, wherein the third weight matrix is related to the input data.
Step S904: and updating the first processing model according to a preset updating mode and the third weight matrix.
Specifically, the first processing model is updated based on the third weight matrix, so as to replace one group of weight matrix with the third weight matrix.
The updating method can refer to the method in embodiments 2-3, and details thereof are not described in this embodiment.
It should be noted that, in this embodiment, the steps S903-904 are only used to explain the process of updating the first processing model, and in a specific implementation, the process may be performed simultaneously with the process of obtaining the prediction result in the step S902, and in this embodiment, the execution sequence of the two is not limited.
In summary, in the information processing method provided in this embodiment, the processing the input data according to the first processing model that is trained in advance, and updating the first weight matrix in the first processing model based on the input data includes: the input data is input into the first processing model to obtain a third weight matrix; and updating the first processing model according to a preset updating mode and the third weight matrix. By adopting the method, when a prediction result is obtained based on input data, a third weight matrix is obtained, and the first processing model is updated based on the third weight matrix, so that the first processing model is updated for each input, historical sample input data is considered, the sample input data of the time is also considered, and the data calculation load is reduced under the condition that the stability of the prediction value is ensured.
Corresponding to the embodiment of the information processing method provided by the application, the application also provides an embodiment of the electronic equipment applying the information processing method.
As shown in fig. 10, a schematic structural diagram of an embodiment 1 of an electronic device provided in the present application, where the electronic device includes the following structures: an ontology 1001 and a processor 1002;
The processor 1002 is configured to obtain input data of at least two dimensions, where the input data of any dimension characterizes a feature of an object to be predicted; and processing the input data according to a first processing model which is trained in advance to obtain a prediction result, and updating a first weight matrix in the first processing model based on the input data, wherein the first processing model comprises at least two groups of weight matrices.
In particular, the processor is configured to have information processing capability, such as a CPU (central processing unit ) or the like.
Preferably, the processor is further configured to: pre-training a first process model;
the method specifically comprises the following steps:
acquiring sample input characteristics;
extracting the sample input features as at least one group of feature pairs as sample input data according to a preset data processing method;
constructing a first processing model, wherein the first processing model comprises at least T groups of weight matrixes, the weights in the memory matrixes are random assignment, and the T values are natural numbers larger than 1;
acquiring sample input data;
inputting the sample input data into the first processing model to obtain a prediction result and a prediction group weight matrix;
Based on the fact that the predicted result is not matched with a preset result corresponding to the sample input data, adjusting the weight of a first processing model according to a preset adjusting mode and the preset result, and returning to the cyclic execution to input the sample input data into the first processing model;
and updating the first processing model according to a preset updating mode and the prediction group weight matrix based on the matching of the prediction result and a preset result corresponding to the sample input data.
Preferably, the processor is specifically configured to:
comparing the deviation between the at least T groups of weight matrixes and the prediction group of weight matrixes in the first processing model, and enabling the first weight matrix with the smallest deviation to be the same;
based on the fact that the first weight matrix with the smallest deviation is the prediction group weight matrix, selecting a second group of matrix weight matrices with the largest intra-group variance in the first processing model, and replacing the first weight matrix with the first weight matrix;
according to a preset mean algorithm and a first processing model comprising a first weight matrix, calculating to obtain a second weight matrix;
and replacing the first weight matrix in the first processing model with the second weight matrix to obtain an updated first processing model.
Preferably, the processor is specifically configured to:
comprising the following steps:
extracting the sample input data by adopting a preset first rule to obtain at least two arrays, wherein any one array represents one characteristic of an object to be predicted;
and processing the at least two groups according to a preset second rule, and extracting at least one group meeting preset conditions from the at least two groups as sample input data.
Preferably, the processor is specifically configured to:
inputting the input data into the first processing model to obtain a prediction group weight matrix;
and updating the first processing model according to a preset updating mode and the prediction group weight matrix.
In summary, in the electronic device provided in this embodiment, during the processing of the input features based on the first processing model, the first matrix of the electronic device is also adjusted according to the received input data, and during the data processing, the electronic device is adjusted according to the input features received each time, so that the remaining matrix is kept unchanged. Therefore, when the first processing model is adopted for processing input data, factors of historical input data are considered, and factors of the input data are referred to, so that the data calculation burden is reduced under the condition that the stability of a predicted value is ensured.
As shown in fig. 11, a schematic structural diagram of an embodiment 2 of an electronic device provided in the present application, where the electronic device includes the following structures: an acquisition module 1101 and a processing module 1102;
the obtaining module 1101 is configured to obtain input data of at least two dimensions, where the input data of any dimension characterizes a feature of an object to be predicted;
the processing module 1102 is configured to process the input data according to a first processing model that is trained in advance to obtain a prediction result, and update a first weight matrix in the first processing model based on the input data, where the first processing model includes at least two sets of weight matrices.
Preferably, the method further comprises: the training module is used for training the first processing model in advance;
the training module specifically comprises:
the acquisition unit is used for acquiring sample input characteristics;
the extraction unit is used for extracting the sample input features into at least one group of feature pairs as sample input data according to a preset data processing method;
the building unit is used for building a first processing model, the first processing model comprises at least T groups of weight matrixes, the weights in the memory matrixes are random assignment, and the T value is a natural number larger than 1;
The prediction unit is used for inputting the sample input data into the first processing model to obtain a prediction result and a prediction group weight matrix;
the adjusting unit is used for adjusting the weight of the first processing model according to a preset adjusting mode and a preset result based on the fact that the prediction result is not matched with the preset result corresponding to the sample input data, and returning to the cyclic execution to input the sample input data into the first processing model;
and the updating unit is used for updating the first processing model according to a preset updating mode and the prediction group weight matrix based on the fact that the prediction result is matched with a preset result corresponding to the sample input data.
Preferably, the updating unit is specifically configured to:
comparing the deviation between the at least T groups of weight matrixes and the prediction group of weight matrixes in the first processing model, and enabling the first weight matrix with the smallest deviation to be the same;
based on the fact that the first weight matrix with the smallest deviation is the prediction group weight matrix, selecting a second group of matrix weight matrices with the largest intra-group variance in the first processing model, and replacing the first weight matrix with the first weight matrix;
according to a preset mean algorithm and a first processing model comprising a first weight matrix, calculating to obtain a second weight matrix;
And replacing the first weight matrix in the first processing model with the second weight matrix to obtain an updated first processing model.
Preferably, the extraction unit is specifically configured to:
extracting the sample input data by adopting a preset first rule to obtain at least two arrays, wherein any one array represents one characteristic of an object to be predicted;
and processing the at least two groups according to a preset second rule, and extracting at least one group meeting preset conditions from the at least two groups as sample input data.
Preferably, the processing module is specifically configured to:
inputting the input data into the first processing model to obtain a prediction group weight matrix;
and updating the first processing model according to a preset updating mode and the prediction group weight matrix.
In summary, in the electronic device provided in this embodiment, during the processing of the input features based on the first processing model, the first matrix of the electronic device is also adjusted according to the received input data, and during the data processing, the electronic device is adjusted according to the input features received each time, so that the remaining matrix is kept unchanged. Therefore, when the first processing model is adopted for processing input data, factors of historical input data are considered, and factors of the input data are referred to, so that the data calculation burden is reduced under the condition that the stability of a predicted value is ensured.
In the present specification, each embodiment is described in a progressive manner, and each embodiment is mainly described in a different point from other embodiments, and identical and similar parts between the embodiments are all enough to refer to each other. The device provided in the embodiment corresponds to the method provided in the embodiment, so that the description is simpler, and the relevant points refer to the description of the method.
The previous description of the provided embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features provided herein.
Claims (8)
1. An information processing method, comprising:
acquiring input data of at least two dimensions, wherein the input data of any dimension represents one characteristic of an object to be predicted; the input data of at least two dimensions characterizes the preference of the user so as to realize the prediction of the user based on the input data of at least two dimensions;
Processing the input data according to a first processing model which is trained in advance to obtain a prediction result, and updating a first weight matrix in the first processing model based on the input data, wherein the first processing model comprises at least two groups of weight matrices, and other weight matrices except the first weight matrix are kept unchanged; the first processing model is provided with rules related to a prediction object, wherein the prediction object refers to the content to be predicted of a user and comprises the preference and the gender of the user;
the prediction result is used as a label of a predicted user, so that commodity and application recommendation is performed on the predicted user based on the label of the predicted user;
wherein: pre-training a first process model;
the method specifically comprises the following steps:
acquiring sample input characteristics;
extracting sample input features as at least one group of feature pairs as sample input data according to a preset data processing method;
constructing a first processing model, wherein the first processing model comprises at least T groups of weight matrixes, the weights in the weight matrixes are random assignment, and the T values are natural numbers larger than 1;
inputting the sample input data into the first processing model to obtain a prediction result and a prediction group weight matrix;
Based on the fact that the predicted result is not matched with a preset result corresponding to the sample input data, adjusting the weight of a first processing model according to a preset adjusting mode and the preset result, and returning to the cyclic execution to input the sample input data into the first processing model;
and updating the first processing model according to a preset updating mode and the prediction group weight matrix based on the matching of the prediction result and a preset result corresponding to the sample input data.
2. The method of claim 1, wherein updating the first processing model according to the preset updating manner and the prediction group weight matrix comprises:
comparing the deviation in the at least T groups of weight matrixes and the deviation between the at least T groups of weight matrixes and the prediction group of weight matrixes in the first processing model to obtain a comparison result;
based on the comparison result, representing that the deviation between the at least T groups of weight matrixes and the prediction group of weight matrixes is smaller than the deviation in the at least T groups of weight matrixes, selecting a first weight matrix with the largest intra-group variance in the first processing model, and replacing the first weight matrix with the prediction group of weight matrixes;
According to a preset mean algorithm and a first processing model comprising a first weight matrix, calculating to obtain a second weight matrix;
and replacing the first weight matrix in the first processing model with the second weight matrix to obtain an updated first processing model.
3. The method according to claim 1, wherein the extracting the sample input data of the at least two dimensions as at least one set of feature pairs according to a preset data processing method includes:
extracting the sample input data by adopting a preset first rule to obtain at least two arrays, wherein any one array represents one characteristic of an object to be predicted;
and processing the at least two groups according to a preset second rule, and extracting at least one group meeting preset conditions from the at least two groups as sample input data.
4. The method of claim 1, processing the input data according to a pre-trained first processing model, updating a first weight matrix in the first processing model based on the input data, comprising:
inputting the input data into the first processing model to obtain a third weight matrix;
And updating the first processing model according to a preset updating mode and the third weight matrix.
5. An electronic device, comprising:
a body;
the processor is used for acquiring input data of at least two dimensions, and the input data of any dimension characterizes one feature of an object to be predicted; the input data of at least two dimensions characterizes the preference of the user so as to realize the prediction of the user based on the input data of at least two dimensions; processing the input data according to a first processing model which is trained in advance to obtain a prediction result, and updating a first weight matrix in the first processing model based on the input data, wherein the first processing model comprises at least two groups of weight matrices, and other weight matrices except the first weight matrix are kept unchanged; the first processing model is provided with rules related to a prediction object, wherein the prediction object refers to the content to be predicted of a user and comprises the preference and the gender of the user; the prediction result is used as a label of a predicted user, so that commodity and application recommendation is performed on the predicted user based on the label of the predicted user;
wherein, the liquid crystal display device comprises a liquid crystal display device,
The processor is further configured to: pre-training a first process model;
the method specifically comprises the following steps:
acquiring sample input characteristics;
extracting sample input features as at least one group of feature pairs as sample input data according to a preset data processing method;
constructing a first processing model, wherein the first processing model comprises at least T groups of weight matrixes, the weights in the weight matrixes are random assignment, and the T values are natural numbers larger than 1;
inputting the sample input data into the first processing model to obtain a prediction result and a prediction group weight matrix;
based on the fact that the predicted result is not matched with a preset result corresponding to the sample input data, adjusting the weight of a first processing model according to a preset adjusting mode and the preset result, and returning to the cyclic execution to input the sample input data into the first processing model;
and updating the first processing model according to a preset updating mode and the prediction group weight matrix based on the matching of the prediction result and a preset result corresponding to the sample input data.
6. The electronic device of claim 5, the processor being specifically configured to:
comparing the deviation between the at least T groups of weight matrixes and the prediction group of weight matrixes in the first processing model, and enabling the first weight matrix with the smallest deviation to be the same;
Based on the fact that the first weight matrix with the smallest deviation is the prediction group weight matrix, selecting a second group of matrix weight matrices with the largest intra-group variance in the first processing model, and replacing the first weight matrix with the first weight matrix;
according to a preset mean algorithm and a first processing model comprising a first weight matrix, calculating to obtain a second weight matrix;
and replacing the first weight matrix in the first processing model with the second weight matrix to obtain an updated first processing model.
7. The electronic device of claim 5, the processor being specifically configured to:
extracting the sample input data by adopting a preset first rule to obtain at least two arrays, wherein any one array represents one characteristic of an object to be predicted;
and processing the at least two groups according to a preset second rule, and extracting at least one group meeting preset conditions from the at least two groups as sample input data.
8. An electronic device, comprising:
the acquisition module is used for acquiring input data of at least two dimensions, and the input data of any dimension represents one characteristic of an object to be predicted; the input data of at least two dimensions characterizes the preference of the user so as to realize the prediction of the user based on the input data of at least two dimensions; the processing module is used for processing the input data according to a first processing model which is trained in advance to obtain a prediction result, updating a first weight matrix in the first processing model based on the input data, wherein the first processing model comprises at least two groups of weight matrices, and other weight matrices except the first weight matrix are kept unchanged; the first processing model is provided with rules related to a prediction object, wherein the prediction object refers to the content to be predicted of a user and comprises the preference and the gender of the user; the prediction result is used as a label of a predicted user, so that commodity and application recommendation is performed on the predicted user based on the label of the predicted user;
Wherein: pre-training a first process model;
the method specifically comprises the following steps:
acquiring sample input characteristics;
extracting sample input features as at least one group of feature pairs as sample input data according to a preset data processing method;
constructing a first processing model, wherein the first processing model comprises at least T groups of weight matrixes, the weights in the weight matrixes are random assignment, and the T values are natural numbers larger than 1;
inputting the sample input data into the first processing model to obtain a prediction result and a prediction group weight matrix;
based on the fact that the predicted result is not matched with a preset result corresponding to the sample input data, adjusting the weight of a first processing model according to a preset adjusting mode and the preset result, and returning to the cyclic execution to input the sample input data into the first processing model;
and updating the first processing model according to a preset updating mode and the prediction group weight matrix based on the matching of the prediction result and a preset result corresponding to the sample input data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811475767.4A CN109584020B (en) | 2018-12-04 | 2018-12-04 | Information processing method and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811475767.4A CN109584020B (en) | 2018-12-04 | 2018-12-04 | Information processing method and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109584020A CN109584020A (en) | 2019-04-05 |
| CN109584020B true CN109584020B (en) | 2023-07-21 |
Family
ID=65927422
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811475767.4A Active CN109584020B (en) | 2018-12-04 | 2018-12-04 | Information processing method and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109584020B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111340276B (en) * | 2020-02-19 | 2022-08-19 | 联想(北京)有限公司 | Method and system for generating prediction data |
| CN112395280B (en) * | 2021-01-19 | 2021-04-23 | 睿至科技集团有限公司 | Data quality detection method and system |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2010073170A (en) * | 2008-09-22 | 2010-04-02 | Yahoo Japan Corp | Recommended commodity selection device, recommended commodity selection program and commodity retrieval device |
| CN106022865A (en) * | 2016-05-10 | 2016-10-12 | 江苏大学 | Goods recommendation method based on scores and user behaviors |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102346899A (en) * | 2011-10-08 | 2012-02-08 | 亿赞普(北京)科技有限公司 | Method and device for predicting advertisement click rate based on user behaviors |
| US20170278114A1 (en) * | 2016-03-24 | 2017-09-28 | Xerox Corporation | Adaptive collaborative filtering with extended kalman filters and multi-armed bandits |
| CN106557846A (en) * | 2016-11-30 | 2017-04-05 | 成都寻道科技有限公司 | Based on university students school data graduation whereabouts Forecasting Methodology |
| CN108305094B (en) * | 2017-12-18 | 2021-07-23 | 北京三快在线科技有限公司 | User behavior prediction method and device and electronic equipment |
| CN108804266A (en) * | 2018-05-22 | 2018-11-13 | 郑州云海信息技术有限公司 | A kind of performance of storage system test method, device and computer readable storage medium |
-
2018
- 2018-12-04 CN CN201811475767.4A patent/CN109584020B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2010073170A (en) * | 2008-09-22 | 2010-04-02 | Yahoo Japan Corp | Recommended commodity selection device, recommended commodity selection program and commodity retrieval device |
| CN106022865A (en) * | 2016-05-10 | 2016-10-12 | 江苏大学 | Goods recommendation method based on scores and user behaviors |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109584020A (en) | 2019-04-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108090093B (en) | Method and device for generating recommendation result | |
| WO2019114413A1 (en) | Model training | |
| CN109754105B (en) | Prediction method, terminal and server | |
| US11494639B2 (en) | Bayesian-optimization-based query-efficient black-box adversarial attacks | |
| CN108496189A (en) | The regularization of machine learning model | |
| Eckles et al. | Thompson sampling with the online bootstrap | |
| CN110838020A (en) | Recommendation method and device based on vector migration, computer equipment and storage medium | |
| CN108304354B (en) | Prediction model training method and device, storage medium and electronic equipment | |
| CN113377964B (en) | Knowledge graph link prediction method, device, equipment and storage medium | |
| CN109584020B (en) | Information processing method and electronic equipment | |
| CN109117442B (en) | Application recommendation method and device | |
| US20220147540A1 (en) | Personalized visualization recommendation system | |
| WO2016132588A1 (en) | Data analysis device, data analysis method, and data analysis program | |
| US20200250715A1 (en) | Automatic rule generation for recommendation engine using hybrid machine learning | |
| Gu et al. | Variational mixtures of ODEs for inferring cellular gene expression dynamics | |
| CN113779380B (en) | Cross-domain recommendation and content recommendation methods, devices and equipment | |
| CN111402003B (en) | System and method for realizing user-related recommendation | |
| CN112905809A (en) | Knowledge graph learning method and system | |
| CN110288444B (en) | Method and system for realizing user related recommendation | |
| CN111260056A (en) | Network model distillation method and device | |
| CN110024036B (en) | Analytical prediction of antibiotic susceptibility | |
| CN116308644A (en) | Recommendation method, device, equipment, storage medium and product of target product | |
| Bai et al. | Neural maximum common subgraph detection with guided subgraph extraction | |
| CN115456039A (en) | Click rate estimation model training method, click rate estimation method and electronic equipment | |
| US11410220B2 (en) | Exploration for interactive recommendation system, method, and computer program product |
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 |