CN116028882B - User labeling and classifying method, device, equipment and storage medium - Google Patents
User labeling and classifying method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN116028882B CN116028882B CN202310320115.8A CN202310320115A CN116028882B CN 116028882 B CN116028882 B CN 116028882B CN 202310320115 A CN202310320115 A CN 202310320115A CN 116028882 B CN116028882 B CN 116028882B
- Authority
- CN
- China
- Prior art keywords
- user
- historical behavior
- users
- labeling
- index
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Abstract
The invention discloses a user labeling and classifying method, a device, equipment and a storage medium, wherein the method sequentially carries out aggregation operation, differential operation, symbol operation and summation operation on historical behavior indexes of a user to obtain a user classifying label; and constructing and training a time sequence model by using the user historical behavior index and the user classification label, inputting the acquired user historical behavior index of the target user in a preset time period into the preset time sequence model to obtain a corresponding prediction classification label and a prediction classification probability, and finally classifying the label of the target user according to the prediction classification label and the prediction classification probability. According to the method and the system, the user historical behavior index is predicted by constructing the preset time sequence model, so that the corresponding prediction classification label and the prediction classification probability are obtained, the target user can be accurately classified, further operators can adopt an accurate marketing strategy, the activity of the user is improved, and the potential value of the user is maximally mined.
Description
Technical Field
The present invention relates to the field of data processing technologies, and in particular, to a method, an apparatus, a device, and a storage medium for labeling and classifying users.
Background
For an internet enterprise, a user is the owner of the internet enterprise. When the user product scale is smaller, operators can adopt non-differentiated operation means and strategies for all users. But as the size of users continues to expand, the means and policies of operation should be skewed toward high-risk users as well as potential users, thereby reducing users and maximizing the potential value of mining users. Therefore, how to accurately classify users has great significance and value for internet operation.
Traditional user operation and user classification are mainly based on attribute differences of users, such as classification by means of "one-touch" based on attributes such as gender, age, region and the like. Because users with the same attribute may have different product behavior habits, the rough method for classifying the user population is difficult to meet the differentiated user demands.
The foregoing is provided merely for the purpose of facilitating understanding of the technical solutions of the present invention and is not intended to represent an admission that the foregoing is prior art.
Disclosure of Invention
The invention mainly aims to provide a user labeling and classifying method, device, equipment and storage medium, which aim to solve the technical problem of how to accurately classify users.
In order to achieve the above object, the present invention provides a user labeling and classifying method, including:
preprocessing historical behavior indexes of a user, and obtaining user classification labels according to preprocessing results, wherein the preprocessing process comprises the steps of sequentially performing aggregation operation, differential operation, symbol operation and summation operation on the historical behavior indexes;
acquiring a user history behavior index of a target user in a preset time period;
inputting the user historical behavior index into a preset time sequence model to obtain a prediction classification label and a prediction classification probability corresponding to the target user, wherein the preset time sequence model is constructed and trained based on the historical behavior index and the user classification label, and the preset time sequence model is used for performing prediction classification on the target user according to the user historical behavior index and outputting a prediction result;
and classifying the labels of the target users according to the prediction classification labels and the prediction classification probability.
Optionally, before the step of obtaining the user history behavior index of the target user in the preset time period, the method further includes:
Acquiring a sample training set from a preset tag data set, wherein the preset tag data set is obtained based on the historical behavior index and the user classification tag structure;
dividing the sample training set into a training set and a test set in proportion, wherein the training set is used for model training, and the test set is used for model test;
inputting the training set and the user classification labels into an initial cyclic neural network model for iterative training to obtain a training model;
inputting the test set into the training model, and testing the advantages and disadvantages of the training model according to the output result;
when the output result reaches the optimal model parameter, taking a training model corresponding to the output result reaching the optimal model parameter as a preset time sequence model.
Optionally, the step of preprocessing the historical behavior index of the user and obtaining the user classification label according to the preprocessing result includes:
acquiring historical behavior indexes of a user;
performing data preprocessing on the historical behavior indexes according to a preset time period, and determining a behavior trend and a user classification label corresponding to the user based on a processing result;
labeling the users according to the behavior trend and the user classification labels to obtain labeling results, wherein the user classification labels comprise potential users, high-risk users and stable users;
And constructing a preset tag data set according to the historical behavior index and the labeling result.
Optionally, when the distribution of the user tag categories marked in the preset tag data set is uneven, the step of acquiring the sample training set from the preset tag data set includes:
randomly extracting historical behavior indexes corresponding to the stationary users in the preset tag data set in a downsampling mode to obtain stationary index data;
resampling historical behavior indexes corresponding to potential users in the preset tag data set in an up-sampling mode to obtain potential index data;
resampling historical behavior indexes corresponding to the high-risk users in the preset tag data set in an up-sampling mode to obtain high-risk index data;
and when the number of samples of the stable index data, the potential index data and the high-risk index data reaches a preset number, constructing a sample training set according to the stable index data, the potential index data and the high-risk index data corresponding to the number of samples reaching the preset number.
Optionally, the step of preprocessing the data of the historical behavior indexes according to a preset time period and determining the behavior trend and the user classification label corresponding to the user based on the processing result includes:
Performing aggregation operation on the historical behavior indexes according to a preset time period to obtain an aggregation operation result corresponding to the historical behavior indexes;
performing differential operation on the aggregation operation result to obtain a differential operation result corresponding to the aggregation operation result;
performing symbol operation on the differential operation result to obtain a periodic symbol sequence corresponding to the historical behavior index;
and summing the periodic symbol sequences, and determining the corresponding behavior trend of the user based on the summation result.
Optionally, the difference operation result is a positive value, which indicates that the behavior trend of the user is enhanced in the adjacent preset time period;
the difference operation result is a negative value, and represents the behavior trend attenuation of the user in the adjacent preset time period;
and the difference operation result is zero, which indicates that the behavior trend of the user is unchanged in the adjacent preset time period.
Optionally, after the step of classifying the target user according to the prediction classification label and the prediction classification probability, the method further includes:
when the prediction classification probability reaches a preset probability threshold, sequencing the target users according to the prediction classification probability to obtain a sequencing result;
And marking the high-risk users and the potential users in the target users according to the sorting results, and sending the marking results to a server so that the server can adjust the operation strategies of the high-risk users and the potential users according to the marking results.
In addition, to achieve the above object, the present invention also proposes a user labeling and classifying device, the device comprising:
the user labeling module is used for preprocessing the historical behavior indexes of the user and obtaining user classification labels according to the preprocessing result, wherein the preprocessing process comprises the steps of sequentially carrying out aggregation operation, differential operation, symbol operation and summation operation on the historical behavior indexes;
a data acquisition module; the method comprises the steps of acquiring a user history behavior index of a target user in a preset time period;
the model application module is used for inputting the user historical behavior index into a preset time sequence model to obtain a prediction classification label and a prediction classification probability corresponding to the target user, the preset time sequence model is constructed and trained based on the historical behavior index and the user classification label, and the preset time sequence model is used for performing prediction classification on the target user according to the user historical behavior index and outputting a prediction result;
And the user classification module is used for classifying the labels of the target users according to the prediction classification labels and the prediction classification probability.
In addition, to achieve the above object, the present invention also proposes a user labeling and classifying device, the device comprising: the system comprises a memory, a processor, and a user labeling and classifying program stored on the memory and executable on the processor, the user labeling and classifying program configured to implement the steps of the user labeling and classifying method as described above.
In addition, to achieve the above object, the present invention also proposes a storage medium having stored thereon a user labeling and classifying program which, when executed by a processor, implements the steps of the user labeling and classifying method as described above.
The method comprises the steps of preprocessing historical behavior indexes of a user, and obtaining user classification labels according to preprocessing results, wherein the preprocessing process comprises the steps of sequentially performing aggregation operation, differential operation, symbol operation and summation operation on the historical behavior indexes; acquiring a user history behavior index of a target user in a preset time period; inputting the user historical behavior index into a preset time sequence model to obtain a prediction classification label and a prediction classification probability corresponding to the target user, wherein the preset time sequence model is constructed and trained based on the historical behavior index and the user classification label, and the preset time sequence model is used for performing prediction classification on the target user according to the user historical behavior index and outputting a prediction result; and classifying the labels of the target users according to the prediction classification labels and the prediction classification probability. Compared with the existing classification mode based on the user attribute difference, the method predicts the user history behavior index of the target user through the constructed preset time sequence model to obtain the prediction classification label and the prediction classification probability corresponding to the target user, so that the target user can be accurately classified, operators can adopt an accurate marketing strategy, the activity of the user is improved, and the potential value of the mining user is maximized.
Drawings
FIG. 1 is a schematic diagram of a user labeling and classifying device of a hardware operating environment according to an embodiment of the present invention;
FIG. 2 is a flow chart of a first embodiment of the user labeling and classifying method of the present invention;
FIG. 3 is a flow chart of a second embodiment of the user labeling and classification method of the present invention;
FIG. 4 is a schematic diagram of an application flow in a second embodiment of the user labeling and classifying method of the present invention;
FIG. 5 is a block diagram of a first embodiment of a user labeling and sorting device of the present invention.
The achievement of the objects, functional features and advantages of the present invention will be further described with reference to the accompanying drawings, in conjunction with the embodiments.
Detailed Description
It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
Referring to fig. 1, fig. 1 is a schematic diagram of a user labeling and classifying device of a hardware running environment according to an embodiment of the present invention.
As shown in fig. 1, the user labeling and classifying device may include: a processor 1001, such as a central processing unit (Central Processing Unit, CPU), a communication bus 1002, a user interface 1003, a network interface 1004, a memory 1005. Wherein the communication bus 1002 is used to enable connected communication between these components. The user interface 1003 may include a Display, an input unit such as a Keyboard (Keyboard), and the optional user interface 1003 may further include a standard wired interface, a wireless interface. The network interface 1004 may optionally include a standard wired interface, a Wireless interface (e.g., a Wireless-Fidelity (Wi-Fi) interface). The Memory 1005 may be a high-speed random access Memory (Random Access Memory, RAM) or a stable nonvolatile Memory (NVM), such as a disk Memory. The memory 1005 may also optionally be a storage device separate from the processor 1001 described above.
It will be appreciated by those skilled in the art that the structure shown in fig. 1 is not limiting of the user labeling and sorting device, and may include more or fewer components than shown, or certain components in combination, or a different arrangement of components.
As shown in fig. 1, an operating system, a network communication module, a user interface module, and a user labeling and classifying program may be included in the memory 1005 as one type of storage medium.
In the user labeling and sorting device shown in FIG. 1, the network interface 1004 is primarily used for data communication with a network server; the user interface 1003 is mainly used for data interaction with a user; the processor 1001 and the memory 1005 in the user labeling and classifying device of the present invention may be provided in the user labeling and classifying device, and the user labeling and classifying device invokes the user labeling and classifying program stored in the memory 1005 through the processor 1001 and executes the user labeling and classifying method provided by the embodiment of the present invention.
An embodiment of the present invention provides a user labeling and classifying method, and referring to fig. 2, fig. 2 is a schematic flow chart of a first embodiment of the user labeling and classifying method of the present invention.
In this embodiment, the user labeling and classifying method includes the following steps:
Step S10: preprocessing the historical behavior indexes of the user, and obtaining user classification labels according to the preprocessing result, wherein the preprocessing process comprises the steps of sequentially performing aggregation operation, differential operation, sign operation and summation operation on the historical behavior indexes.
It should be noted that, the execution body of the method of the present embodiment may be a computing service device with functions of data processing, network communication and program running, such as a mobile phone, a tablet computer, a personal computer, etc., or may be other electronic devices capable of implementing the same or similar functions, such as the user labeling and classifying devices described above, which is not limited in this embodiment. The user labeling and classifying device (hereinafter referred to as classifying device) in this embodiment and the user labeling and classifying method provided in each embodiment described below will be specifically described.
It is understood that the historical behavior index is a historical behavior data index of the user held in a database operated by the user. For example, taking home broadband user operation as an example, the historical behavior index may be a daily movie watching duration or a time record of a starting state of a user in the past year, or total flow data used by the user in half a year, and specific content of the historical behavior index may be selected according to user operation content of different enterprises, which is not limited in this embodiment.
It should be understood that the user classification label is a classification label obtained by classifying the user according to operation requirements of different enterprises, for example, the user classification label may be a stationary user, a low-potential user, a high-potential user, an important user, a high-risk user, or the like, or may be other classification standards, which is not limited in this embodiment; further users can be categorized into three general categories of user category labels: smooth users, high risk users, potential users. Taking the time of using the enterprise software as an example, if the time of using the software in the historical behavior index of the user does not change greatly, the user is a stable user; if the time for using the software in the historical behavior index of the user is gradually reduced, the user is a high-risk user; if the time for using the software in the historical behavior index of the user gradually increases, the user is a potential user.
In a specific implementation, the classification device may perform an aggregation operation, a difference operation, a sign operation and a summation operation on the historical behavior indexes to obtain a preprocessing result, for example, if the preprocessing result is stable and unchanged, the user tag is a stable user, and the user classification tag is obtained according to different data of the preprocessing result.
Step S20: and acquiring a user history behavior index of the target user in a preset time period.
It should be noted that, the user history behavior index is a history behavior index of the target user, and is different from the above history behavior index in that it is a history behavior index obtained for a specific user.
It should be understood that the preset time period is a time period to which the user history behavior index is acquired, which is preset by the classification device. In general, in order to accurately predict the recent situation of a target user, a user history behavior index of the target user for a recent period of time (for example, three months, half a year, etc.) may be acquired, and the longer the acquired period of time of the user history behavior index, the more accurate the prediction of the recent situation of the target user.
It is understood that the target user is a user who is designated in the user operation process and needs to be operated. For example, taking home broadband user operation as an example, after a new user applies for home broadband, the user can pay attention to behavior index data in a period of time after the new user, and the classification of the user is predicted, so that the operation is performed according to the classification, the retention and activation of the new user are effectively realized, and the viscosity and conversion rate of the new user are improved; or other target users, this embodiment is not limited in this regard.
Step S30: the user historical behavior index is input into a preset time sequence model, a prediction classification label and a prediction classification probability corresponding to the target user are obtained, the preset time sequence model is constructed and trained based on the historical behavior index and the user classification label, and the preset time sequence model is used for performing prediction classification on the target user according to the user historical behavior index and outputting a prediction result.
The preset time-series model is a model for predictive classification of the target user, which is set in the classification apparatus in advance. The model can predict according to a series of data taking time as an independent variable, for example, the login time in half a year of a user, the traffic service condition in three months, the operation directions aimed at by different enterprises are different, the acquired historical behavior data of the user are also different, and the embodiment is not limited to the above; the model can capture the rules of time series, so that the future trend data of the historical behavior indexes of the user can be predicted, and the target user can be classified according to the predicted results.
It can be understood that the prediction classification label is a user classification label obtained by predicting the target user by using a preset time sequence model. For example, the user classification labels may be stationary users, low potential users, high potential users, important users, high risk users, and the like. According to the historical behavior index of the user, the preset time sequence model can predict the trend of the target user in a period of time in the future and classify the trend to output a prediction classification label.
It should be understood that the prediction classification probability is the probability of the preset time sequence model predicting the user classification label to which the target user belongs. Generally, the user history behavior index is fluctuation data in a period of time, the larger the fluctuation is, the more inaccurate the predicted result is, when the prediction classification label is output, the prediction accuracy probability (namely the prediction classification probability) can be output according to the data fluctuation degree of the user history behavior index, and the classification equipment can classify the target user according to whether the prediction classification label is adopted by the prediction classification probability or not, so that the classification accuracy of the classified user can be improved.
Step S40: and classifying the labels of the target users according to the prediction classification labels and the prediction classification probability.
In practical consideration, taking three user classifications (potential users, stationary users, high-risk users) as an example, for a given user historical behavior index, a preset time series model can predict three probability values: p1, p2 and p3, and p1+p2+p3=1, and the predictive class label is the class label corresponding to the largest probability value.
Further, in order to enhance the discovery and mining of high-risk users and potential users, the method further includes, after step S40: when the prediction classification probability reaches a preset probability threshold, sequencing the target users according to the prediction classification probability to obtain a sequencing result; and marking the high-risk users and the potential users in the target users according to the sorting results, and sending the marking results to a server so that the server can adjust the operation strategies of the high-risk users and the potential users according to the marking results.
It should be noted that the preset probability threshold is a preset probability value. The preset probability threshold may be adjusted according to the number of users, and the preset probability threshold may be increased due to an excessive number of users, which is not limited in this embodiment.
It is understood that an operation policy is a policy that an internet enterprise operates for different users. Taking three user classifications (potential users, stable users and high-risk users) and the time of using enterprise software as examples, if the time of using the software in the user history behavior index of the target user does not change greatly, the target user is the stable user; if the time for using the software in the user history behavior index of the target user is gradually reduced, the target user is a high-risk user; if the time for the target user to use the software in the user history behavior index gradually increases, the target user is a potential user. According to different users, different operation strategies can be used for operation, so that user retention and activation are effectively realized, and user viscosity and conversion rate are improved.
In a specific implementation, when the prediction classification probability reaches a preset probability threshold, for example, eighty percent, at this time, the higher the prediction classification probability exceeding the preset probability threshold, the more accurate the prediction classification label of the target user; then sorting can be carried out according to the high-low target users of the prediction classification probability, and a sorting result is obtained; in the operation process, stable users generally occupy most and are not easy to change, in order to effectively realize user retention and activation, the operation can be focused on high-risk users and potential users, and the high-risk users and potential users in target users are marked according to the sorting result; finally, the marking result is sent to a server, so that the server adjusts the operation strategies of the high-risk users and potential users according to the marking result, and for the high-risk users, operators can take targeted retention measures and means to reduce and avoid user loss; for potential users, operators can adopt a strategy of accurate marketing, so that the activity of the users is improved, the potential value of the users is maximally mined, and the profit maximization of products is realized.
The method comprises the steps of preprocessing historical behavior indexes of a user, and obtaining user classification labels according to preprocessing results, wherein the preprocessing process comprises the steps of sequentially performing aggregation operation, differential operation, sign operation and summation operation on the historical behavior indexes; acquiring a user history behavior index of a target user in a preset time period; inputting the user historical behavior index into a preset time sequence model to obtain a prediction classification label and a prediction classification probability corresponding to the target user, wherein the preset time sequence model is constructed and trained based on the historical behavior index and the user classification label, and the preset time sequence model is used for performing prediction classification on the target user according to the user historical behavior index and outputting a prediction result; and classifying the labels of the target users according to the prediction classification labels and the prediction classification probability. Compared with the existing classification mode based on the user attribute difference, the embodiment predicts the user history behavior index of the target user through the constructed preset time sequence model to obtain the prediction classification label and the prediction classification probability corresponding to the target user, so that the target user can be accurately classified, operators can adopt an accurate marketing strategy, the activity of the user is improved, and the potential value of the mining user is maximized. For high-risk users, operators can take targeted retention measures and means, so that user loss is reduced and avoided; for potential users, operators can adopt a strategy of accurate marketing, so that the activity of the users is improved, the potential value of the users is maximally mined, and the profit maximization of enterprise products is realized.
Referring to fig. 3, fig. 3 is a flowchart illustrating a second embodiment of the user labeling and classifying method according to the present invention.
Based on the first embodiment, in this embodiment, in consideration of the establishment of the preset time series model, before step S10, the method further includes:
step S01: and acquiring a sample training set from a preset tag data set, wherein the preset tag data set is acquired based on the historical behavior index and the user classification tag structure.
It should be noted that the preset tag data set is a data set corresponding to the preset user classification tag and the user history behavior index data. Before the preset time sequence model is constructed, historical behavior index data of users of enterprise products can be obtained, the users are initially classified according to the historical behavior index data, and user classification labels are marked, so that a preset label data set is constructed according to the historical behavior index and the user classification labels.
It can be understood that, in order to train the preset time sequence model, a certain amount of sample data (i.e. a sample training set) can be extracted from the preset label data set for training, the sample data can be extracted according to the amount, and the sample data can be extracted according to different classification of the user classification labels.
Step S02: the sample training set is proportionally divided into a training set and a testing set, wherein the training set is used for model training, and the testing set is used for model testing.
Step S03: and inputting the training set and the user classification labels into an initial cyclic neural network model for iterative training to obtain a training model.
It should be noted that, the sample training set may be divided into a training set and a test set according to a preset ratio (for example, 8:2), where the training set is used for training a model, and the test set is used for evaluating and selecting the trained model; meanwhile, the weight coefficient of the model can be obtained by training the model on the training set, and the optimal model can be selected according to the optimal weight coefficient when the model is evaluated on the testing set.
It should be understood that the initial recurrent neural network model is a network model constructed from a recurrent neural network that trains the training set. On the premise of giving historical behavior indexes of users and user classification labels, a model can be trained in a supervised learning mode. Because the behavior index of a user at a certain time point is influenced by the behavior indexes of adjacent time points of the time point, a model can be constructed by adopting a cyclic neural network, such as a LSTM, GRU, TCN classical cyclic neural network structure, and the historical behavior index of the user corresponds to the behavior indexes of different time points of the user, namely, the user can be understood to be given a time sequence, so that the classification of the user can be realized by adopting the cyclic neural network structure training time sequence model.
In a specific implementation, the model training process is approximately: given a weight coefficient (W) and a historical behavior index (X) of a model, the model calculates a prediction classification label, and compares the fitness of the prediction classification label and a user label (a true value), and the fitness is generally expressed by loss function values. And meanwhile, the weight coefficient (W) is adjusted, and a new prediction label and the fitness are obtained through calculation, wherein the higher the fitness is, the better the fitness is.
Step S04: and inputting the test set into the training model, and testing the advantages and disadvantages of the training model according to the output result.
Step S05: when the output result reaches the optimal model parameter, taking a training model corresponding to the output result reaching the optimal model parameter as a preset time sequence model.
The optimal model parameter is an adjustment weight coefficient (W) of the model when the fitness is optimal.
In a specific implementation, after performing iterative training on an initial cyclic neural network model to obtain a training model, in order to improve the accuracy of a preset time sequence model, a test set can be input into the training model to predict, a series of weight coefficients (W) obtained during model training and historical behavior indexes on the test set are calculated to obtain a prediction classification label, the fitness of the prediction classification label and a user label (true value) is compared, and a weight coefficient with the highest fitness is selected as an optimal solution of the model. It can be appreciated that the higher the optimal model parameter setting, the higher the accuracy of the finally obtained pre-set time series model.
Further, considering that the user classification labels of the users need to be labeled before the preset time series model is built, in this embodiment, the step S10 includes: acquiring historical behavior indexes of a user; performing data preprocessing on the historical behavior indexes according to a preset time period, and determining a behavior trend and a user classification label corresponding to the user based on a processing result; labeling the users according to the behavior trend and the user classification labels to obtain labeling results, wherein the user classification labels comprise potential users, high-risk users and stable users; and constructing a preset tag data set according to the historical behavior index and the labeling result.
It should be noted that the preset time period is a preset time period, and the behavior change of the user is reflected by the change of the time period. For example, the historical behavior index data of the user in one year is acquired, and the historical behavior index may be processed in weeks or months, or in days, which is not limited in this embodiment.
It can be understood that the behavior trends are in one-to-one correspondence with the above user classification labels, for example, the stationary user corresponds to the stationary trend, the high-risk user corresponds to the decay trend, and the potential user corresponds to the enhancement trend, which is not repeated in this embodiment.
When the historical behavior indexes are subjected to data preprocessing, a discount chart can be constructed according to the historical behavior indexes, and behavior trends corresponding to users can be judged according to the trends of the discount chart; the historical behavior indexes can be subjected to a series of operations such as differential operation, symbol operation and the like, and the behavior trend corresponding to the user is reflected by the change of the data; or otherwise, the present embodiment is not limited thereto.
In a specific implementation, the classification device obtains a historical behavior index of a user; performing data preprocessing on the historical behavior indexes according to a preset time period, and determining a behavior trend corresponding to the user and a user classification label based on a processing result; labeling the users according to the behavior trend and the user classification labels, for example, labeling potential users, high-risk users, stable users and the like; and finally, constructing a preset tag data set according to the historical behavior indexes and the corresponding labeling results, so that accurate data can be provided for training of a preset time sequence model, and the prediction accuracy of the preset time sequence model is improved.
Further, considering that the number of stationary users, high-risk users and potential users in the users is not consistent, when the distribution of the user tag categories marked in the preset tag data set is not uniform, the step of acquiring the sample training set from the preset tag data set in the embodiment includes: randomly extracting historical behavior indexes corresponding to the stationary users in the preset tag data set in a downsampling mode to obtain stationary index data; resampling historical behavior indexes corresponding to potential users in the preset tag data set in an up-sampling mode to obtain potential index data; resampling historical behavior indexes corresponding to the high-risk users in the preset tag data set in an up-sampling mode to obtain high-risk index data; and when the number of samples of the stable index data, the potential index data and the high-risk index data reaches a preset number, constructing a sample training set according to the stable index data, the potential index data and the high-risk index data corresponding to the number of samples reaching the preset number.
It should be noted that, the stationary index data is a historical behavior index corresponding to the stationary user, the potential index data is a historical behavior index corresponding to the potential user, and the high-risk index data corresponds to a historical behavior index corresponding to the high-risk user.
It is understood that the preset number is a preset number of samples for acquiring the stability index data, the potential index data, and the high-risk index data. The preset number may be a specific number or a number ratio, which is not limited in this embodiment.
In practical considerations, a sample construction sample dataset may be extracted from the pre-set tag datasets of initially labeled stationary, potential, and high risk users. In general, the smooth user, high risk user, and potential user quantity distributions for user tagging process annotations are highly non-uniform. The sample size of the stable users in enterprise product users is often larger, and the sample size of the stable users can be reduced in a downsampling mode; samples marked as high-risk users and potential users are often fewer, and resampling can be performed in an upsampling mode to appropriately increase the sample sizes of the high-risk users and potential users. The sampling principle is that potential users and high-risk users sample as many as possible, and the sample sizes of stationary users, potential users and high-risk users are close to 1:1:1. after the number of samples reaches 1 of the preset number: 1: and 1, constructing a sample training set according to the corresponding stable index data, potential index data and high-risk index data. Therefore, the accuracy of prediction of high-risk users and potential users among the users can be improved when the preset time sequence model is trained.
Further, considering accuracy of data processing on the historical behavior indexes, in this embodiment, the step of performing data preprocessing on the historical behavior indexes according to a preset time period in the step, and determining the behavior trend and the user classification label corresponding to the user based on the processing result includes: performing aggregation operation on the historical behavior indexes according to a preset time period to obtain an aggregation operation result corresponding to the historical behavior indexes; performing differential operation on the aggregation operation result to obtain a differential operation result corresponding to the aggregation operation result; performing symbol operation on the differential operation result to obtain a periodic symbol sequence corresponding to the historical behavior index; and summing the periodic symbol sequences, and determining the corresponding behavior trend of the user based on the summation result.
It should be noted that, the aggregation operation is an aggregation operation such as summing or averaging the historical behavior indexes according to a given period T, and aims to eliminate the influence of short-term periodic fluctuation on the long-term behavior trend of the user. The historical behavior index may be expressed as:
x 1 ，…，x T ；x T+1 ，…，x 2T ；……；x （n-1）T+1 ，…，x nT ；
wherein x is 1 To x T For data represented by historical behavior index in the first period T, x T+1 To x 2T Data represented by historical behavior index in the second period T, and so on, x （n-1）T+1 To x nT Is the data represented by the historical behavior index in the nth period T. Taking the example of the mean value aggregation operation, the formula can be expressed as follows:
wherein x is n And (3) obtaining an aggregation operation result for the historical behavior index data in the nth period.
It is understood that the differential operation may be a first order differential operation, which may reflect the trend of the aggregate result change. Further, in this embodiment, the difference operation result is a positive value, which indicates that the behavior trend of the user is enhanced in the adjacent preset time period; the difference operation result is a negative value, and represents the behavior trend attenuation of the user in the adjacent preset time period; and the difference operation result is zero, which indicates that the behavior trend of the user is unchanged in the adjacent preset time period. The larger the absolute value of the first order difference, the larger the user behavior trend change rate representing the adjacent time period. The differential operation formula may be:
wherein x is n Aggregation operation result obtained for historical behavior index in nth period, x n-1 Aggregate operation result, deltax, obtained for historical behavior index in n-1 th period n-1 The difference operation result of the n-1 th is obtained from the aggregation operation result of the n-1 th and the aggregation operation result of the n-th.
It should be understood that the sign operation is to operate on the differential operation result, and the formula may be:
wherein Deltax is n For the nth differential operation result, f (x) represents the sign operation result obtained from the nth differential operation result, x>0 has a value of 1; x is x<0 has a value of-1; the value of x=0 is 0, and the difference operation result is subjected to sign operation to obtain a periodic sign sequence consisting of-1 and 1 (which may also contain 0). Only the direction of the user behavior trend in adjacent time periods can be considered through the symbolic operation, and the magnitude of the local behavior trend change is ignored. The symbolic operation eliminates the influence of the local behavior trend extreme value, and the result is more biased to reflect the trend of the long-term change of the user behavior.
In a specific implementation, the periodic symbol sequence may be summed, and the result may represent a trend in the user behavior. The larger the value is, the trend of the user behavior is enhanced; the smaller the value, the less the user behavior trend decays; a value near 0 indicates a smooth trend of user behavior. Therefore, the users can be initially classified based on the summation result of the periodic symbol sequences, the behavior trend of the users is determined, and the potential users, the high-risk users and the stable users are sequentially corresponding, so that the accuracy of processing the historical behavior indexes is improved.
In practical consideration, referring to fig. 4, fig. 4 is a schematic application flow chart of a second embodiment of the user labeling and classifying method according to the present invention.
First, user marking is performed. The user mark is a process of realizing preliminary classification of users based on user behavior trend, taking home broadband user operation as an example, by researching historical user behavior indexes (such as daily movie watching time length or starting state of users in the past 1 year), carrying out aggregation operation (summation or averaging) on the historical behavior indexes according to a given period (week or month) to eliminate the influence of periodic fluctuation on the long-term trend of user behavior; performing differential operation (for example, first-order differential operation) on the aggregation operation result in a given period, wherein the differential operation result can reflect the change trend of the user index; carrying out symbol operation on the differential operation result, wherein the symbol operation result can reflect the behavior change trend of the adjacent period; finally, the symbol operation result (such as-1, 1 or 0) of each period is summed and divided. The closer the result of the summation is to 0, the more smooth the user behavior trend is; the larger the summation result is, the trend of the user behavior is enhanced; the smaller the summation result is, the less the user behavior trend decays. Therefore, the user category can be initially divided and marked, and the user category labels are respectively and correspondingly marked as stable users, potential users and high-risk users based on the user behavior trend, so that the user category labels are obtained.
Model training is then performed. The model training is a process of training a supervised learning model based on historical behavior indexes of users and given user classification labels, and adopts a cyclic neural network structure to train a time sequence model, so that classification of user groups can be realized, and high-risk users and potential users can be mined.
And finally, performing model application. According to the behavior index of the user to be mined, the time sequence model can predict the classification label and the classification probability of the user to be mined, and based on the predicted classification label and classification probability, high-risk users and potential users can be found and mined, so that the retention and the activation of the users are improved.
For a given user historical behavior index, the embodiment performs aggregation operation on the user behavior index according to a given period, performs differential operation and sign operation on an aggregation operation result, sums and classifies the sign operation result, and primarily screens and marks high-risk users, potential users and stable users. And taking a given user behavior index as input, taking the marked user type as a user classification label, adopting a cyclic neural network to construct a training time sequence model, and realizing classification of users, thereby realizing discovery and mining of high-risk users and potential users. Therefore, the retention and activation of the user are effectively realized, the viscosity and conversion rate of the user are improved, and the profit maximization is realized.
In addition, the embodiment of the invention also provides a storage medium, wherein the storage medium is stored with a user labeling and classifying program, and the user labeling and classifying program realizes the steps of the user labeling and classifying method when being executed by a processor.
Referring to fig. 5, fig. 5 is a block diagram showing the construction of a first embodiment of the user labeling and classifying apparatus according to the present invention.
As shown in fig. 5, the user labeling and classifying device according to the embodiment of the present invention includes:
the user labeling module 501 is configured to preprocess a historical behavior index of a user, and obtain a user classification label according to a preprocessing result, where the preprocessing process includes sequentially performing an aggregation operation, a difference operation, a sign operation, and a summation operation on the historical behavior index.
A data acquisition module 502; the method comprises the steps of acquiring a user history behavior index of a target user in a preset time period;
the model application module 503 is configured to input the user historical behavior index into a preset time sequence model, obtain a prediction classification label and a prediction classification probability corresponding to the target user, construct and train the preset time sequence model based on the historical behavior index and the user classification label, and perform prediction classification on the target user according to the user historical behavior index and output a prediction result;
And the user classification module 504 is configured to classify the target user according to the prediction classification label and the prediction classification probability.
In the embodiment, aggregation operation, differential operation, symbol operation and summation operation are sequentially carried out on the historical behavior indexes of the user, so that a user classification label is obtained; and constructing and training a time sequence model by using the user historical behavior index and the user classification label, inputting the acquired user historical behavior index of the target user in a preset time period into the preset time sequence model to obtain a corresponding prediction classification label and a prediction classification probability, and finally classifying the label of the target user according to the prediction classification label and the prediction classification probability. According to the method and the system, the user historical behavior index is predicted by constructing the preset time sequence model, so that the corresponding prediction classification label and the prediction classification probability are obtained, the target user can be accurately classified, further operators can adopt an accurate marketing strategy, the activity of the user is improved, and the potential value of the user is maximally mined.
Other embodiments or specific implementations of the user labeling and classifying device of the present invention may refer to the above method embodiments, and will not be described herein.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or system. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or system that comprises the element.
The foregoing embodiment numbers of the present invention are merely for the purpose of description, and do not represent the advantages or disadvantages of the embodiments.
From the above description of embodiments, it will be clear to a person skilled in the art that the above embodiment method may be implemented by means of software plus a necessary general hardware platform, but may of course also be implemented by means of hardware, but in many cases the former is a preferred embodiment. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art in the form of a software product stored in a storage medium (e.g. read-only memory/random-access memory, magnetic disk, optical disk), comprising instructions for causing a terminal device (which may be a mobile phone, a computer, a server, or a network device, etc.) to perform the method according to the embodiments of the present invention.
The foregoing description is only of the preferred embodiments of the present invention, and is not intended to limit the scope of the invention, but rather is intended to cover any equivalents of the structures or equivalent processes disclosed herein or in the alternative, which may be employed directly or indirectly in other related arts.
Claims (10)
1. A method of user labeling and categorizing, the method comprising:
preprocessing historical behavior indexes of a user, and obtaining user classification labels according to preprocessing results, wherein the preprocessing process comprises the steps of sequentially performing aggregation operation, differential operation, symbol operation and summation operation on the historical behavior indexes;
acquiring a user history behavior index of a target user in a preset time period;
inputting the user historical behavior index into a preset time sequence model to obtain a prediction classification label and a prediction classification probability corresponding to the target user, wherein the preset time sequence model is constructed and trained based on the historical behavior index and the user classification label, and the preset time sequence model is used for performing prediction classification on the target user according to the user historical behavior index and outputting a prediction result;
And classifying the labels of the target users according to the prediction classification labels and the prediction classification probability.
2. The method for labeling and classifying a user as set forth in claim 1, wherein said step of obtaining a user history behavior index of the target user for a predetermined period of time is preceded by the step of:
acquiring a sample training set from a preset tag data set, wherein the preset tag data set is obtained based on the historical behavior index and the user classification tag structure;
dividing the sample training set into a training set and a test set in proportion, wherein the training set is used for model training, and the test set is used for model test;
inputting the training set and the user classification labels into an initial cyclic neural network model for iterative training to obtain a training model;
inputting the test set into the training model, and testing the advantages and disadvantages of the training model according to the output result;
when the output result reaches the optimal model parameter, taking a training model corresponding to the output result reaching the optimal model parameter as a preset time sequence model.
3. The user labeling and classifying method as set forth in claim 2, wherein the step of preprocessing the historical behavior index of the user and obtaining the user classification label according to the preprocessing result comprises:
Acquiring historical behavior indexes of a user;
performing data preprocessing on the historical behavior indexes according to a preset time period, and determining a behavior trend and a user classification label corresponding to the user based on a processing result;
labeling the users according to the behavior trend and the user classification labels to obtain labeling results, wherein the user classification labels comprise potential users, high-risk users and stable users;
and constructing a preset tag data set according to the historical behavior index and the labeling result.
4. The user labeling and classifying method according to claim 3, wherein said step of acquiring the sample training set from the preset tag data set when the distribution of the user tag categories marked in the preset tag data set is not uniform comprises:
randomly extracting historical behavior indexes corresponding to the stationary users in the preset tag data set in a downsampling mode to obtain stationary index data;
resampling historical behavior indexes corresponding to potential users in the preset tag data set in an up-sampling mode to obtain potential index data;
resampling historical behavior indexes corresponding to the high-risk users in the preset tag data set in an up-sampling mode to obtain high-risk index data;
And when the number of samples of the stable index data, the potential index data and the high-risk index data reaches a preset number, constructing a sample training set according to the stable index data, the potential index data and the high-risk index data corresponding to the number of samples reaching the preset number.
5. The user labeling and classifying method as set forth in claim 3, wherein said step of preprocessing said historical behavior index data for a predetermined period of time and determining said user's corresponding behavior trend and user classification label based on the processing result comprises:
performing aggregation operation on the historical behavior indexes according to a preset time period to obtain an aggregation operation result corresponding to the historical behavior indexes;
performing differential operation on the aggregation operation result to obtain a differential operation result corresponding to the aggregation operation result;
performing symbol operation on the differential operation result to obtain a periodic symbol sequence corresponding to the historical behavior index;
and summing the periodic symbol sequences, and determining the corresponding behavior trend of the user based on the summation result.
6. The user labeling and classifying method according to claim 5, wherein the difference operation result is a positive value, which indicates that the behavior trend of the user is enhanced in the adjacent preset time period;
The difference operation result is a negative value, and represents the behavior trend attenuation of the user in the adjacent preset time period;
and the difference operation result is zero, which indicates that the behavior trend of the user is unchanged in the adjacent preset time period.
7. The user labeling and categorizing method of claim 1, wherein after the step of categorizing the target user based on the predictive categorization label and the predictive categorization probability, further comprising:
when the prediction classification probability reaches a preset probability threshold, sequencing the target users according to the prediction classification probability to obtain a sequencing result;
and marking the high-risk users and the potential users in the target users according to the sorting results, and sending the marking results to a server so that the server can adjust the operation strategies of the high-risk users and the potential users according to the marking results.
8. A user labeling and sorting device, the device comprising:
the user labeling module is used for preprocessing the historical behavior indexes of the user and obtaining user classification labels according to the preprocessing result, wherein the preprocessing process comprises the steps of sequentially carrying out aggregation operation, differential operation, symbol operation and summation operation on the historical behavior indexes;
A data acquisition module; the method comprises the steps of acquiring a user history behavior index of a target user in a preset time period;
the model application module is used for inputting the user historical behavior index into a preset time sequence model to obtain a prediction classification label and a prediction classification probability corresponding to the target user, the preset time sequence model is constructed and trained based on the historical behavior index and the user classification label, and the preset time sequence model is used for performing prediction classification on the target user according to the user historical behavior index and outputting a prediction result;
and the user classification module is used for classifying the labels of the target users according to the prediction classification labels and the prediction classification probability.
9. A user labeling and sorting device, the device comprising: memory, a processor and a user labeling and classifying program stored on the memory and executable on the processor, the user labeling and classifying program being configured to implement the steps of the user labeling and classifying method according to any of claims 1 to 7.
10. A storage medium having stored thereon a user labeling and sorting program which when executed by a processor performs the steps of the user labeling and sorting method of any of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310320115.8A CN116028882B (en) | 2023-03-29 | 2023-03-29 | User labeling and classifying method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310320115.8A CN116028882B (en) | 2023-03-29 | 2023-03-29 | User labeling and classifying method, device, equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116028882A CN116028882A (en) | 2023-04-28 |
| CN116028882B true CN116028882B (en) | 2023-06-02 |
Family
ID=86089678
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310320115.8A Active CN116028882B (en) | 2023-03-29 | 2023-03-29 | User labeling and classifying method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116028882B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116842330B (en) * | 2023-08-31 | 2023-11-24 | 庆云县人民医院 | Health care information processing method and device capable of comparing histories |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107358156A (en) * | 2017-06-06 | 2017-11-17 | 华南理工大学 | The feature extracting method of Ultrasonic tissue characterization based on Hilbert-Huang transform |
| WO2017219548A1 (en) * | 2016-06-20 | 2017-12-28 | 乐视控股（北京）有限公司 | Method and device for predicting user attributes |
| CN112163617A (en) * | 2020-09-25 | 2021-01-01 | 深圳市傲天科技股份有限公司 | Label-free numerical value type feature classification method, device, equipment and readable storage medium |
| CN112200016A (en) * | 2020-09-17 | 2021-01-08 | 东北林业大学 | Electroencephalogram signal emotion recognition based on ensemble learning method AdaBoost |
| CN113392237A (en) * | 2021-06-15 | 2021-09-14 | 青岛聚看云科技有限公司 | Classified label display method, server and display equipment |
| CN114154866A (en) * | 2021-12-02 | 2022-03-08 | 北京顶象技术有限公司 | Marketing enterprise financial risk early warning method and system |
| CN114445252A (en) * | 2021-11-15 | 2022-05-06 | 南方科技大学 | Data completion method and device, electronic equipment and storage medium |
| CN115186764A (en) * | 2022-08-03 | 2022-10-14 | 腾讯科技（北京）有限公司 | Data processing method and device, electronic equipment and storage medium |
| CN115456258A (en) * | 2022-08-26 | 2022-12-09 | 中远海运科技(北京)有限公司 | Method for predicting transportation capacity of competitor ship and computer readable medium |
| CN115688798A (en) * | 2022-10-29 | 2023-02-03 | 中国人民解放军战略支援部队信息工程大学 | Interpretable false information detection method and device based on credible evidence reasoning and temporal-spatial feature aggregation |
| CN115858675A (en) * | 2022-12-05 | 2023-03-28 | 西安电子科技大学 | Non-independent same-distribution data processing method based on federal learning framework |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP4230964A1 (en) * | 2017-01-12 | 2023-08-23 | Mobileye Vision Technologies Ltd. | Navigation based on vehicle activity |
| US20200160177A1 (en) * | 2018-11-16 | 2020-05-21 | Royal Bank Of Canada | System and method for a convolutional neural network for multi-label classification with partial annotations |
| US20210405008A1 (en) * | 2020-06-29 | 2021-12-30 | Dräger Safety AG & Co. KGaA | Monitoring system |
| US20220012773A1 (en) * | 2020-07-09 | 2022-01-13 | Shopify Inc. | Systems and methods for detecting multiple users of an online account |
-
2023
- 2023-03-29 CN CN202310320115.8A patent/CN116028882B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017219548A1 (en) * | 2016-06-20 | 2017-12-28 | 乐视控股（北京）有限公司 | Method and device for predicting user attributes |
| CN107358156A (en) * | 2017-06-06 | 2017-11-17 | 华南理工大学 | The feature extracting method of Ultrasonic tissue characterization based on Hilbert-Huang transform |
| CN112200016A (en) * | 2020-09-17 | 2021-01-08 | 东北林业大学 | Electroencephalogram signal emotion recognition based on ensemble learning method AdaBoost |
| CN112163617A (en) * | 2020-09-25 | 2021-01-01 | 深圳市傲天科技股份有限公司 | Label-free numerical value type feature classification method, device, equipment and readable storage medium |
| CN113392237A (en) * | 2021-06-15 | 2021-09-14 | 青岛聚看云科技有限公司 | Classified label display method, server and display equipment |
| CN114445252A (en) * | 2021-11-15 | 2022-05-06 | 南方科技大学 | Data completion method and device, electronic equipment and storage medium |
| CN114154866A (en) * | 2021-12-02 | 2022-03-08 | 北京顶象技术有限公司 | Marketing enterprise financial risk early warning method and system |
| CN115186764A (en) * | 2022-08-03 | 2022-10-14 | 腾讯科技（北京）有限公司 | Data processing method and device, electronic equipment and storage medium |
| CN115456258A (en) * | 2022-08-26 | 2022-12-09 | 中远海运科技(北京)有限公司 | Method for predicting transportation capacity of competitor ship and computer readable medium |
| CN115688798A (en) * | 2022-10-29 | 2023-02-03 | 中国人民解放军战略支援部队信息工程大学 | Interpretable false information detection method and device based on credible evidence reasoning and temporal-spatial feature aggregation |
| CN115858675A (en) * | 2022-12-05 | 2023-03-28 | 西安电子科技大学 | Non-independent same-distribution data processing method based on federal learning framework |
Non-Patent Citations (1)
| Title |
|---|
| 基于用户行为的特征工程构建与应用研究;白肇强;中国优秀硕士学位论文全文数据库 经济与管理科学辑(第12期);J157-458 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116028882A (en) | 2023-04-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20210224832A1 (en) | Method and apparatus for predicting customer purchase intention, electronic device and medium | |
| CN108563722B (en) | Industry classification method, system, computer device and storage medium for text information | |
| CN108596630B (en) | Fraud transaction identification method, system and storage medium based on deep learning | |
| US20140074614A1 (en) | Time series-based entity behavior classification | |
| CN111210335B (en) | User risk identification method and device and electronic equipment | |
| CN110705719A (en) | Method and apparatus for performing automatic machine learning | |
| CN110705592A (en) | Classification model training method, device, equipment and computer readable storage medium | |
| CN111753087A (en) | Public opinion text classification method and device, computer equipment and storage medium | |
| CN116028882B (en) | User labeling and classifying method, device, equipment and storage medium | |
| CN112101520A (en) | Risk assessment model training method, business risk assessment method and other equipment | |
| CN111145009A (en) | Method and device for evaluating risk after user loan and electronic equipment | |
| CN111210336A (en) | User risk model generation method and device and electronic equipment | |
| CN111275492A (en) | User portrait generation method, device, storage medium and equipment | |
| CN111582314A (en) | Target user determination method and device and electronic equipment | |
| CN108182633B (en) | Loan data processing method, loan data processing device, loan data processing program, and computer device and storage medium | |
| CN112994960A (en) | Method and device for detecting business data abnormity and computing equipment | |
| CN112995414B (en) | Behavior quality inspection method, device, equipment and storage medium based on voice call | |
| Jinbo et al. | The application ofadaboost in customer churn prediction | |
| CN111898249A (en) | Landslide displacement nonparametric probability density prediction method, equipment and storage medium | |
| CN116029760A (en) | Message pushing method, device, computer equipment and storage medium | |
| CN113887633B (en) | Malicious behavior identification method and system for closed source power industrial control system based on IL | |
| US11556935B1 (en) | Financial risk management based on transactions portrait | |
| US20190114673A1 (en) | Digital experience targeting using bayesian approach | |
| CN114091789A (en) | Project data fluctuation prediction method and device, computer equipment and storage medium | |
| CN116304358B (en) | User data acquisition method |
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 |