Method and apparatus for determining a marketing target

Abstract

The application provides a marketing object determination method and device. The method comprises the following steps: generating multiple decision trees based on a marketing object data set and according to a decision tree algorithm for maximizing a specific target value, wherein one decision tree corresponds to one marketing object screening rule generated by a combination of marketing object characteristics; extracting the marketing object screening rules corresponding to the multiple decision trees and composing a marketing object screening rule set; screening the marketing object data set based on the marketing object screening rule set to determine a target marketing object satisfying all screening rules in the marketing object screening rule set. Through the application, the problem that an existing gain model based on a decision tree cannot accurately select a target marketing object is solved, and the effect of avoiding waste of marketing resources is achieved.

Description

Technical Field

[0001] This invention relates to the field of machine learning technology, and more specifically, to a method and apparatus for determining marketing targets. Background Technology

[0002] With the development of machine learning and artificial intelligence technologies, intelligent marketing is permeating all industries. How to measure and predict the "incremental improvement" brought about by marketing interventions—that is, to find the people who "buy because of the marketing campaign"—instead of wasting marketing budgets on people who "would have bought anyway," has become the most important challenge for intelligent marketing algorithms.

[0003] Uplift models are a crucial technology in intelligent marketing, but most existing uplift models suffer from a significant problem: a lack of interpretability. This means that people cannot understand the logic behind the model's predictions, making it difficult to fully accept or reject the results.

[0004] Existing interpretable gain models are based on decision trees, which treat the experimental and control groups as two probability distributions of Y. Following the standard Classification and Regression Tree (CART) model, they design splitting criteria to maximize the difference in distribution between the two groups after splitting, thus achieving precise stratification. The most commonly used splitting criteria are Kullback-Leibler divergence, Euclidean distance, and chi-square distance. However, this method, in building the decision tree, does not directly use maximizing the difference in probability distribution of Y (i.e., gain) as the splitting criterion. Instead, it uses existing measures of distribution difference (such as KL divergence and Euclidean distance), resulting in inaccurate gain predictions and an inability to accurately select target marketing objects, thus wasting marketing resources. Summary of the Invention

[0005] This invention provides a method and apparatus for determining marketing targets, at least to solve the problem that existing decision tree-based gain models in related technologies cannot accurately select target marketing targets.

[0006] According to an embodiment of the present invention, a method for determining marketing targets is provided, comprising: generating multiple decision trees based on a marketing target dataset and according to a decision tree algorithm for maximizing a specific target value, wherein each decision tree corresponds to a marketing target screening rule composed of multiple marketing target features; extracting the marketing target screening rules corresponding to the multiple decision trees and forming a marketing target screening rule set; and screening the marketing target dataset based on the marketing target screening rule set to determine target marketing targets that satisfy all screening rules in the marketing target screening rule set.

[0007] In one exemplary embodiment, each decision tree is a single-branch decision tree, and a single-branch decision tree contains multiple of the marketing object features.

[0008] In one exemplary embodiment, generating multiple decision trees includes: calculating the gain of all marketing object features in the marketing object dataset according to a decision tree algorithm that maximizes a specific target value; taking the final split point corresponding to the first marketing object feature with the largest gain value as the first child node of the first decision tree, and taking the splitting rule corresponding to the first marketing object feature as the first filtering rule of the first child node; calculating the gain of all marketing object features in the marketing object data not covered by the first filtering rule according to the decision tree algorithm that maximizes a specific target value, to obtain other child nodes of the first decision tree, until the depth of the first decision tree reaches a first threshold.

[0009] In one exemplary embodiment, generating multiple decision trees further includes: performing gain calculation on the marketing object data not covered by the filtering rules of the first decision tree according to the decision tree algorithm for maximizing a specific target value, to obtain a second decision tree; and performing gain calculation on the marketing object data not covered by the filtering rules of the first and second decision trees according to the decision tree algorithm for maximizing a specific target value, to obtain other decision trees, until the number of decision trees reaches a second threshold.

[0010] In an exemplary embodiment, the gain calculation for all marketing object features in the marketing object dataset further includes: for any marketing object feature, calculating the gain by using each value of the marketing object feature as a split point; using the value corresponding to the maximum gain value as the final split point of the marketing object feature, and using the splitting rule corresponding to the maximum gain value as the filtering rule for the marketing object feature.

[0011] In one exemplary embodiment, the method further includes: performing pre-pruning during the generation of the decision tree to stop generating the decision tree if a third threshold is not met, wherein the third threshold includes at least one of the following: number of marketing targets, recall rate, gain, and marketing cost.

[0012] According to another embodiment of the present invention, a device for determining marketing targets is provided, comprising: a generation module, configured to generate multiple decision trees based on a marketing target dataset and according to a decision tree algorithm for maximizing a specific target value, wherein each decision tree corresponds to a marketing target screening rule composed of multiple marketing target features; an extraction module, configured to extract the marketing target screening rules corresponding to the multiple decision trees and form a marketing target screening rule set; and a determination module, configured to screen the marketing target dataset based on the marketing target screening rule set to determine target marketing targets that satisfy all screening rules in the marketing target screening rule set. In an exemplary embodiment, each decision tree is a single-branch decision tree, and a single-branch decision tree contains multiple marketing target features.

[0013] In one exemplary embodiment, the system further includes a stopping module for pre-pruning during the generation of the decision tree to stop generating the decision tree if a third threshold is not met, wherein the third threshold includes at least one of the following: number of marketing targets, recall rate, gain, and marketing cost.

[0014] According to yet another embodiment of the present invention, a storage medium is also provided, wherein a computer program is stored therein, wherein the computer program is configured to execute the steps in any of the above method embodiments when running.

[0015] According to yet another embodiment of the present invention, an electronic device is also provided, including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform the steps in any of the above method embodiments.

[0016] This invention, by generating decision trees based on algorithms that maximize specific target values, provides more accurate gain predictions, thus facilitating the precise selection of target marketing objects. Therefore, it solves the problem that existing decision tree-based gain models cannot accurately select target marketing objects, thereby avoiding the waste of marketing resources. Attached Figure Description

[0017] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings: Figure 1 This is a hardware structure block diagram of a computer terminal for implementing the method for determining marketing targets according to an embodiment of the present invention; Figure 2 This is a flowchart of a method for determining marketing targets according to an embodiment of the present invention; Figure 3This is a structural block diagram of a marketing target determination device according to an embodiment of the present invention; Figure 4 This is a structural block diagram of a marketing target determination device according to another embodiment of the present invention; Figure 5 This is a schematic diagram of pseudocode for selecting feature splitting points according to an embodiment of the present invention; Figure 6 This is a schematic diagram of pseudocode for constructing a single tree and automatically extracting rules according to an embodiment of the present invention; Figure 7 This is a schematic diagram of the pseudocode for a set of combined rules according to an embodiment of the present invention. Detailed Implementation

[0018] The present invention will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the present application can be combined with each other.

[0019] It should be noted that the terms "first," "second," etc., in the specification, claims, and drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0020] Example 1

[0021] The method embodiment provided in Embodiment 1 of this application can be executed on a mobile terminal, computer terminal, or similar computing device. Taking its operation on a computer terminal as an example, Figure 1 This is a hardware structure block diagram of a computer terminal for implementing the method for determining marketing targets according to an embodiment of the present invention. Figure 1 As shown, a computer terminal may include one or more ( Figure 1 Only one is shown in the diagram. A processor 102 (which may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.) and a memory 104 for storing data are also shown. Optionally, the computer terminal may further include a transmission device 106 for communication functions and an input / output device 108. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the computer terminal described above. For example, the computer terminal may also include components that are more complex than those described above. Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0022] The memory 104 can be used to store computer programs, such as application software programs and modules, like the computer program corresponding to the marketing target determination method in this embodiment of the invention. The processor 102 executes various functional applications and data processing by running the computer programs stored in the memory 104, thereby implementing the above-described method. The memory 104 may include high-speed random access memory and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to a computer terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0023] The transmission device 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by a communication provider for the computer terminal. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 may be a Radio Frequency (RF) module used for wireless communication with the Internet.

[0024] This embodiment provides a method for determining marketing targets running on the aforementioned computer terminal. Figure 2 This is a flowchart of a method for determining marketing targets according to an embodiment of the present invention, such as... Figure 2 As shown, the process includes the following steps: Step S202: Based on the marketing object dataset and according to the decision tree algorithm for maximizing specific target values, generate multiple decision trees, wherein each decision tree corresponds to a marketing object screening rule composed of multiple marketing object features; Step S204: Extract the marketing object screening rules corresponding to the multiple decision trees and form a marketing object screening rule set; Step S206: Filter the marketing object dataset based on the marketing object filtering rule set to determine the target marketing objects that satisfy all filtering rules in the marketing object filtering rule set.

[0025] In this embodiment, each decision tree is a single-branch decision tree, and a single-branch decision tree contains multiple marketing object features.

[0026] In step S202 of this embodiment, the following steps are included: calculating the gain of all marketing object features in the marketing object dataset according to the decision tree algorithm for maximizing a specific target value; taking the final split point corresponding to the first marketing object feature with the largest gain value as the first child node of the first decision tree, and taking the splitting rule corresponding to the first marketing object feature as the first filtering rule of the first child node; calculating the gain of all marketing object features in the marketing object data not covered by the first filtering rule according to the decision tree algorithm for maximizing a specific target value, and obtaining other child nodes of the first decision tree, until the depth of the first decision tree reaches a first threshold.

[0027] In step S202 of this embodiment, the method further includes: performing gain calculation on the marketing object data not covered by the filtering rules of the first decision tree according to the decision tree algorithm for maximizing a specific target value, so as to obtain a second decision tree; and performing gain calculation on the marketing object data not covered by the filtering rules of the first decision tree and the second decision tree according to the decision tree algorithm for maximizing a specific target value, so as to obtain other decision trees, until the number of trees in the decision tree reaches a second threshold.

[0028] In this embodiment, the method further includes: calculating the gain of all marketing object features in the marketing object dataset, and also includes: For any of the marketing object features, each value of the marketing object feature is used as a split point for gain calculation; The value corresponding to the maximum gain is taken as the final split point of the marketing object feature, and the division rule corresponding to the maximum gain is taken as the screening rule of the marketing object feature.

[0029] In one exemplary embodiment, the method further includes: performing pre-pruning during the generation of the decision tree to stop generating the decision tree if a third threshold is not met, wherein the third threshold includes at least one of the following: number of marketing targets, recall rate, gain, and marketing cost.

[0030] Through the steps described above, since decision trees are generated based on algorithms that maximize specific target values, their gain predictions are more accurate, which is beneficial for accurately selecting target marketing objects. Therefore, this solves the problem that existing decision tree-based gain models cannot accurately select target marketing objects, thus avoiding the waste of marketing resources.

[0031] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0032] This embodiment also provides a marketing target identification device for implementing the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0033] Figure 3 This is a structural block diagram of a marketing target determination device according to an embodiment of the present invention, such as... Figure 3 As shown, the device includes: a generation module 10, an extraction module 20, and a determination module 30.

[0034] The generation module 10 is used to generate multiple decision trees based on the marketing object dataset and according to the decision tree algorithm that maximizes a specific target value. Each decision tree corresponds to a marketing object screening rule composed of multiple marketing object features. Extraction module 20 is used to extract the marketing object screening rules corresponding to the multiple decision trees and form a marketing object screening rule set; The determination module 30 is used to filter the marketing object dataset based on the marketing object filtering rule set in order to determine the target marketing objects that satisfy all filtering rules in the marketing object filtering rule set.

[0035] In one exemplary embodiment, each decision tree is a single-branch decision tree, and a single-branch decision tree contains multiple of the marketing object features.

[0036] Figure 4 This is a structural block diagram of a marketing target determination device according to another embodiment of the present invention, such as... Figure 4 As shown, the device includes Figure 3 In addition to all modules, it also includes: The stopping module 40 is used to perform pre-pruning during the generation of the decision tree so as to stop generating the decision tree when a third threshold is not met, wherein the third threshold includes at least one of the following: number of marketing targets, recall rate, gain, and marketing cost.

[0037] It should be noted that the above modules can be implemented by software or hardware. For the latter, they can be implemented in the following ways, but are not limited to: all the above modules are located in the same processor; or, the above modules are located in different processors in any combination.

[0038] To facilitate understanding of the technical solutions provided by this invention, detailed descriptions will be given below in conjunction with specific scenario embodiments.

[0039] This invention proposes a marketing gain decision tree model based on causal learning, designing two directly gain-based splitting criteria to improve the accuracy of gain prediction after splitting. The decision tree model generates multiple decision trees by maximizing the difference in probability distributions between the experimental and control groups (Y), and automatically extracts rules from each tree to form a rule set. Pre-pruning is used to avoid overfitting. The model features fast computation speed and strong interpretability.

[0040] In existing decision tree-based gain models, the gain at the split point is shown in equation (1): Formula (1) in, Let these represent the probability distributions of the experimental group and the control group, respectively. This represents a measure of the difference in distribution. Indicate using features Measurement of differences after splitting.

[0041] The marketing gain decision tree model based on causal learning proposed in this embodiment of the invention designs two methods for directly measuring gain. and : Formula (2) Formula (3) in, It is a rule Number of samples covered and The rules in the experimental group and the control group are respectively The probability distribution of Y over the covered samples. and This represents the initial probability distribution of Y in the experimental and control groups.

[0042] use and Replace gain calculation This yielded two different splitting criteria. and .

[0043] Based on the above formulas (2) and (3) and in combination with specific implementation scenarios, the following will elaborate on the construction of decision trees and the determination of target marketing objects.

[0044] This embodiment uses a histogram-based method to select feature splitting points. The specific process includes: for the dataset... and any feature , where the dataset This is a tabular dataset where each column represents a feature and each row represents a sample (i.e., a marketing target). First, calculate... Number of positive samples in the experimental group Total number of samples and the number of positive samples in the control group. Total number of samples And according to formula (4) we get and , Formula (4); Right now, ， .

[0045] Then features any eigenvalue in As a feature split point; calculate features The eigenvalues ​​are not greater than The number of positive samples in the experimental group Total number of samples and the number of positive samples in the control group. Total number of samples And according to formula (4), we get and ,in, R For eigenvalues ​​not greater than .

[0046] In this embodiment , , and Substitute into formula (2) or formula (3) above to calculate the result. Gain at the split point and corresponding .if The maximum value is greater than or equal to If the maximum value is , then the eigenpoint split point is The rule judge is The principle of splitting is The gain of this partition is The maximum value; conversely, the characteristic split point is The rule judge is The principle of splitting is The gain of this partition is The maximum value.

[0047] For example, features For the user's age, a feature value of 25 was selected as the feature split point, and the split point was calculated for samples where the user's age is no greater than 25. and and in the sample of users older than 25 and Furthermore, the calculations were performed. and ,Right now:

[0048] or, ; or,

[0049] Repeat the above process to obtain other age characteristic values. and ,Compare maximum value and The maximum value, if If the maximum value is larger, then only the left child node is retained; the splitting criterion is... Conversely, only the right child node is retained; the splitting criterion is... For example, in the examples above... The maximum value is ,like The maximum value is ,and Therefore, eigenvalue 25 is the optimal split point for the age feature. The splitting criterion is that the age is no greater than 25. The gain of this split is... .

[0050] Specifically, the pseudocode for the feature split point is as follows: Figure 5 As shown.

[0051] In this embodiment, the splitting criterion is directly based on gain, which more intuitively reflects the goal of maximizing the difference between the probability distributions of the experimental group and the control group regarding Y compared to KL divergence and Euclidean distance. Simultaneously, the rule set is extracted using an ensemble tree approach, reducing the correlation between rules and maintaining the good interpretability of the tree model.

[0052] Once the splitting criterion is determined, a decision tree can be constructed. In this embodiment, a top-down greedy algorithm is used to build a decision tree based on... and The pseudocode for constructing a decision tree and extracting rules, and automatically extracting rules from a single tree is as follows: Figure 6 As shown, the specific process is as follows: On each child node, the dataset is processed... Features of all samples on Perform a traversal and determine the features according to the above calculation method. The gain of each feature value is calculated, and the feature with the largest gain, its corresponding split point, and the splitting rule are selected as the optimal splitting rule for this child node. Samples not covered by the rule of this child node enter the next child node for a new round of rule extraction, ultimately resulting in multiple single-branch decision trees. Combining the splitting rules of multiple trees yields a rule set.

[0053] In this embodiment, the ensemble of decision trees enables the model to both improve prediction performance and provide explicit interpretability. (In the dataset...) The algorithm described above for building a single tree is used to obtain a decision tree and extract one rule. Then, the data not covered by this rule is used as a new dataset, and the algorithm is used again to generate another decision tree and extract rules. This process is repeated until the number of trees reaches a preset value. Combining all extracted rules forms a rule set. The pseudocode for this process is as follows: Figure 7 As shown in the figure. In this embodiment, a set of rules is automatically extracted by ensemble tree. The data used by different trees are separated, which avoids strong correlation between rules and maintains good interpretability.

[0054] In this embodiment, to prevent overfitting, the decision tree model is pre-pruned; that is, node splitting stops when a threshold is not met, thus halting the generation of the decision tree. This threshold includes the minimum number of samples for each child node. Minimum recall rate Minimum gain and marketing costs .

[0055] Embodiments of the present invention also provide a storage medium storing a computer program, wherein the computer program is configured to execute the steps in any of the above method embodiments when running.

[0056] Optionally, in this embodiment, the storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0057] Embodiments of the present invention also provide an electronic device including a memory and a processor, the memory storing a computer program and the processor being configured to run the computer program to perform the steps in any of the above method embodiments.

[0058] Optionally, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor and the input / output device is connected to the processor.

[0059] It is obvious to those skilled in the art that the modules or steps of the present invention described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, thereby storing them in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those presented herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0060] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for determining marketing targets, characterized in that, include: Based on the marketing object dataset and according to the decision tree algorithm that maximizes a specific target value, multiple decision trees are generated. Each decision tree corresponds to a marketing object selection rule composed of multiple marketing object features. Extract the marketing target selection rules corresponding to the multiple decision trees and form a marketing target selection rule set; The marketing object dataset is filtered based on the marketing object filtering rule set to determine the target marketing objects that meet all the filtering rules in the marketing object filtering rule set; The generation of multiple decision trees includes: Based on the decision tree algorithm for maximizing a specific target value, gain calculation is performed on all marketing object features in the marketing object dataset; The final split point corresponding to the first marketing object feature with the largest gain value is taken as the first child node of the first decision tree, and the splitting rule corresponding to the first marketing object feature is taken as the first screening rule of the first child node. According to the decision tree algorithm for maximizing a specific target value, gain calculation is performed on all marketing object features of the marketing object data that are not covered by the first screening rule to obtain other child nodes of the first decision tree, until the depth of the first decision tree reaches a first threshold. According to the decision tree algorithm for maximizing a specific target value, gain calculation is performed on the marketing object data that is not covered by the filtering rules of the first decision tree to obtain a second decision tree; According to the decision tree algorithm for maximizing a specific target value, gain calculation is performed on the marketing object data that is not covered by the filtering rules of the first decision tree and the second decision tree to obtain other decision trees until the number of decision trees reaches the second threshold. The gain is determined by the following formula: ， or , in, It is a rule Number of samples covered and The rules in the experimental group and the control group are respectively The probability distribution of Y over the covered samples. and This represents the initial probability distribution of Y in the experimental and control groups.

2. The method according to claim 1, characterized in that, in, Each decision tree is a single-branch decision tree, and a single-branch decision tree contains multiple of the aforementioned marketing object characteristics.

3. The method according to claim 1, characterized in that, The gain calculation for all marketing object features in the marketing object dataset also includes: For any of the marketing object features, each value of the marketing object feature is used as a split point for gain calculation; The value corresponding to the maximum gain is taken as the final split point of the marketing object feature, and the division rule corresponding to the maximum gain is taken as the screening rule of the marketing object feature.

4. The method according to claim 1, characterized in that, The method further includes: Pre-pruning is performed during the generation of the decision tree to stop generating the decision tree if a third threshold is not met, wherein the third threshold includes at least one of the following: number of marketing targets, recall rate, gain, and marketing cost.

5. A device for determining marketing targets, characterized in that, include: The generation module is used to generate multiple decision trees based on the marketing object dataset and according to the decision tree algorithm that maximizes a specific target value. Each decision tree corresponds to a marketing object selection rule composed of multiple marketing object features. The extraction module is used to extract the marketing object screening rules corresponding to the multiple decision trees and form a marketing object screening rule set; The determination module is used to filter the marketing object dataset based on the marketing object filtering rule set to determine the target marketing objects that satisfy all filtering rules in the marketing object filtering rule set; The generation module is further configured to: calculate the gain of all marketing object features in the marketing object dataset according to a decision tree algorithm that maximizes a specific target value; take the final split point corresponding to the first marketing object feature with the largest gain value as the first child node of the first decision tree, and take the splitting rule corresponding to the first marketing object feature as the first filtering rule of the first child node; calculate the gain of all marketing object features of the marketing object data not covered by the first filtering rule according to the decision tree algorithm that maximizes a specific target value, to obtain other child nodes of the first decision tree, until the depth of the first decision tree reaches a first threshold; calculate the gain of the marketing object data not covered by the filtering rule of the first decision tree according to the decision tree algorithm that maximizes a specific target value, to obtain a second decision tree; and calculate the gain of the marketing object data not covered by the filtering rules of the first and second decision trees according to the decision tree algorithm that maximizes a specific target value, to obtain other decision trees, until the number of decision trees reaches a second threshold. The gain is determined by the following formula: ， or , in, It is a rule Number of samples covered and The rules in the experimental group and the control group are respectively The probability distribution of Y over the covered samples. and This represents the initial probability distribution of Y in the experimental and control groups.

6. The apparatus according to claim 5, characterized in that, in, Each decision tree is a single-branch decision tree, and a single-branch decision tree contains multiple of the aforementioned marketing object characteristics.

7. The apparatus according to claim 5, characterized in that, include: A stopping module is used to perform pre-pruning during the generation of the decision tree so as to stop generating the decision tree if a third threshold is not met, wherein the third threshold includes at least one of the following: number of marketing targets, recall rate, gain, and marketing cost.

8. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, wherein the computer program is configured to execute the method described in any one of claims 1 to 4 when it is run.

9. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to run the computer program to perform the method as described in any one of claims 1 to 4.

Images