Marketing scheme generation method and device

Abstract

The invention provides a marketing scheme generation method and apparatus. The method comprises the steps of obtaining customer data of a target customer; based on a preset division rule, dividing the client data into client positive data and client negative data; inputting the client forward data into a pre-trained marketing probability influence model to obtain a first marketing probability; inputting the initial marketing probability and the client negative data into a pre-trained marketing value evaluation model to obtain a second marketing probability; performing score conversion based on the second marketing probability to obtain a customer value score; and generating a marketing scheme for the target customer based on the customer value score. In the scheme, a large amount of customer data is divided into two types, the marketing probability influence model and the marketing value evaluation model are utilized to analyze the two types of customer data respectively to obtain the marketing probability of the customer, and finally, the marketing scheme of the customer is generated based on the marketing probability, so that the data analysis efficiency is improved, and the marketing efficiency of the customer is improved. And the accuracy of the marketing scheme is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a method and device for generating a marketing plan. Background Art

[0002] Data has become an important production factor in all industries and the core engine of current business growth and development decisions.

[0003] In the process of bank customer marketing, with the accelerated popularization of digitalization, banks need to face and process more and more customer data. Large amounts of useless, lengthy and redundant customer data not only affect the efficiency of data analysis, but also affect the accuracy of marketing plans.

[0004] Therefore, how to improve the efficiency of customer data analysis and ensure the accuracy of marketing plans is an issue that needs to be addressed urgently. Summary of the invention

[0005] In view of this, an embodiment of the present invention provides a method and device for generating a marketing plan, so as to achieve the purpose of improving data analysis efficiency and improving the accuracy of the marketing plan.

[0006] To achieve the above objectives, the embodiments of the present invention provide the following technical solutions:

[0007] A first aspect of an embodiment of the present invention discloses a method for generating a marketing plan, the method comprising:

[0008] Obtain customer data of target customers;

[0009] Based on a preset division rule, the customer data is divided into positive customer data and negative customer data;

[0010] Inputting the customer positive data into a pre-trained marketing probability influence model to obtain a first marketing probability;

[0011] Inputting the preliminary marketing probability and the customer negative data into a pre-trained marketing value assessment model to obtain a second marketing probability;

[0012] Performing score conversion based on the second marketing probability to obtain a customer value score;

[0013] Based on the customer value score, a marketing plan for the target customer is generated.

[0014] Preferably, after obtaining the customer data of the target customer, the method further comprises:

[0015] Preprocessing the customer data;

[0016] The preprocessing includes: deleting features with unique values ​​in the customer data, deleting features with missing values ​​greater than a target threshold in the customer data, data normalization, deleting irrelevant features in the customer data using regression coefficients, and deleting features with IV values ​​less than or equal to 0.02 in the customer data.

[0017] Preferably, the training process of the marketing probability influence model includes:

[0018] Constructing a marketing probability influence model to be trained, and defining a loss function applied to a regression task in the marketing probability influence model to be trained; the marketing probability influence model to be trained is an XGBoost model;

[0019] Dividing a pre-acquired first data set into a first training set and a first test set; the first data set includes customer positive data among customer data corresponding to a plurality of customers;

[0020] Based on the first training set and the grid search method, obtaining optimal parameters;

[0021] Using the optimal parameters and the first training set, training the marketing probability influence model to be trained to obtain a trained marketing probability influence model;

[0022] Using the first test set to evaluate the trained marketing probability impact model;

[0023] If the evaluation is passed, the trained marketing probability influence model is determined to be obtained.

[0024] Preferably, the training process of the marketing value assessment model includes:

[0025] Constructing a marketing value assessment model to be trained; the marketing value assessment model is a logistic regression model;

[0026] The pre-acquired second data set is divided into a second training set and a second test set; the second data set includes: inputting positive customer data in customer data corresponding to multiple customers into the trained marketing probability influence model to obtain the first marketing probability and negative customer data in customer data corresponding to multiple customers;

[0027] Based on the second training set and the stepwise regression method, the optimal sample variables are screened out;

[0028] Using the optimal sample variables and the second training set, training the marketing value evaluation model to be trained to obtain a trained marketing value evaluation model;

[0029] Using the second test set to evaluate the trained marketing value evaluation model;

[0030] If the evaluation is passed, the trained marketing value evaluation model is determined to be obtained.

[0031] Preferably, performing score conversion based on the second marketing probability to obtain a customer value score includes:

[0032] Based on a preset mapping relationship, the second marketing probability is converted into a customer value score.

[0033] Preferably, generating a marketing plan for the target customer based on the customer value score includes:

[0034] Based on the correspondence between the preset customer value score and the value grade, the value grade of the target customer is obtained;

[0035] Generate a marketing report that matches the value level of the target customer according to the preset report generation rules;

[0036] Obtaining marketing strategies for said marketing reports;

[0037] Using the positive customer data and the negative customer data, the customer portrait of the target customer is annotated to obtain the target customer portrait; the customer portrait is constructed based on the real-time collection of relevant indicators of the target customer;

[0038] Based on the customer value score, the marketing report, the target customer portrait and the marketing strategy, a marketing plan for the target customer is generated.

[0039] Preferably, the method further comprises:

[0040] Obtaining feedback data from the target customers during the process of marketing to the target customers based on the marketing plan;

[0041] Based on the feedback data, the target customer portrait in the marketing plan is updated.

[0042] A second aspect of an embodiment of the present invention discloses a device for generating a marketing plan, the device comprising:

[0043] An acquisition unit, used to acquire customer data of target customers;

[0044] A division unit, configured to divide the customer data into positive customer data and negative customer data based on a preset division rule;

[0045] A first prediction unit, configured to input the customer positive data into a pre-trained marketing probability influence model to obtain a first marketing probability;

[0046] A second prediction unit, used for inputting the preliminary marketing probability and the customer negative data into a pre-trained marketing value evaluation model to obtain a second marketing probability;

[0047] a conversion unit, configured to perform score conversion based on the second marketing probability to obtain a customer value score;

[0048] A plan generating unit is used to generate a marketing plan for the target customer based on the customer value score.

[0049] Preferably, the device further comprises:

[0050] A preprocessing unit, configured to preprocess the customer data of the target customer after acquiring the customer data of the target customer;

[0051] The preprocessing includes: deleting features with unique values ​​in the customer data, deleting features with missing values ​​greater than a target threshold in the customer data, data normalization, deleting irrelevant features in the customer data using regression coefficients, and deleting features with IV values ​​less than or equal to 0.02 in the customer data.

[0052] Preferably, the device further comprises:

[0053] A first training unit is used to construct a marketing probability influence model to be trained, and define a loss function applied to a regression task in the marketing probability influence model to be trained; the marketing probability influence model to be trained is an XGBoost model;

[0054] Dividing a pre-acquired first data set into a first training set and a first test set; the first data set includes customer positive data among customer data corresponding to a plurality of customers;

[0055] Based on the first training set and the grid search method, obtaining optimal parameters;

[0056] Using the optimal parameters and the first training set, training the marketing probability influence model to be trained to obtain a trained marketing probability influence model;

[0057] Using the first test set to evaluate the trained marketing probability impact model;

[0058] If the evaluation is passed, the trained marketing probability influence model is determined to be obtained.

[0059] Based on the method and device for generating a marketing plan provided by the above-mentioned embodiment of the present invention, customer data of a target customer is obtained; based on a preset division rule, the customer data is divided into customer positive data and customer negative data; the customer positive data is input into a pre-trained marketing probability influence model to obtain a first marketing probability; the preliminary marketing probability and the customer negative data are input into a pre-trained marketing value evaluation model to obtain a second marketing probability; a score conversion is performed based on the second marketing probability to obtain a customer value score; based on the customer value score, a marketing plan for the target customer is generated. In this solution, a large amount of customer data is divided into two categories, and the marketing probability influence model and the marketing value evaluation model are used to analyze the two categories of customer data respectively to obtain the customer's marketing probability, and finally a marketing plan for the customer is generated based on the marketing probability, thereby achieving the purpose of improving data analysis efficiency and improving the accuracy of marketing plans. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0061] Figure 1 A schematic diagram generated for a marketing scheme disclosed in an embodiment of the present invention;

[0062] Figure 2 A flowchart of a method for generating a marketing plan disclosed in an embodiment of the present invention;

[0063] Figure 3 This is a structural diagram of a device for generating a marketing plan disclosed in an embodiment of the present invention. DETAILED DESCRIPTION

[0064] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0065] In this application, the terms "comprises", "comprising" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device comprising the element.

[0066] As can be seen from the background technology, in the process of bank customer marketing, with the accelerated popularization of digitalization, banks need to face and process more and more customer data. Large amounts of useless, lengthy, and redundant customer data not only affect the efficiency of data analysis, but also affect the accuracy of marketing plans. In summary, how to improve the efficiency of customer data analysis and ensure the accuracy of marketing plans is an urgent problem to be solved.

[0067] Therefore, an embodiment of the present invention discloses a method and device for generating a marketing plan. In this solution, a large amount of customer data is divided into two categories. The two categories of customer data are analyzed separately using a marketing probability influence model and a marketing value evaluation model to obtain the customer's marketing probability, and finally a marketing plan for the customer is generated based on the marketing probability, thereby achieving the purpose of improving data analysis efficiency and improving the accuracy of the marketing plan.

[0068] like Figure 1 As shown, it is a schematic diagram of generating a marketing plan disclosed in an embodiment of the present invention.

[0069] Among them, omni-channel collects customer data on natural attributes and social attributes, including customer data obtained from multiple systems such as integrated views, multi-system fixed reports, and enterprise-level customer tag libraries. These customer data can be divided into two categories based on expert experience: data-increasing and cost-reducing. The data-increasing category refers to positive customer data, which is data that has a positive impact on customer evaluation, such as asset data and investment preference data; the cost-reducing category refers to negative customer data, which is data that has a negative impact on customer evaluation, such as data indicating that the customer is a dishonest debtor.

[0070] Then, the customer's internal bank operating data is taken as part of the customer data and pushed to the bank marketing platform together with the customer data collected above.

[0071] The bank marketing platform uses the marketing probability impact model and the marketing value assessment model to analyze the two types of customer data separately, obtain the customer's marketing probability, and finally generate a marketing plan for the customer based on the marketing probability for marketing personnel to market to the customer.

[0072] like Figure 2FIG. 1 is a flowchart of a method for generating a marketing plan disclosed in an embodiment of the present invention, comprising the following steps:

[0073] Step S101: Acquire customer data of target customers.

[0074] In step S101, customer data refers to customer data of natural attributes and social attributes collected through all channels, including customer data obtained from multiple systems such as integrated views, multi-system fixed reports, and enterprise-level customer tag libraries.

[0075] Step S102: based on a preset division rule, divide the customer data into customer positive data and customer negative data.

[0076] In step S102, the preset division rule is a division rule artificially formulated based on expert experience.

[0077] Data enhancement (customer positive data): asset allocation data: customer deposits, loans, funds, wealth management, insurance and other product data; transaction behavior data: transaction behavior, product preferences, transaction preferences, monthly activity, etc. Extended data: reference to merchant data such as smart relationship chain, Longxin Merchant, and company research reports; reference to the head office customer tag library and bank platform operation data.

[0078] Cost reduction (negative customer data): Introduce external data such as judicial freeze, anti-money laundering risk list, marketing SMS unsubscribe list, judicial cases, dishonest debtors, and tax arrears list.

[0079] In one embodiment, after obtaining the customer data, preprocessing is required, as follows:

[0080] Delete features with unique values ​​in customer data, delete features with missing values ​​greater than the target threshold in customer data, normalize data, use regression coefficients to delete irrelevant features in customer data, and delete features with IV values ​​less than or equal to 0.02 in customer data.

[0081] Among them, features with missing values ​​greater than 60% are unavailable; Data normalization: For list-type data, they are converted into binary data with values ​​of 0 or 1, categorical variables are reclassified, zero values ​​or mode interpolation is performed, and unique hot encoding is performed, date variables are split into year, month, and day, and numerical variables are interpolated by mean, mode, and the most recent observation value; IV value (Information Value) is an indicator for evaluating the predictive ability of features for target variables. It evaluates the importance of features by measuring the ability of features to distinguish target variables after binning.

[0082] Step S103: Input the customer's positive data into the pre-trained marketing probability influence model to obtain a first marketing probability.

[0083] The training process of the marketing probability influence model includes:

[0084] Step S201: construct a marketing probability influence model to be trained, and define a loss function applied to a regression task in the marketing probability influence model to be trained.

[0085] Among them, the marketing probability influence model to be trained is the XGBoost model. The XGBoost model performs well in processing large-scale data, can process in parallel, has high efficiency, and has good performance in prediction accuracy through iterative optimization model.

[0086] In step S201, since the task type of this training is set as a regression task and the model output is to predict the first marketing probability of the customer, a loss function applied to the regression task is defined in the marketing probability influence model to be trained.

[0087] Step S202: Divide the pre-acquired first data set into a first training set and a first test set.

[0088] The first data set includes customer positive data in customer data corresponding to multiple customers.

[0089] It should be noted that the process of obtaining the first training set is the same as steps S101 to S102, and the customer positive data in the first data set is the customer positive data that has been preprocessed, and the preprocessing process is consistent with the above.

[0090] Step S203: Obtaining optimal parameters based on the first training set and the grid search method.

[0091] In the specific implementation process of step S203, grid search (GridSearchCV) is used on the training set to improve the generalization ability of the model. GridSearchCV integrates automatic parameter adjustment and cross-validation functions, and defines a parameter grid (param_grid), which lists the parameters to be adjusted and all possible value combinations. Then a model is trained for each parameter combination, and cross-validation is used to evaluate the performance of the model, and finally the parameter combination with the highest average test score is selected as the optimal parameter.

[0092] For example, the optimal parameters are: colsample_bytree (the proportion of features randomly sampled when building each tree) is 0.8, learning_rate (learning rate) is 0.025, max_depth (maximum depth of the tree) is 2, and subsampe (the proportion of random sampling of samples for each tree) is 0.9.

[0093] Step S204: using the optimal parameters and the first training set to train the marketing probability influence model to be trained, and obtaining the trained marketing probability influence model.

[0094] Step S205: Use the first test set to evaluate the trained marketing probability impact model.

[0095] The final effect of the model is evaluated using indicators such as recall rate and AUC (Area Under the Curve). For example, if the recall rate is 65.16% and the AUC is 0.89, it is determined that the marketing probability after training affects the model evaluation.

[0096] Step S206: If the evaluation is passed, it is determined that the trained marketing probability influence model is obtained.

[0097] Step S104: Input the preliminary marketing probability and customer negative data into a pre-trained marketing value assessment model to obtain a second marketing probability.

[0098] It should be noted that the use of cost-reduction data can adjust the risk resistance of the model and correct the results of the model affected by the marketing probability. If a customer has a higher risk, then no matter how high the marketing probability is, the customer has no marketing value. Some more serious risk indicators, such as customers who have been judicially frozen, may directly have a "veto power."

[0099] Among them, the logistic regression model is selected for this task. In this task, the model input is the output result of the previous model marketing probability and the processed cost reduction data samples. For the list data, they are all converted into binary data with values ​​of 0 or 1, and the output is the predicted second marketing probability of the customer. The training process of the marketing value evaluation model is as follows:

[0100] Step S301: construct a marketing value assessment model to be trained.

[0101] Among them, the marketing value evaluation model is a logistic regression model.

[0102] Step S302: Divide the pre-acquired second data set into a second training set and a second test set.

[0103] Among them, the second data set includes: inputting positive customer data in the customer data corresponding to multiple customers into the trained marketing probability influence model to obtain the first marketing probability and negative customer data in the customer data corresponding to multiple customers, that is, cost reduction data.

[0104] Among them, the first marketing probability corresponds to the customer's negative data. For example, the customer data summary of customer A includes customer positive data and customer negative data. The customer's positive data is input into the marketing probability influence model to obtain the first marketing probability. Then the second data set includes customer A's first marketing probability and its corresponding customer negative data.

[0105] Step S304: Based on the second training set and the stepwise regression method, the optimal sample variables are screened out.

[0106] In the specific implementation process of step S304, first, all variables are traversed, each variable is modeled separately with the target variable, and the variable with the best model result is selected as the variable selected in the first round. On the basis of the variables selected in the first round, a second variable is added, and the remaining variables are traversed again, and the variable that can make the model result optimal is selected as the variable selected in the second round. The above process is repeated until no variable can significantly improve the model performance or prevent overfitting, and the optimal sample variable is obtained.

[0107] Here, the variable refers to a certain item of customer data, such as whether the customer is a dishonest person, which is a variable of 0 or 1.

[0108] Step S305: using the optimal sample variables and the second training set to train the marketing value evaluation model to be trained, and obtaining the trained marketing value evaluation model.

[0109] Step S305: Use the second test set to evaluate the trained marketing value evaluation model.

[0110] For example, the F1 value is used as an evaluation indicator to evaluate the final effect of the model. The final F1 value is 0.90, which indicates that the marketing value evaluation model after training performs well and passes the evaluation.

[0111] Step S306: If the evaluation is passed, it is determined that a trained marketing value evaluation model is obtained.

[0112] Step S105: Perform score conversion based on the second marketing probability to obtain a customer value score.

[0113] In the specific implementation process of step S105, the second marketing probability is converted into a customer value score based on a preset mapping relationship.

[0114] When the model predicts customer value, the output is usually a probability value, which indicates the size of the customer's marketing value. In order to more intuitively measure the marketing value of each customer, this probability value can be converted into a score, which is a numerical value that quantifies the customer's marketing value, such as a score between 0 and 100, where a higher score indicates a greater marketing value.

[0115] It should be noted that the second marketing probability takes a value between 0 and 1. According to the pre-established mapping relationship between the second marketing probability 0-1 and the score 0-100, the customer value score can be obtained according to the second marketing probability.

[0116] Step S106: Generate a marketing plan for target customers based on the customer value score.

[0117] The specific implementation process of step S106 includes the following steps:

[0118] Step S401: based on the preset correspondence between the customer value score and the value grade, the value grade of the target customer is obtained.

[0119] In step S401, customers can be divided into different levels according to customer value scores. For example, 0-30 points represent low-value customers, 30-70 points represent medium-value customers, and 70-100 points represent high-value customers. This classification helps banks quickly identify marketing targets that need to be focused on, improve marketing efficiency, and reduce ineffective marketing.

[0120] Step S402: Generate a marketing report that matches the value level of the target customer according to preset report generation rules.

[0121] In step S402, the report generation rules analyze in detail the customer characteristics of each value level, including their behavior patterns, preferences and needs, so as to generate a marketing report that matches the value level of the target customer.

[0122] Step S403: Obtain a marketing strategy for the marketing report.

[0123] For example, high-value customers may require more personalized services and product recommendations, while low-value customers may be more suited to generic marketing messages.

[0124] Step S404: using the positive customer data and the negative customer data, annotate the customer portrait of the target customer to obtain the target customer portrait.

[0125] Among them, the customer portrait is constructed based on the real-time collection of relevant indicators of the target customers. Specifically, customer data records related indicators will be collected at fixed time intervals. Based on the regular records, the dynamic change characteristics of customer behavior will be constructed to generate customer portraits. The data collected in the "increase data" part will be used to display the customer's product, transaction, upstream and downstream data dimensions. The data in the "reduce cost" part will be used to mark the portrait in red, generate early warning signs, remind the account manager of possible risk points, and obtain the target customer portrait.

[0126] Step S405: Generate a marketing plan for target customers based on the customer value score, marketing report, target customer portrait and marketing strategy.

[0127] In an embodiment of the present invention, the account manager can combine the customer value score, marketing report, marketing strategy and target customer portrait in the marketing plan to determine the marketing method for the target customer.

[0128] In one embodiment, the customer feedback obtained from the account manager's marketing is combined with the registered customer answering preferences, time preferences, channel preferences and other information to feed back the customer profile.

[0129] Specifically, the target customer feedback data is obtained during the process of marketing to the target customers based on the marketing plan; based on the feedback data, the target customer portrait in the marketing plan is updated.

[0130] Based on the method for generating a marketing plan disclosed in the above-mentioned embodiment of the present invention, customer data of target customers is obtained; based on a preset division rule, customer data is divided into customer positive data and customer negative data; customer positive data is input into a pre-trained marketing probability influence model to obtain a first marketing probability; preliminary marketing probability and customer negative data are input into a pre-trained marketing value evaluation model to obtain a second marketing probability; score conversion is performed based on the second marketing probability to obtain a customer value score; based on the customer value score, a marketing plan for the target customer is generated. In this solution, a large amount of customer data is divided into two categories, and the marketing probability influence model and the marketing value evaluation model are used to analyze the two categories of customer data respectively to obtain the customer's marketing probability, and finally a marketing plan for the customer is generated based on the marketing probability, thereby achieving the purpose of improving data analysis efficiency and improving the accuracy of the marketing plan.

[0131] Corresponding to the method for generating a marketing plan disclosed in the above embodiment of the present invention, as Figure 3 , which is a structural diagram of a marketing plan generation device disclosed in an embodiment of the present invention, including: an acquisition unit 301, a division unit 302, a first prediction unit 303, a second prediction unit 304, a conversion unit 305 and a plan generation unit 306.

[0132] The acquisition unit 301 is used to acquire the customer data of the target customer.

[0133] The division unit 302 is used to divide the customer data into customer positive data and customer negative data based on a preset division rule.

[0134] In one embodiment, the marketing plan generating device further includes:

[0135] A preprocessing unit, used for preprocessing the customer data after acquiring the customer data of the target customer;

[0136] Preprocessing includes: deleting features with unique values ​​in customer data, deleting features with missing values ​​greater than the target threshold in customer data, normalizing data, using regression coefficients to delete irrelevant features in customer data, and deleting features with IV values ​​less than or equal to 0.02 in customer data.

[0137] The first prediction unit 303 is used to input customer positive data into a pre-trained marketing probability influence model to obtain a first marketing probability.

[0138] In one embodiment, the marketing plan generating device further includes:

[0139] The first training unit is used to construct a marketing probability influence model to be trained, and define a loss function applied to a regression task in the marketing probability influence model to be trained; the marketing probability influence model to be trained is an XGBoost model;

[0140] Dividing a pre-acquired first data set into a first training set and a first test set; the first data set includes customer positive data among customer data corresponding to a plurality of customers;

[0141] Based on the first training set and the grid search method, the optimal parameters are obtained;

[0142] Using the optimal parameters and the first training set, training the marketing probability influence model to be trained to obtain a trained marketing probability influence model;

[0143] The trained marketing probability impact model is evaluated using the first test set;

[0144] If the evaluation is passed, the trained marketing probability influence model is determined to be obtained.

[0145] The second prediction unit 304 is used to input the preliminary marketing probability and the customer negative data into the pre-trained marketing value evaluation model to obtain the second marketing probability.

[0146] In one embodiment, the marketing plan generating device further includes:

[0147] The second training unit is used to construct a marketing value evaluation model to be trained; the marketing value evaluation model is a logistic regression model;

[0148] The pre-acquired second data set is divided into a second training set and a second test set; the second data set includes: inputting positive customer data in customer data corresponding to multiple customers into the trained marketing probability influence model to obtain the first marketing probability and negative customer data in customer data corresponding to multiple customers;

[0149] Based on the second training set and stepwise regression method, the optimal sample variables are screened out;

[0150] Using the optimal sample variables and the second training set, training the marketing value evaluation model to be trained to obtain a trained marketing value evaluation model;

[0151] Using the second test set to evaluate the trained marketing value assessment model;

[0152] If the evaluation is passed, the trained marketing value evaluation model is determined to be obtained.

[0153] The conversion unit 305 is used to perform score conversion based on the second marketing probability to obtain a customer value score.

[0154] In one embodiment, the conversion unit 305 is specifically configured to:

[0155] Based on the preset mapping relationship, the second marketing probability is converted into a customer value score.

[0156] The plan generating unit 306 is used to generate a marketing plan for target customers based on the customer value score.

[0157] In one embodiment, the solution generating unit 306 is specifically configured to:

[0158] Based on the correspondence between the preset customer value score and the value level, the value level of the target customer is obtained;

[0159] Generate marketing reports that match the value level of target customers according to preset report generation rules;

[0160] Get marketing strategies for marketing reports;

[0161] Using the positive and negative customer data, the customer portrait of the target customer is annotated to obtain the target customer portrait; the customer portrait is constructed based on the relevant indicators of the target customer collected in real time;

[0162] Generate marketing plans for target customers based on customer value scores, marketing reports, target customer portraits and marketing strategies.

[0163] In one embodiment, the marketing plan generating device further includes:

[0164] An updating unit, used for obtaining feedback data from target customers during the process of marketing to target customers based on the marketing plan;

[0165] Based on feedback data, update the target customer profile in the marketing plan.

[0166] Based on a device for generating a marketing plan disclosed in the above-mentioned embodiment of the present invention, customer data of target customers is obtained; based on a preset division rule, the customer data is divided into customer positive data and customer negative data; the customer positive data is input into a pre-trained marketing probability influence model to obtain a first marketing probability; the preliminary marketing probability and the customer negative data are input into a pre-trained marketing value evaluation model to obtain a second marketing probability; a score conversion is performed based on the second marketing probability to obtain a customer value score; based on the customer value score, a marketing plan for the target customer is generated. In this solution, a large amount of customer data is divided into two categories, and the marketing probability influence model and the marketing value evaluation model are used to analyze the two categories of customer data respectively to obtain the customer's marketing probability, and finally a marketing plan for the customer is generated based on the marketing probability, thereby achieving the purpose of improving data analysis efficiency and improving the accuracy of the marketing plan.

[0167] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can refer to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system or system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiment. The system and system embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without creative work.

[0168] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0169] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for generating a marketing plan, characterized in that: The method comprises: Obtain customer data of target customers; Based on a preset division rule, the customer data is divided into positive customer data and negative customer data; Inputting the customer positive data into a pre-trained marketing probability influence model to obtain a first marketing probability; Inputting the preliminary marketing probability and the customer negative data into a pre-trained marketing value assessment model to obtain a second marketing probability; Performing score conversion based on the second marketing probability to obtain a customer value score; Based on the customer value score, a marketing plan for the target customer is generated.

2. The method according to claim 1, characterized in that After acquiring the customer data of the target customer, the method further includes: Preprocessing the customer data; The preprocessing includes: deleting features with unique values ​​in the customer data, deleting features with missing values ​​greater than a target threshold in the customer data, data normalization, deleting irrelevant features in the customer data using regression coefficients, and deleting features with IV values ​​less than or equal to 0.02 in the customer data.

3. The method according to claim 1, characterized in that The training process of the marketing probability influence model includes: Constructing a marketing probability influence model to be trained, and defining a loss function applied to a regression task in the marketing probability influence model to be trained; the marketing probability influence model to be trained is an XGBoost model; Dividing a pre-acquired first data set into a first training set and a first test set; the first data set includes customer positive data among customer data corresponding to a plurality of customers; Based on the first training set and the grid search method, obtaining optimal parameters; Using the optimal parameters and the first training set, training the marketing probability influence model to be trained to obtain a trained marketing probability influence model; Using the first test set to evaluate the trained marketing probability impact model; If the evaluation is passed, the trained marketing probability influence model is determined to be obtained.

4. The method according to claim 3, characterized in that The training process of the marketing value assessment model includes: Constructing a marketing value assessment model to be trained; the marketing value assessment model is a logistic regression model; The pre-acquired second data set is divided into a second training set and a second test set; the second data set includes: inputting positive customer data in customer data corresponding to multiple customers into the trained marketing probability influence model to obtain the first marketing probability and negative customer data in customer data corresponding to multiple customers; Based on the second training set and the stepwise regression method, the optimal sample variables are screened out; Using the optimal sample variables and the second training set, training the marketing value evaluation model to be trained to obtain a trained marketing value evaluation model; Using the second test set to evaluate the trained marketing value evaluation model; If the evaluation is passed, the trained marketing value evaluation model is determined to be obtained.

5. The method according to claim 1, characterized in that The performing score conversion based on the second marketing probability to obtain a customer value score includes: Based on a preset mapping relationship, the second marketing probability is converted into a customer value score.

6. The method according to claim 1, characterized in that Generating a marketing plan for the target customer based on the customer value score includes: Based on the correspondence between the preset customer value score and the value grade, the value grade of the target customer is obtained; Generate a marketing report that matches the value level of the target customer according to the preset report generation rules; Obtaining marketing strategies for said marketing reports; Using the positive customer data and the negative customer data, the customer portrait of the target customer is annotated to obtain the target customer portrait; the customer portrait is constructed based on the real-time collection of relevant indicators of the target customer; Based on the customer value score, the marketing report, the target customer portrait and the marketing strategy, a marketing plan for the target customer is generated.

7. The method according to claim 6, characterized in that The method further comprises: Obtaining feedback data from the target customers during the process of marketing to the target customers based on the marketing plan; Based on the feedback data, the target customer portrait in the marketing plan is updated.

8. A device for generating a marketing plan, characterized in that: The device comprises: An acquisition unit, used to acquire customer data of target customers; A division unit, configured to divide the customer data into positive customer data and negative customer data based on a preset division rule; A first prediction unit, used for inputting the customer positive data into a pre-trained marketing probability influence model to obtain a first marketing probability; A second prediction unit, used for inputting the preliminary marketing probability and the customer negative data into a pre-trained marketing value evaluation model to obtain a second marketing probability; a conversion unit, configured to perform score conversion based on the second marketing probability to obtain a customer value score; A plan generating unit is used to generate a marketing plan for the target customer based on the customer value score.

9. The device according to claim 8, characterized in that The device also includes: A preprocessing unit, configured to preprocess the customer data of the target customer after acquiring the customer data of the target customer; The preprocessing includes: deleting features with unique values ​​in the customer data, deleting features with missing values ​​greater than a target threshold in the customer data, data normalization, deleting irrelevant features in the customer data using regression coefficients, and deleting features with IV values ​​less than or equal to 0.02 in the customer data.

10. The device according to claim 8, characterized in that The device also includes: A first training unit is used to construct a marketing probability influence model to be trained, and define a loss function applied to a regression task in the marketing probability influence model to be trained; the marketing probability influence model to be trained is an XGBoost model; Dividing a pre-acquired first data set into a first training set and a first test set; the first data set includes customer positive data among customer data corresponding to a plurality of customers; Based on the first training set and the grid search method, obtaining optimal parameters; Using the optimal parameters and the first training set, training the marketing probability influence model to be trained to obtain a trained marketing probability influence model; Using the first test set to evaluate the trained marketing probability impact model; If the evaluation is passed, the trained marketing probability influence model is determined to be obtained.

Images