A data classification optimization method and system based on clustering algorithm

Abstract

The present invention discloses a data classification optimization method and system based on a clustering algorithm, which relates to the field of computer technology. The method divides historical marketing activities into K activity clusters, and then divides marketing populations into R marketing population clusters, and statistically calculates historical marketing effect data of each historical marketing. For each activity cluster, based on the historical marketing content data and historical marketing effect data of all corresponding historical marketing activities, a corresponding marketing effect prediction model is trained, and marketing content data of new marketing activities is collected. Based on the marketing content data, a target cluster to which the new marketing activity belongs is obtained; based on the marketing content data and the marketing effect prediction model of the target cluster, a predicted marketing effect is obtained; the subjectivity and limitations of manually designed content types are avoided, and the effect of marketing activities is improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a data classification optimization method and system based on a clustering algorithm. Background Art

[0002] In the current marketing environment, enterprises are facing fierce market competition. In order to effectively attract target customers and improve the marketing effect of marketing activities (marketing effect generally includes quantifiable indicators such as customer conversion rate and return on investment), enterprises need to evaluate the potential effect of marketing activities in advance to optimize marketing strategies. However, traditional marketing effect evaluation methods have some shortcomings.

[0003] Some companies use simple linear models or rule systems to predict marketing effectiveness, but these models usually have overly idealistic assumptions and find it difficult to capture the complex interactions between marketing activities, target audiences, and the marketing environment.

[0004] In addition, existing marketing effectiveness evaluation methods usually treat the content of marketing activities and the target audience separately, and are unable to effectively explore the intrinsic connection between the two, resulting in limited accuracy and generalization ability of prediction results.

[0005] The Chinese patent with the authorization announcement number CN110033156B discloses a method and device for determining the effect of a business activity, which uses the business activity feature information of users who participate in the business activity and users who do not participate in the business activity to calculate the propensity scores of users who participate in the business activity and users who do not participate in the business activity. Based on the calculated propensity scores, stratified sampling is performed to screen out a homogeneous user group of users who participate in the business activity, that is, to obtain users in the experimental group and users in the control group, and the effect of the business activity is determined based on the behavioral information of users in the experimental group and users in the control group; however, this solution tends to solve the situation where users have been determined and the number of users is small, and it is not applicable in an environment such as the Internet where users are unpredictable and the number of users is huge;

[0006] To this end, the present invention proposes a data classification optimization method and system based on a clustering algorithm. Summary of the invention

[0007] The present invention aims to solve at least one of the technical problems existing in the prior art. To this end, the present invention proposes a data classification optimization method and system based on a clustering algorithm, which avoids the subjectivity and limitations of manually designed content types and improves the effectiveness of marketing activities.

[0008] To achieve the above purpose, a data classification optimization method based on clustering algorithm is proposed, which includes the following steps:

[0009] Step 1: Collect historical marketing content data and historical marketing population characteristic data of N historical marketing activities in advance; N is the number of historical marketing activities collected;

[0010] Step 2: Use a clustering algorithm on the historical marketing content data of all historical marketing activities to divide the historical marketing activities into K activity clusters; K is the number of preset activity clusters;

[0011] Step 3: Divide the marketing population into R marketing population clusters for the historical marketing population characteristic data of all historical marketing activities; R is the number of preset marketing population clusters;

[0012] Step 4: For each historical marketing activity, based on the results of the marketing population cluster division and the historical marketing population characteristic data, calculate the historical marketing effect data;

[0013] Step 5: For each activity cluster, based on the historical marketing content data and historical marketing effect data of all corresponding historical marketing activities, train the corresponding marketing effect prediction model;

[0014] Step 6: Collect marketing content data of the new marketing campaign, and based on the marketing content data, obtain the target cluster to which the new marketing campaign belongs; based on the marketing content data and the marketing effect prediction model of the target cluster, obtain the predicted marketing effect;

[0015] The method of pre-collecting the historical marketing content data and historical marketing population characteristic data of N historical marketing activities is as follows:

[0016] Read the marketing content of all the historical business activities conducted from the database background of the social network platform as historical marketing content data, and collect the characteristic data of the crowd participating in the marketing in all the historical business activities as historical marketing crowd characteristic data;

[0017] The marketing content is a text description in the form of copywriting, a display picture in the form of pictures, or a combination of copywriting and pictures;

[0018] The demographic data includes public consumer information of each consumer in the marketing population;

[0019] The method of using a clustering algorithm on the historical marketing content data of all historical marketing activities to divide the historical marketing activities into K activity clusters includes the following steps:

[0020] Step 21: convert the historical marketing content data of each historical marketing campaign into a content feature vector using a pre-trained feature extraction model;

[0021] The content feature vector conversion method includes:

[0022] When the historical marketing content data is of the copywriting type, the conversion method of the content feature vector is:

[0023] Use word segmentation tools to split the text description into word sequences according to word boundaries;

[0024] Use the bag-of-words model to represent each word in the word sequence as a word frequency vector, calculate the weight of each word according to the word frequency inverse document frequency, and use the Word2Vec neural network to map the words to a vector space of fixed dimension according to the weight of each word; use the BERT model to convert the vector space set corresponding to the word sequence into a text feature vector according to the context words, and the text feature vector is the content feature vector;

[0025] When the historical marketing content data is of the image type, the content feature vector is converted as follows:

[0026] Scale the pixel values ​​of the displayed image to the interval [0, 1];

[0027] Use a pre-trained image processing neural network model to extract image feature vectors of displayed images;

[0028] When the historical marketing content data includes both text and image data, the text feature vector of the text class and the image feature vector of the image class are concatenated, and the concatenated vector is used as the content feature vector; each element in the content feature vector is a quantized content feature;

[0029] Step 22: Setting the clustering number parameter of the K-Means algorithm to K, and setting the first maximum number of iterations, and marking the program corresponding to the K-Means algorithm after the setting as the first clustering program;

[0030] Step 23: Input the content feature vectors of all historical marketing activities into the first clustering program to obtain K activity clusters;

[0031] The process of obtaining K activity clusters using the K-Means algorithm includes the following steps:

[0032] Step 231: Treat each content feature vector as an M-dimensional discrete point, wherein M is the dimension of the content feature vector, and the coordinate of the M-dimensional discrete point in each dimension corresponds to a content feature in the content feature vector;

[0033] Step 232: randomly selecting K M-dimensional discrete points as initial cluster centers, and the remaining M-dimensional discrete points as non-initial cluster centers;

[0034] Step 233: For each non-initial cluster center, calculate the Euclidean distance to each initial cluster center, and divide each M-dimensional discrete point into the cluster where the nearest initial cluster center is located;

[0035] Step 234: Calculate the average value of the content features corresponding to each dimension of the M-dimensional discrete points in each cluster, form a new M-dimensional discrete point with the average value of the content features corresponding to each dimension, and use the new M-dimensional discrete point as the new cluster center of the cluster;

[0036] Step 235: recalculate the Euclidean distance from each M-dimensional discrete point to each new cluster center, and re-divide each M-dimensional discrete point into the cluster where the nearest new cluster center is located;

[0037] Step 236: Repeat steps 234 to 235 until the M-dimensional discrete points in all the divided clusters no longer change, and mark the number of each cluster as k, k=1, 2, 3, ..., K; obtain all the M-dimensional discrete points contained in each cluster, and form a discrete point set from all the M-dimensional discrete points; the k-th activity cluster includes the M-dimensional space coordinates of the new cluster center of the k-th cluster, and the historical marketing activities corresponding to each M-dimensional discrete point in the k-th cluster;

[0038] The method of dividing the marketing population into R marketing population clusters based on the historical marketing population characteristic data of all historical marketing activities is as follows:

[0039] The public consumer information of each marketing participant in each historical marketing activity is combined into a set of consumer feature vectors; each consumer feature in the consumer feature vector corresponds to a feature in the public consumer information;

[0040] The cluster number parameter of the K-Means algorithm is set to R, and the second maximum number of iterations is set, and the program corresponding to the K-Means algorithm after the setting is marked as the second clustering program;

[0041] All consumer feature vectors are input into the second clustering procedure to obtain R marketing population clusters;

[0042] For each historical marketing activity, based on the division results of the marketing population clusters and the historical marketing population characteristic data, the method of calculating the historical marketing effect data is as follows:

[0043] Determine marketing effectiveness indicators in advance;

[0044] For each historical marketing activity, the indicator value of the marketing effect indicator is calculated based on the corresponding historical marketing population characteristic data and marketing population cluster;

[0045] For each activity cluster, based on the historical marketing content data and historical marketing effect data of all corresponding historical marketing activities, the method of training the corresponding marketing effect prediction model is as follows:

[0046] For the kth active cluster:

[0047] The content feature vectors of all historical marketing activities corresponding to the kth activity cluster are used as inputs of a marketing effect prediction model, and the marketing effect prediction model takes the predicted value of the marketing effect indicator corresponding to each historical marketing activity as output; the marketing effect prediction model takes the indicator value label of the marketing effect indicator corresponding to each historical marketing activity as a prediction target, takes the difference between the predicted value of the marketing effect indicator and the indicator value of the marketing effect indicator as a prediction error, and takes minimizing the mean of the sum of squares of the prediction errors of each historical marketing activity as a training target; the marketing effect prediction model is trained until the mean of the sum of squares of the prediction errors reaches convergence, and the training is stopped; the marketing effect prediction model is any one of the regression models;

[0048] The method of obtaining the activity cluster to which the new marketing activity belongs as the target cluster based on the marketing content data is as follows:

[0049] The marketing content data is converted into a content feature vector using a pre-trained feature extraction model, and the content feature vector converted from the marketing content data is represented as a target content vector;

[0050] The Euclidean distance between the target content vector and the center point of each activity cluster is calculated, and the activity cluster with the closest Euclidean distance is taken as the target cluster.

[0051] A data classification optimization system based on clustering algorithm is proposed, which includes a historical marketing data collection module, a clustering analysis module, a marketing effect prediction model training module and a marketing effect prediction module; wherein each module is electrically connected;

[0052] A historical marketing data collection module collects historical marketing content data and historical marketing population characteristic data of N historical marketing activities in advance, and sends the historical marketing content data and historical marketing population characteristic data to the cluster analysis module and the marketing effect prediction model training module;

[0053] The cluster analysis module uses a clustering algorithm on the historical marketing content data of all historical marketing activities to divide the historical marketing activities into K activity clusters; K is the number of preset activity clusters; for the historical marketing population feature data of all historical marketing activities, the marketing population is divided into R marketing population clusters; R is the number of preset marketing population clusters; and all activity clusters and marketing population clusters are sent to the marketing effect prediction model training module and the marketing effect prediction module;

[0054] The marketing effect prediction model training module calculates historical marketing effect data for each historical marketing activity based on the division results of the marketing population clusters and the historical marketing population feature data. For each activity cluster, the corresponding marketing effect prediction model is trained based on the historical marketing content data and historical marketing effect data of all corresponding historical marketing activities, and the marketing effect prediction model is sent to the marketing effect prediction module.

[0055] The marketing effect prediction module collects the marketing content data of the new marketing activity, and based on the marketing content data, obtains the target cluster to which the new marketing activity belongs; and obtains the predicted marketing effect based on the marketing content data and the marketing effect prediction model of the target cluster.

[0056] An electronic device is proposed, comprising: a processor and a memory, wherein the memory stores a computer program that can be called by the processor;

[0057] The processor executes the above-mentioned data classification optimization method based on clustering algorithm by calling the computer program stored in the memory.

[0058] A computer-readable storage medium is provided, on which a rewritable computer program is stored;

[0059] When the computer program is executed on a computer device, the computer device is enabled to execute the above-mentioned data classification optimization method based on clustering algorithm.

[0060] Compared with the prior art, the present invention has the following beneficial effects:

[0061] The present invention collects historical marketing content data and historical marketing population characteristic data of N historical marketing activities in advance, uses a clustering algorithm on the historical marketing content data of all historical marketing activities, divides the historical marketing activities into K activity clusters, and divides the marketing population into R marketing population clusters for the historical marketing population characteristic data of all historical marketing activities. For each historical marketing activity, based on the division results of the marketing population clusters and the historical marketing population characteristic data, the historical marketing effect data are statistically calculated. For each activity cluster, based on the historical marketing content data and historical marketing effect data of all corresponding historical marketing activities, the corresponding marketing effect prediction model is trained, the marketing content data of new marketing activities is collected, and based on the marketing content data, the target cluster to which the new marketing activity belongs is obtained; based on the marketing content data and the marketing effect prediction model of the target cluster, the predicted marketing effect is obtained; by clustering the marketing content and the marketing population respectively, and then training a model for predicting the marketing effect index for each type of marketing content, it is achieved that before the marketing activity is carried out, the planner of the marketing activity continuously modifies the marketing content to achieve the target marketing effect, thereby avoiding the subjectivity and limitations of manually designed content types and improving the effect of the marketing activity. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 This is a flow chart of a data classification optimization method based on a clustering algorithm in Example 1 of the present invention;

[0063] Figure 2 A module connection relationship diagram of a data classification optimization system based on a clustering algorithm in Example 2 of the present invention;

[0064] Figure 3 This is a schematic diagram of the structure of an electronic device in Embodiment 3 of the present invention;

[0065] Figure 4 This is a schematic diagram of the computer-readable storage medium structure in Example 4 of the present invention. DETAILED DESCRIPTION

[0066] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than 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.

[0067] Example 1

[0068] like Figure 1 As shown, a data classification optimization method based on a clustering algorithm comprises the following steps:

[0069] Step 1: Collect historical marketing content data and historical marketing population characteristic data of N historical marketing activities in advance; N is the number of historical marketing activities collected;

[0070] Step 2: Use a clustering algorithm on the historical marketing content data of all historical marketing activities to divide the historical marketing activities into K activity clusters; K is the number of preset activity clusters;

[0071] Step 3: Divide the marketing population into R marketing population clusters for the historical marketing population characteristic data of all historical marketing activities; R is the number of preset marketing population clusters;

[0072] Step 4: For each historical marketing activity, based on the results of the marketing population cluster division and the historical marketing population characteristic data, calculate the historical marketing effect data;

[0073] Step 5: For each activity cluster, based on the historical marketing content data and historical marketing effect data of all corresponding historical marketing activities, train the corresponding marketing effect prediction model;

[0074] Step 6: Collect marketing content data of the new marketing campaign, and based on the marketing content data, obtain the target cluster to which the new marketing campaign belongs; based on the marketing content data and the marketing effect prediction model of the target cluster, obtain the predicted marketing effect;

[0075] The method of pre-collecting the historical marketing content data and historical marketing population characteristic data of N historical marketing activities is as follows:

[0076] Read the marketing content of all the historical business activities conducted from the database background of the social network platform as historical marketing content data, and collect the characteristic data of the crowd participating in the marketing in all the historical business activities as historical marketing crowd characteristic data;

[0077] Specifically, the marketing content may be a text description in the form of a copy, or a display picture in the form of a picture, or a combination of a copy and a picture;

[0078] The population characteristic data includes the public information of each consumer in the marketing population; the public information of consumers includes but is not limited to age, gender, occupation, location, consumption level, etc.;

[0079] Furthermore, the method of using a clustering algorithm on the historical marketing content data of all historical marketing activities to divide the historical marketing activities into K activity clusters includes the following steps:

[0080] Step 21: convert the historical marketing content data of each historical marketing campaign into a content feature vector using a pre-trained feature extraction model;

[0081] Specifically, the content feature vector is transformed as follows:

[0082] When the historical marketing content data is of the copywriting type, the conversion method of the content feature vector is:

[0083] Use a word segmentation tool to segment the text description into word sequences according to word boundaries; the word segmentation tool may be a word segmentation tool such as Jieba;

[0084] Use the bag-of-words model to represent each word in the word sequence as a word frequency vector, calculate the weight of each word according to the word frequency inverse document frequency, and use the Word2Vec neural network to map the words to a vector space of fixed dimension according to the weight of each word; use the BERT model to convert the vector space set corresponding to the word sequence into a text feature vector according to the context words, and the text feature vector is the content feature vector;

[0085] When the historical marketing content data is of the image type, the content feature vector is converted as follows:

[0086] Scale the pixel values ​​of the displayed image to the interval [0, 1], and further perform transformations such as rotation and scaling on the displayed image to expand the data set;

[0087] Extracting a picture feature vector of the displayed picture using a pre-trained graphics processing neural network model; the graphics processing neural network model may be any one of VGG and ResNet, and it is understood that the picture feature vector may be obtained from an intermediate layer of the graphics processing neural network model;

[0088] After obtaining the image feature vector, the principal component analysis method can be used to project the high-dimensional image feature vector features into a low-dimensional feature vector. The projected low-dimensional feature vector can be used as the content feature vector.

[0089] When the historical marketing content data includes both text and image data, the text feature vector of the text class and the image feature vector of the image class (when the principal component analysis method is used, a low-dimensional feature vector is used) are concatenated, and the concatenated vector is used as the content feature vector; each element in the content feature vector is a quantized content feature;

[0090] Step 22: Setting the clustering number parameter of the K-Means algorithm to K, and setting the first maximum number of iterations, and marking the program corresponding to the K-Means algorithm after the setting as the first clustering program;

[0091] Step 23: Input the content feature vectors of all historical marketing activities into the first clustering program to obtain K activity clusters;

[0092] It can be understood that each historical marketing activity will be clustered into an activity cluster, and the historical marketing activities in each activity cluster have similar marketing product types. For example, historical marketing activities of marketing clothing, marketing food, marketing health products, and marketing fitness equipment will be clustered into the same activity cluster; and when K is large enough, each marketing product type can be further subdivided;

[0093] Specifically, the process of obtaining K activity clusters using the K-Means algorithm includes the following steps:

[0094] Step 231: Treat each content feature vector as an M-dimensional discrete point, wherein M is the dimension of the content feature vector, and the coordinate of the M-dimensional discrete point in each dimension corresponds to a content feature in the content feature vector;

[0095] Step 232: randomly selecting K M-dimensional discrete points as initial cluster centers, and the remaining M-dimensional discrete points as non-initial cluster centers;

[0096] Step 233: For each non-initial cluster center, calculate the Euclidean distance to each initial cluster center, and divide each M-dimensional discrete point into the cluster where the nearest initial cluster center is located;

[0097] Step 234: Calculate the average value of the content features corresponding to each dimension of the M-dimensional discrete points in each cluster, form a new M-dimensional discrete point with the average value of the content features corresponding to each dimension, and use the new M-dimensional discrete point as the new cluster center of the cluster;

[0098] Step 235: recalculate the Euclidean distance from each M-dimensional discrete point to each new cluster center, and re-divide each M-dimensional discrete point into the cluster where the nearest new cluster center is located;

[0099] Step 236: Repeat steps 234 to 235 until the M-dimensional discrete points in all the divided clusters no longer change, and mark the number of each cluster as k, k=1, 2, 3, ..., K; obtain all the M-dimensional discrete points contained in each cluster, and form a discrete point set from all the M-dimensional discrete points; the k-th activity cluster includes the M-dimensional space coordinates of the new cluster center of the k-th cluster, and the historical marketing activities corresponding to each M-dimensional discrete point in the k-th cluster;

[0100] Furthermore, the method of dividing the marketing population into R marketing population clusters for the historical marketing population characteristic data of all historical marketing activities is as follows:

[0101] The public consumer information of each marketing participant in each historical marketing activity is combined into a set of consumer feature vectors; each consumer feature in the consumer feature vector corresponds to a feature in the public consumer information;

[0102] The cluster number parameter of the K-Means algorithm is set to R, and the second maximum number of iterations is set, and the program corresponding to the K-Means algorithm after the setting is marked as the second clustering program;

[0103] All consumer feature vectors are input into the second clustering program to obtain R marketing population clusters. It can be understood that each marketing population cluster contains a number of consumers, and the consumers in each marketing population cluster belong to the same consumption feature, such as males of XX age group in XX region. Obviously, when R is larger, the consumption feature types divided into consumer populations are more detailed. The process of obtaining R marketing population clusters using the second clustering program is consistent with the operation process of the first clustering program, and the only difference is the input data, which will not be repeated in the present invention.

[0104] Furthermore, for each historical marketing activity, based on the division results of the marketing population clusters and the historical marketing population characteristic data, the method of calculating the historical marketing effect data is as follows:

[0105] Predetermine the marketing effect indicators; the marketing effect indicators may be the number of product purchasers, the type distribution of the purchasing consumer group, the ratio of the number of final order placements to the number of participants, the total purchase volume of the product, etc. The marketing effect indicators are determined according to actual needs;

[0106] For each historical marketing activity, the indicator value of the marketing effect indicator is calculated based on the corresponding historical marketing population characteristic data and marketing population clusters; for example, the Euclidean distance between the consumer characteristic vector of each consumer in the historical marketing population characteristic data and the center of each marketing population cluster is calculated, and each consumer is divided into the marketing population cluster with the closest Euclidean distance, so as to achieve the purpose of calculating the number and proportion relationship of each marketing population cluster among the consumer groups of the historical marketing activity;

[0107] Furthermore, for each activity cluster, based on the historical marketing content data and historical marketing effect data of all corresponding historical marketing activities, the corresponding marketing effect prediction model is trained in the following manner:

[0108] For the kth active cluster:

[0109] The content feature vectors of all historical marketing activities corresponding to the kth activity cluster are used as inputs of a marketing effect prediction model, and the marketing effect prediction model takes the predicted value of the marketing effect indicator corresponding to each historical marketing activity as output; the marketing effect prediction model takes the indicator value label of the marketing effect indicator corresponding to each historical marketing activity as a prediction target, takes the difference between the predicted value of the marketing effect indicator and the indicator value of the marketing effect indicator as a prediction error, and takes minimizing the mean of the sum of squares of the prediction errors of each historical marketing activity as a training target; the marketing effect prediction model is trained until the mean of the sum of squares of the prediction errors reaches convergence, and the training is stopped; the marketing effect prediction model is any one of the regression models; the regression model includes but is not limited to the SVR model, the polynomial regression model, etc.; the sum of squares of the prediction errors is a mean square error function;

[0110] Further, the method of collecting marketing content data of new marketing activities is consistent with the method of collecting historical marketing content data of historical marketing activities;

[0111] The method of obtaining the activity cluster to which the new marketing activity belongs as the target cluster based on the marketing content data is as follows:

[0112] The marketing content data is converted into a content feature vector using a pre-trained feature extraction model, and the content feature vector converted from the marketing content data is represented as a target content vector;

[0113] Calculate the Euclidean distance between the target content vector and the center point of each activity cluster, and take the activity cluster with the closest Euclidean distance as the target cluster;

[0114] Furthermore, the marketing effect prediction model based on the marketing content data and the target clusters obtains the predicted marketing effect in the following manner:

[0115] Inputting the content feature vector of the marketing content data into the marketing effect prediction model corresponding to the target cluster, and obtaining the predicted value of the marketing effect index of the marketing content data output by the marketing effect prediction model;

[0116] It is understandable that when the predicted value of the marketing effectiveness indicator fails to meet the expectations of the marketing activity planner, adjustments can be made to the marketing content before the marketing activity begins, and the marketing effectiveness indicator can be re-predicted until the expectations of the marketing activity planner are met; thereby automatically discovering different types of marketing content patterns, avoiding the subjectivity and limitations of manually designed content types, and improving the effectiveness of marketing activities.

[0117] Example 2

[0118] like Figure 2As shown, a data classification optimization system based on clustering algorithm includes a historical marketing data collection module, a clustering analysis module, a marketing effect prediction model training module and a marketing effect prediction module; wherein each module is electrically connected;

[0119] A historical marketing data collection module collects historical marketing content data and historical marketing population characteristic data of N historical marketing activities in advance, and sends the historical marketing content data and historical marketing population characteristic data to the cluster analysis module and the marketing effect prediction model training module;

[0120] The cluster analysis module uses a clustering algorithm on the historical marketing content data of all historical marketing activities to divide the historical marketing activities into K activity clusters; K is the number of preset activity clusters; for the historical marketing population feature data of all historical marketing activities, the marketing population is divided into R marketing population clusters; R is the number of preset marketing population clusters; and all activity clusters and marketing population clusters are sent to the marketing effect prediction model training module and the marketing effect prediction module;

[0121] The marketing effect prediction model training module calculates historical marketing effect data for each historical marketing activity based on the division results of the marketing population clusters and the historical marketing population feature data. For each activity cluster, the corresponding marketing effect prediction model is trained based on the historical marketing content data and historical marketing effect data of all corresponding historical marketing activities, and the marketing effect prediction model is sent to the marketing effect prediction module.

[0122] The marketing effect prediction module collects the marketing content data of the new marketing activity, and based on the marketing content data, obtains the target cluster to which the new marketing activity belongs; and obtains the predicted marketing effect based on the marketing content data and the marketing effect prediction model of the target cluster.

[0123] Example 3

[0124] Figure 3 1 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present application. Figure 3 As shown, according to another aspect of the present application, an electronic device 100 is also provided. The electronic device 100 may include one or more processors and one or more memories. The memories store computer readable codes, and when the computer readable codes are executed by one or more processors, a data classification optimization method based on a clustering algorithm as described above may be executed.

[0125] The method or device according to the embodiment of the present application can also be used by Figure 3 The electronic device architecture shown in FIG. Figure 3As shown, the electronic device 100 may include a bus 101, one or more CPUs 102, a ROM 103, a RAM 104, a communication port 105 connected to a network, an input / output component 106, a hard disk 107, etc. The storage device in the electronic device 100, such as the ROM 103 or the hard disk 107, may store a data classification optimization method based on a clustering algorithm provided in the present application.

[0126] Furthermore, the electronic device 100 may further include a user interface 108. Figure 3 The architecture shown is only exemplary and can be omitted according to actual needs when implementing different devices. Figure 3 One or more components of an electronic device are shown.

[0127] Example 4

[0128] Figure 4 Schematic diagram of the computer-readable storage medium structure provided by an embodiment of the present application. Figure 4 As shown, it is a computer-readable storage medium 200 according to an embodiment of the present application. Computer-readable instructions are stored on the computer-readable storage medium 200. When the computer-readable instructions are executed by the processor, a data classification optimization method based on a clustering algorithm according to an embodiment of the present application described with reference to the above figures can be executed. The computer-readable storage medium 200 includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory (cache), etc. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0129] In addition, according to the implementation of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium, the non-transitory machine-readable storage medium stores machine-readable instructions, the machine-readable instructions can be run by a processor to execute instructions corresponding to the method steps provided by the present application, and when the computer program is executed by a central processing unit (CPU), the above functions defined in the method of the present application are executed.

[0130] The methods, apparatuses, and devices of the present application may be implemented in many ways. For example, the methods, apparatuses, and devices of the present application may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is for illustration only, and the steps of the method of the present application are not limited to the order specifically described above unless otherwise specified. In addition, in some embodiments, the present application may also be implemented as programs recorded in a recording medium, which include machine-readable instructions for implementing the method according to the present application. Therefore, the present application also covers recording media storing programs for executing the method according to the present application.

[0131] In addition, the parts of the above-mentioned technical solutions provided in the embodiments of the present application that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail to avoid excessive redundancy.

[0132] The specific implementation modes as described above further describe the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation mode of the present invention and is not intended to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0133] The above preset parameters or preset thresholds are all set by technicians in this field according to actual conditions or obtained through large-scale data simulation.

[0134] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A data classification optimization method based on clustering algorithm, characterized in that: The following steps are involved: Step 1: Collect historical marketing content data and historical marketing population characteristic data of N historical marketing activities in advance; N is the number of historical marketing activities collected; Step 2: Use a clustering algorithm on the historical marketing content data of all historical marketing activities to divide the historical marketing activities into K activity clusters; K is the number of preset activity clusters; Step 3: Divide the marketing population into R marketing population clusters for the historical marketing population characteristic data of all historical marketing activities; R is the number of preset marketing population clusters; Step 4: For each historical marketing activity, based on the results of the marketing population cluster division and the historical marketing population characteristic data, calculate the historical marketing effect data; Step 5: For each activity cluster, based on the historical marketing content data and historical marketing effect data of all corresponding historical marketing activities, train the corresponding marketing effect prediction model; Step 6: Collect marketing content data of the new marketing campaign, and based on the marketing content data, obtain the target cluster to which the new marketing campaign belongs; based on the marketing content data and the marketing effect prediction model of the target cluster, obtain the predicted marketing effect; The method of pre-collecting the historical marketing content data and historical marketing population characteristic data of N historical marketing activities is as follows: Read the marketing content of all the historical business activities conducted from the database background of the social network platform as historical marketing content data, and collect the characteristic data of the crowd participating in the marketing in all the historical business activities as historical marketing crowd characteristic data; The marketing content is a text description in the form of copywriting, a display picture in the form of pictures, or a combination of copywriting and pictures; The demographic data includes public consumer information of each consumer in the marketing population; For each historical marketing activity, based on the division results of the marketing population clusters and the historical marketing population characteristic data, the method of calculating the historical marketing effect data is as follows: Determine marketing effectiveness indicators in advance; For each historical marketing activity, the indicator value of the marketing effect indicator is calculated based on the corresponding historical marketing population characteristic data and marketing population cluster; The marketing effectiveness indicators are the number of people who purchase the product, the type distribution of purchasing consumer groups, the proportion of the number of people who ultimately place orders to the number of participants, and the total purchase volume of the product.

2. The data classification optimization method based on clustering algorithm according to claim 1 is characterized in that: The method of using a clustering algorithm on the historical marketing content data of all historical marketing activities to divide the historical marketing activities into K activity clusters includes the following steps: Step 21: convert the historical marketing content data of each historical marketing campaign into a content feature vector using a pre-trained feature extraction model; Step 22: Set the clustering number parameter of the K-Means algorithm to K, set the first maximum number of iterations, and mark the program corresponding to the K-Means algorithm after the setting as the first clustering program; Step 23: Input the content feature vectors of all historical marketing activities into the first clustering program to obtain K activity clusters.

3. The data classification optimization method based on clustering algorithm according to claim 2 is characterized in that: The content feature vector conversion method includes: When the historical marketing content data is of the copywriting type, the conversion method of the content feature vector is: Use word segmentation tools to split the text description into word sequences according to word boundaries; Use the bag-of-words model to represent each word in a word sequence as a word frequency vector, calculate the weight of each word based on the word frequency inverse document frequency, and use the Word2Vec neural network to map the words to a vector space of fixed dimension according to the weight of each word; use the BERT model to convert the vector space set corresponding to the word sequence into a text feature vector based on the context words, and the text feature vector is the content feature vector; When the historical marketing content data is of the image type, the content feature vector is converted as follows: Scale the pixel values ​​of the displayed image to the interval [0, 1]; Use a pre-trained image processing neural network model to extract image feature vectors of displayed images; When the historical marketing content data includes both text and image categories, the text feature vector of the text category and the image feature vector of the image category are spliced, and the spliced ​​vector is used as the content feature vector; each element in the content feature vector is a quantized content feature.

4. The data classification optimization method based on clustering algorithm according to claim 3 is characterized in that: The process of obtaining K activity clusters using the K-Means algorithm includes the following steps: Step 231: Treat each content feature vector as an M-dimensional discrete point, wherein M is the dimension of the content feature vector, and the coordinate of the M-dimensional discrete point in each dimension corresponds to a content feature in the content feature vector; Step 232: randomly selecting K M-dimensional discrete points as initial cluster centers, and the remaining M-dimensional discrete points as non-initial cluster centers; Step 233: For each non-initial cluster center, calculate the Euclidean distance to each initial cluster center, and divide each M-dimensional discrete point into the cluster where the nearest initial cluster center is located; Step 234: Calculate the average value of the content features corresponding to each dimension of the M-dimensional discrete points in each cluster, form a new M-dimensional discrete point with the average value of the content features corresponding to each dimension, and use the new M-dimensional discrete point as the new cluster center of the cluster; Step 235: recalculate the Euclidean distance from each M-dimensional discrete point to each new cluster center, and re-divide each M-dimensional discrete point into the cluster where the nearest new cluster center is located; Step 236: Repeat steps 234-235 until the M-dimensional discrete points in all divided clusters no longer change, and mark the number of each cluster as k, k=1, 2, 3, ...K; obtain all the M-dimensional discrete points contained in each cluster, and form all the M-dimensional discrete points into a discrete point set; the k-th activity cluster includes the M-dimensional space coordinates of the new cluster center of the k-th cluster, and the historical marketing activities corresponding to each M-dimensional discrete point in the k-th cluster.

5. The data classification optimization method based on clustering algorithm according to claim 4 is characterized in that: The method of dividing the marketing population into R marketing population clusters based on the historical marketing population characteristic data of all historical marketing activities is as follows: The public consumer information of each marketing participant in each historical marketing activity is combined into a set of consumer feature vectors; each consumer feature in the consumer feature vector corresponds to a feature in the public consumer information; The clustering number parameter of the K-Means algorithm is set to R, and the second maximum number of iterations is set, and the program corresponding to the K-Means algorithm after the setting is marked as the second clustering program; All consumer feature vectors are input into the second clustering procedure to obtain R marketing population clusters.

6. The data classification optimization method based on clustering algorithm according to claim 5 is characterized in that: For each activity cluster, based on the historical marketing content data and historical marketing effect data of all corresponding historical marketing activities, the method of training the corresponding marketing effect prediction model is as follows: For the kth active cluster: The content feature vectors of all historical marketing activities corresponding to the kth activity cluster are used as the input of the marketing effect prediction model, and the marketing effect prediction model takes the predicted value of the marketing effect indicator corresponding to each historical marketing activity as the output; the marketing effect prediction model takes the indicator value label of the marketing effect indicator corresponding to each historical marketing activity as the prediction target, takes the difference between the predicted value of the marketing effect indicator and the indicator value of the marketing effect indicator as the prediction error, and takes minimizing the mean of the sum of squares of the prediction errors of each historical marketing activity as the training target; the marketing effect prediction model is trained until the mean of the sum of squares of the prediction errors reaches convergence and the training is stopped; the marketing effect prediction model is any one of the regression models.

7. The data classification optimization method based on clustering algorithm according to claim 6 is characterized in that: The method of obtaining the activity cluster to which the new marketing activity belongs as the target cluster based on the marketing content data is as follows: The marketing content data is converted into a content feature vector using a pre-trained feature extraction model, and the content feature vector converted from the marketing content data is represented as a target content vector; The Euclidean distance between the target content vector and the center point of each activity cluster is calculated, and the activity cluster with the closest Euclidean distance is taken as the target cluster.

8. A data classification optimization system based on a clustering algorithm, which is used to implement the data classification optimization method based on a clustering algorithm as described in any one of claims 1 to 7, characterized in that: It includes a historical marketing data collection module, a cluster analysis module, a marketing effect prediction model training module and a marketing effect prediction module; wherein each module is electrically connected; A historical marketing data collection module collects historical marketing content data and historical marketing population characteristic data of N historical marketing activities in advance, and sends the historical marketing content data and historical marketing population characteristic data to the cluster analysis module and the marketing effect prediction model training module; The cluster analysis module uses a clustering algorithm on the historical marketing content data of all historical marketing activities to divide the historical marketing activities into K activity clusters; K is the number of preset activity clusters; for the historical marketing population feature data of all historical marketing activities, the marketing population is divided into R marketing population clusters; R is the number of preset marketing population clusters; and all activity clusters and marketing population clusters are sent to the marketing effect prediction model training module and the marketing effect prediction module; The marketing effect prediction model training module calculates historical marketing effect data for each historical marketing activity based on the division results of the marketing population clusters and the historical marketing population feature data. For each activity cluster, the corresponding marketing effect prediction model is trained based on the historical marketing content data and historical marketing effect data of all corresponding historical marketing activities, and the marketing effect prediction model is sent to the marketing effect prediction module. The marketing effect prediction module collects the marketing content data of the new marketing activity, and based on the marketing content data, obtains the target cluster to which the new marketing activity belongs; based on the marketing effect prediction model of the marketing content data and the target cluster, obtains the predicted marketing effect.

9. An electronic device, characterized in that: include: processor and memory, wherein The memory stores a computer program that can be called by the processor; The processor executes the data classification optimization method based on clustering algorithm described in any one of claims 1 to 7 in the background by calling the computer program stored in the memory.

10. A computer-readable storage medium, characterized in that: A rewritable computer program is stored thereon; When the computer program is executed on a computer device, the computer device executes the data classification optimization method based on a clustering algorithm as described in any one of claims 1 to 7 in the background.

Images