Grouping rule recommendation method and device, electronic equipment and readable storage medium

Abstract

This invention provides a method and apparatus for grouping rule recommendation, an electronic device, and a readable storage medium. The method includes: preprocessing relational data in a recommendation topic to obtain processed relational data; merging the user's feature wide table data and the processed relational data, grouping them by relational target, and analyzing grouping features; adding the user-selected rule items to the grouping features to form grouping rules; generating a selected pool based on the grouping rules, matching the grouping rules in the selected pool with the original rules in a preset recommendation pool, and sorting the matched recommendation rules by recommendation degree; presenting recommendation rules whose recommendation degree ranking meets preset conditions; saving the added rules after the user references the recommendation rules and adding them to the feature wide table data, then returning to the merging step. This invention improves user engagement, scalability, flexibility, and process visibility by iterating and enhancing the accuracy of recommendations.

Description

Technical Field

[0001] This invention relates to the field of Internet technology applications, and in particular to a grouping rule recommendation method and apparatus, electronic device and readable storage medium. Background Technology

[0002] In recent years, the "infusion of intelligence" into traditional software has become a trend. In this process, the combination of human intervention and recommendations is increasingly accepted by users as a convenient, efficient, and intelligent way to enhance software capabilities.

[0003] The customer group management system uses the continuously accumulated customer base characteristics as its data source, allowing business users to select characteristics and set complex rules to ultimately generate customer groups with business characteristics, providing data support for business operation analysis, content recommendation, precision marketing, and other business operations.

[0004] The existing customer segmentation methods are roughly as follows:

[0005] Patent CN202010584585.1 provides a method for generating tag recommendations for target users using collaborative filtering algorithms based on tags selected by similar users. This method has low recommendation accuracy when there are few similar users, and it only recommends customer tags, neglecting more complex calculation rules, thus offering limited improvement in actual business operations. Patent CN201911421024.3 uses a tag-based user profile clustering algorithm to perform secondary grouping of the target group based on user-selected features. This method is relatively simplistic, belonging to a specific grouping method, and cannot flexibly adapt to the variability of business operations, leading to user distrust. Patent CN201... Patent 710429233.7 discloses a method for tag management and recommendation using knowledge graphs. However, it only provides a single tag without conditions or rule-related content, resulting in minimal improvement in rule configuration flexibility. It does not cover intervals, enumerations, or other related aspects, leading to low accuracy. Patent CN202010885050.8 provides a process for tagging users based on customer behavior intent. These tags constitute the user profile in this method. However, this method lacks segmentation rule definitions and rule recommendation steps, making it significantly different from our method. For segmented business users, it is essentially a black box. The relationship between this method and our method can be considered as the completed tags being the feature information upon which our method is based.

[0006] In summary, the methods described above use machine algorithms to model fixed business scenarios and are suitable for scenarios with relatively simple and fixed business operations. However, existing methods cannot provide generalized recommendations based on business segmentation rules, cannot give users more decision-making power, cannot lower the barrier to entry to allow business segmentation with minimal awareness, and cannot solve the problem of relying entirely on experience or algorithms in actual business operations. User participation, recommendation accuracy, scalability, and flexibility all need improvement. Summary of the Invention

[0007] This invention provides a method and apparatus for recommending grouping rules, an electronic device, and a readable storage medium to address the technical deficiencies existing in the prior art.

[0008] This invention provides a grouping rule recommendation method, comprising:

[0009] S1. Preprocess the relational data in the recommended topics to obtain processed relational data; the relational data includes customer identifiers, relational targets, and relational degrees.

[0010] S2. After merging the user's feature wide table data and the processed relational data, group them by relational target and analyze the group characteristics;

[0011] S3. Add the user-selected rule item to the grouping feature to form a clustering rule;

[0012] S4. Based on the grouping rules, generate a selected pool, match the grouping rules of the selected pool with the original rules in the preset recommendation pool, and sort the matched recommendation rules according to their recommendation degree.

[0013] S5. Present the recommendation rules that meet the preset conditions for ranking the recommendation degree;

[0014] S6. Save the added rule after the user references the recommendation rule and add it to the feature wide table data, then return to S2.

[0015] According to the clustering rule recommendation method of the present invention, the method further includes:

[0016] Based on the recommendation rules cited by the user, record the user's attention to the target of the reference relationship;

[0017] The recommendation score is obtained based on the attention level and the relationship level.

[0018] According to the clustering rule recommendation method of the present invention, before preprocessing the relation data in the recommendation topic to obtain the processed relation data, the method includes:

[0019] Create a recommendation topic, which includes: topic name, loss coefficient, and recommendation weight.

[0020] According to the clustering rule recommendation method of the present invention, the step of preprocessing the relation data in the recommendation topic to obtain processed relation data includes:

[0021] Pre-grouping customer identifiers based on the data in the relation target column: If a customer identifier appears multiple times in relation to the relation target during pre-grouping, the relation degree will decrease infinitely. Let the loss coefficient be q, and the formula is:

[0022]

[0023] The relational degree in the relational data is normalized to a value from 0 to 100, using the following formula:

[0024]

[0025] Where X is the degree of relation between the user and the relation target, Xn represents the degree of relation between the customer identifier and the relation target appearing n times; q represents the loss coefficient, a n X represents the degree of relation in the current relation data; Xscale represents the normalized degree of relation; Xmin represents the minimum degree of all relations; and Xmax represents the maximum degree of all relations.

[0026] According to the grouping rule recommendation method of the present invention, the analysis of grouping features includes:

[0027] Cluster and bin the grouped relational data, and record the features, bin intervals, user scale, and sum of interval relation degrees;

[0028] Alternatively, record features, enumerated values, customer size, and the sum of enumerated relation degrees;

[0029] Alternatively, it can record characteristics, user scale, and the sum of relationship levels.

[0030] According to the clustering rule recommendation method of the present invention, the step of matching the clustering rules of the selected pool with the original rules in the preset recommendation pool includes:

[0031] The selected pool's clustering rules are matched with the original rule items, original conditions, and original operators of the original rules in the recommendation pool. The rules that successfully match are used as recommendation rules. The clustering rules include clustering features and conditions, clustering operators, and clustering parameters.

[0032] According to the grouping rule recommendation method of the present invention, after presenting the recommendation rules whose recommendation degree ranking satisfies the preset conditions, the method includes:

[0033] Based on the recommendation rules cited by the user, the recommended content corresponding to the recommendation rules is displayed to the user.

[0034] The present invention also provides a grouping rule recommendation device, comprising:

[0035] The preprocessing module is used to preprocess the relational data in the recommended topics to obtain processed relational data; the relational data includes customer identifiers, relational targets, and relational degrees.

[0036] The grouping module is used to merge the user's feature wide table data and the processed relational data, group them by relational target, and analyze the group characteristics.

[0037] The clustering rule determination module is used to add the rule items selected by the user to the grouping features to form clustering rules;

[0038] The matching module is used to generate a selected pool based on the grouping rules, match the grouping rules of the selected pool with the original rules in the preset recommendation pool, and sort the matched recommendation rules according to the degree of recommendation.

[0039] The recommendation presentation module is used to present the recommendation rules that meet the preset conditions for recommendation ranking;

[0040] The update module is used to save the added rules after the user references the recommendation rules and add them to the feature wide table data, and then return to the grouping module.

[0041] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of any of the above-described clustering rule recommendation methods.

[0042] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the clustering rule recommendation method as described above.

[0043] This invention, while retaining the flexibility of the rules, supplements the rule recommendation capability, enabling users to see clear prompts at any time when segmenting business processes and to adjust them according to their actual situation. The adjusted content can serve as the basis for the next prompt, and this iteration continuously improves the accuracy of rule recommendations. It can effectively and accurately recommend multiple relationship targets within different topics to users, facilitating rapid recommendations. Each selection of topic, relationship target, and rule will serve as the basis for the next recommendation ranking, iterating and strengthening the accuracy of recommendations, thereby improving user engagement, scalability, flexibility, and process visibility. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0045] Figure 1 This is a flowchart illustrating the grouping rule recommendation method provided by the present invention;

[0046] Figure 2 This is a schematic diagram of the grouping rule recommendation device provided by the present invention;

[0047] Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0049] The following is combined Figure 1 This invention describes a grouping rule recommendation method, the method comprising:

[0050] S1. Preprocess the relational data in the recommended topics to obtain processed relational data; the relational data includes customer identifiers, relational targets, and relational degrees.

[0051] Preprocessing the relational data in the recommended topics can be done periodically according to a set cycle. Preprocessing using a specific method is characterized by unifying the relational degree between customers and relational targets to the same range dimension through a specific method, laying the foundation for better recommendation results.

[0052] The customer identifier can be coded information such as an ID card or mobile phone number.

[0053] The relationship objective could be browsing data or ordering data.

[0054] Relationship level can include factors such as importance and liking.

[0055] The preprocessed relational data is roughly in the form shown in Table 1 below:

[0056] Table 1

[0057]

[0058] S2. After merging the user's feature wide table data and the processed relational data, group them by relational target and analyze the group characteristics;

[0059] Each recommendation starts from S2. The user's feature wide table data format is roughly as shown in Table 2 below, where 1 in the "VIP" column indicates "yes".

[0060] Table 2

[0061]

[0062] The processing results of S1 are merged with the customer feature wide table data, then grouped by relationship target, and the grouping characteristics are analyzed. The rules after this processing are roughly as shown in Table 3 below:

[0063] Table 3

[0064] primary key Relationship Goals feature Identifier / Range / Enumeration scale Relationship degree T0001 5G 128 package VIP status 1 200,000 12900 T0002 5G 128 package Expenditure Balance [500-800] 170,000 3459 T0003 5G 128 package Place of origin Beijing 30,000 540 T0004 5G 128 package Place of origin Shanghai 40,000 455 T0005 5G 128 package Place of origin Guizhou 0 million 0

[0065] S3. Add the user-selected rule item to the grouping feature to form a clustering rule;

[0066] When users perform business operations, clear prompts are provided in the relevant area for reference when selecting rule items. Based on historical data within the recommended topic (such as order history, browsing history, and preferences), an initial recommendation is formed without iteration, according to the topic's contextual goals. This initial recommendation then generates original recommendation rules based on specific algorithms and weight settings.

[0067] Then, the segmentation rules created by business users are combined with the original recommendation rules for a secondary recombination. Due to the interdependent nature of rules, this process is repeated multiple times to find the segmentation rules (basic features, feature parameters, operators, etc.) most likely to be needed by users. Here, intelligent prompts can be provided to the user for final selection; this is a semi-automatic mode with step-by-step prompts and customer confirmation. In other words, if the user finds that the recommended rules are inconsistent with their desired settings, they can actively make a selection themselves.

[0068] Specifically, the selected features can be gender, age, monthly data usage, etc., and the set conditions can be male (gender condition), 25-30 (age condition), 300-400M (monthly data usage condition), etc.; the modified operators can be AND, OR, exclusion, parentheses, etc.

[0069] Rule items are divided into three categories: features and conditions, operators (AND, NOT), and rules. The "rule item" type includes operators and operational conditions; the specific definition of the "value" type includes feature IDs and rule symbols such as AND, OR, and NOT; the "relationship degree" provides the basis for rule matching and adopts a specific recommendation grouping rule storage format and content definition. Users are given choices, but they can also choose not to select. If no choice is made, the user-selected rule item does not need to be added. When the size of the matched customer group exceeds a preset value, it is saved as a feature condition, and the operator is saved as "AND"; features and conditions with a proportion of 0 are saved as rule items, and the relationship is saved as "NOT"; the relationship degree is inherited.

[0070] After processing by S3, the general form is shown in Table 4 below:

[0071] Table 4

[0072] Rule Item value Relationship degree Operators ( 12900 Feature conditions T0001 12900 Operators and 540 Feature conditions T0002 540 Operators and 455 Feature conditions T0003 455 Operators ) 12900 Operators not 12900 Feature conditions T0004 12900

[0073] S4. Based on the grouping rules, generate a selected pool, match the grouping rules of the selected pool with the original rules in the preset recommendation pool, and sort the matched recommendation rules according to their recommendation degree.

[0074] When a customer searches for or selects a rule, that tag is added to the selected pool. Rules in the selected pool are matched with rules in the recommendation pool, and then sorted by recommendation score. Recommendation score = Relationship score + (Topic recommendation weight * Attention score) / 10 + Scale percentage * 100. The topic recommendation weight is preset. Attention score refers to the user's level of attention to the referenced relationship target. Attention score is a counter that increments by 1 each time the relationship data is used; it can be used as part of the calculation conditions.

[0075] S5. Present the recommendation rules that meet the preset conditions for ranking the recommendation degree;

[0076] Based on existing rules in the rule pool, the top 20 recommended rules (or recommended content) can be displayed in descending order of recommendation popularity. Recommended content is a combination of several recommended rules. 2. Recommended content includes: topic name, recommendation goal, rule description, recommendation popularity, and predicted number of users; for example, recommended content may include:

[0077] Topic Title: App Recommendations

[0078] Recommended target: Honor of Kings preference

[0079] Rule description: [Online time > 5 hours] and [Age < 30 years old]

[0080] Number of people predicted: 234,405

[0081] Recommendation rating: 8432

[0082] Additional options include: partial citation and full citation; recommended content will appear in the recommendation section when the user selects features, fills in conditions, or modifies operators.

[0083] S6. Save the added rule after the user references the recommendation rule and add it to the feature wide table data, then return to S2.

[0084] Users can continue to add features and parameters after referencing the recommendation rules, modifying AND and NOT logic rules to achieve a specific business goal. Once achieved, these new clustering rules are saved. The new rules are recorded as identifying features and their corresponding rule IDs, which are then added back to the features in step three, continuously optimizing the rule recommendation capabilities. Saved clustering rules generate identifying features, preparing for more precise narrowing of the interval range in the future. If no rules are added at this point, the process can directly return to step S2.

[0085] In the process described above, rule recommendation acts as an assistant, continuously updating the user's current choices. If the user ignores all selections, the rules will be optimized periodically, and the user's choices will be weighted in the algorithm and recommended. Unless manually stopped, this recommendation strategy can be executed repeatedly from S2 to S6, and the process will not end.

[0086] This invention, while retaining the flexibility of the rules, supplements the rule recommendation capability, enabling users to see clear prompts at any time when segmenting business processes and to adjust them according to their actual situation. The adjusted content can serve as the basis for the next prompt, and this iteration continuously improves the accuracy of rule recommendations. It can effectively and accurately recommend multiple relationship targets within different topics to users, facilitating rapid recommendations. Each selection of topic, relationship target, and rule will serve as the basis for the next recommendation ranking, iterating and strengthening the accuracy of recommendations, thereby improving user engagement, scalability, flexibility, and process visibility.

[0087] According to the clustering rule recommendation method of the present invention, the method further includes:

[0088] Based on the recommendation rules cited by the user, record the user's attention to the target of the reference relationship;

[0089] The recommendation score is obtained based on the attention level and the relationship level.

[0090] Business users can select recommended topics based on their intentions and view the top 20 recommended rules. Business users can selectively introduce all or some of the rules, and the system will record the user's attention to the cited topic and the target of the cited relationship. Each time a user cites recommended content, the system will also record the user's attention to this topic or relationship data, which will be used for subsequent recommendation weighting.

[0091] According to the clustering rule recommendation method of the present invention, before preprocessing the relation data in the recommendation topic to obtain the processed relation data, the method includes:

[0092] Create a recommendation topic, which includes: topic name, loss coefficient (1-100), and recommendation weight (1-100).

[0093] Recommended topics are created and pre-stored in a registration history recommendation relationship data model. The data in this model can be browsing data, order data, and the relationship degree can be abstract indicators such as importance, liking, and recommendation degree. This data will eventually be processed and normalized to a uniform range.

[0094] According to the clustering rule recommendation method of the present invention, the step of preprocessing the relation data in the recommendation topic to obtain processed relation data includes:

[0095] Pre-grouping customer identifiers based on the data in the relation target column: If a customer identifier appears multiple times in relation to the relation target during pre-grouping, the relation degree will decrease infinitely. Let the loss coefficient be q, and the formula is:

[0096]

[0097] The relational degree in the relational data is normalized to a value from 0 to 100, using the following formula:

[0098]

[0099] Where X is the relation degree of the grouped users to the relation target, Xn represents the relation degree of the customer identifier appearing n times to the relation target; q represents the loss coefficient (adjustable, default is 10), a n X represents the degree of relation in the current relation data; Xscale represents the normalized degree of relation; Xmin represents the minimum degree of all relations; and Xmax represents the maximum degree of all relations.

[0100] According to the grouping rule recommendation method of the present invention, the analysis of grouping features includes:

[0101] The CART algorithm clusters and bins the grouped relational data, recording features, bin intervals, user size, and the sum of interval relation degrees (continuous features).

[0102] Alternatively, classification algorithms can be used to record features, enumerated values, customer size, and the sum of enumerated relation degrees; (enumerated features).

[0103] Or record characteristics, user scale, and total relationship degree (identifier-type characteristics).

[0104] After the recommended topic is launched, the system will read the relationship data and perform similarity matching with customer characteristics. Based on the feature type, continuous values, enumerated values, multiple intervals, the corresponding relationship degree sum, and customer group size will be saved.

[0105] According to the clustering rule recommendation method of the present invention, the step of matching the clustering rules of the selected pool with the original rules in the preset recommendation pool includes:

[0106] The selected pool's clustering rules are matched with the original rule items, original conditions, and original operators of the original rules in the recommendation pool. The rules that successfully match are used as recommendation rules. The clustering rules include clustering features and conditions, clustering operators, and clustering parameters.

[0107] Based on the consistent grouping rules of this system, the items are placed in the recommendation pool to await matching. The sum of the relationship degree and the customer group size will be weighted to form a new recommendation degree.

[0108] According to the grouping rule recommendation method of the present invention, after presenting the recommendation rules whose recommendation degree ranking satisfies the preset conditions, the method includes:

[0109] Based on the recommendation rules cited by the user, the recommended content corresponding to the recommendation rules is displayed to the user.

[0110] See Figure 2 The following describes the clustering rule recommendation device provided by the present invention. The clustering rule recommendation device described below can be referred to in correspondence with the clustering rule recommendation method described above. The clustering rule recommendation device includes:

[0111] Preprocessing module 10 is used to preprocess the relational data in the recommended topics to obtain processed relational data; the relational data includes customer identifier, relational target, and relational degree;

[0112] Preprocessing the relational data in the recommended topics can be done periodically according to a set cycle. Preprocessing using a specific method is characterized by unifying the relational degree between customers and relational targets to the same range dimension through a specific method, laying the foundation for better recommendation results.

[0113] Grouping module 20 is used to merge the user's feature wide table data and the processed relation data, group them by relation target, and analyze the grouping characteristics;

[0114] Each recommendation starts from grouping module 20, which merges the processing results of preprocessing module 10 with the customer feature wide table data, then groups the data by relationship target and analyzes the group features.

[0115] The grouping rule determination module 30 is used to add the rule items selected by the user to the grouping features to form grouping rules;

[0116] When users perform business operations, clear prompts are provided in the relevant area for reference when selecting rule items. Based on historical data within the recommended topic (such as order history, browsing history, and preferences), an initial recommendation is formed without iteration, according to the topic's contextual goals. This initial recommendation then generates original recommendation rules based on specific algorithms and weight settings.

[0117] Then, the segmentation rules created by business users are combined with the original recommendation rules for a secondary recombination. Due to the interdependent nature of rules, this process is repeated multiple times to find the segmentation rules (basic features, feature parameters, operators, etc.) most likely to be needed by users. Here, intelligent prompts can be provided to the user for final selection; this is a semi-automatic mode with step-by-step prompts and customer confirmation. In other words, if the user finds that the recommended rules are inconsistent with their desired settings, they can actively make a selection themselves.

[0118] The matching module 40 is used to generate a selected pool based on the grouping rules, match the grouping rules of the selected pool with the original rules in the preset recommendation pool, and sort the matched recommendation rules according to the recommendation degree.

[0119] When a customer searches for or selects a rule, that tag is added to the selected pool. Rules in the selected pool are matched with rules in the recommendation pool, and then sorted by recommendation score. Recommendation score = Relationship score + (Topic recommendation weight * Attention score) / 10 + Scale percentage * 100. The topic recommendation weight is preset. Attention score refers to the user's level of attention to the referenced relationship target. Attention score is a counter that increments by 1 each time the relationship data is used; it can be used as part of the calculation conditions.

[0120] The recommendation presentation module 50 is used to present the recommendation rules that meet the preset conditions for recommendation degree ranking;

[0121] Based on existing rules in the rule pool, the top 20 recommended rules (or recommended content) can be displayed in descending order of recommendation popularity. Recommended content is a combination of several recommended rules. 2. Recommended content includes: topic name, recommendation goal, rule description, recommendation popularity, and predicted number of users.

[0122] The update module 60 is used to save the added rules after the user references the recommendation rules and add them to the feature wide table data, and then return to the grouping module 20.

[0123] Users can continue to add features and parameters after referencing recommendation rules, modify AND and NOT logic rules, and save them as new clustering rules once a certain business goal is achieved. The new rules will be recorded as identifying features and their corresponding rule IDs, and this record will be added back to the features from step three, continuously optimizing the rule recommendation capabilities. Saved clustering rules will generate identifying features, preparing for more precise narrowing of intervals in the future. If no rules are added at this point, users can directly return to grouping module 20.

[0124] According to the grouping rule recommendation device of the present invention, the device further includes a recommendation degree determination module, the recommendation degree determination module being used for:

[0125] Based on the recommendation rules cited by the user, record the user's attention to the target of the reference relationship;

[0126] The recommendation score is obtained based on the attention level and the relationship level.

[0127] Business users can select recommended topics based on their intentions and view the top 20 recommended rules. Business users can selectively introduce all or some of the rules, and the system will record the user's attention to the cited topic and the target of the cited relationship. Each time a user cites recommended content, the system will also record the user's attention to this topic or relationship data, which will be used for subsequent recommendation weighting.

[0128] According to the grouping rule recommendation device of the present invention, the device further includes a recommendation topic creation module, the recommendation topic creation module being used for:

[0129] Create a recommendation topic, which includes: topic name, loss coefficient (1-100), and recommendation weight (1-100).

[0130] Recommended topics are created and pre-stored in a registration history recommendation relationship data model. The data in this model can be browsing data, order data, and the relationship degree can be abstract indicators such as importance, liking, and recommendation degree. This data will eventually be processed and normalized to a uniform range.

[0131] According to the grouping rule recommendation apparatus of the present invention, wherein when the preprocessing module 10 is used to:

[0132] Pre-grouping customer identifiers based on the data in the relation target column: If a customer identifier appears multiple times in relation to the relation target during pre-grouping, the relation degree will decrease infinitely. Let the loss coefficient be q, and the formula is:

[0133]

[0134] The relational degree in the relational data is normalized to a value from 0 to 100, using the following formula:

[0135]

[0136] Where X is the relation degree of the grouped users to the relation target, Xn represents the relation degree of the customer identifier appearing n times to the relation target; q represents the loss coefficient (adjustable, default is 10), a n X represents the degree of relation in the current relation data; Xscale represents the normalized degree of relation; Xmin represents the minimum degree of all relations; and Xmax represents the maximum degree of all relations.

[0137] According to the grouping rule recommendation apparatus of the present invention, the grouping module 20 is used for:

[0138] The CART algorithm clusters and bins the grouped relational data, recording features, bin intervals, user size, and the sum of interval relation degrees (continuous features).

[0139] Alternatively, classification algorithms can be used to record features, enumerated values, customer size, and the sum of enumerated relation degrees; (enumerated features).

[0140] Or record characteristics, user scale, and total relationship degree (identifier-type characteristics).

[0141] After the recommended topic is launched, the system will read the relationship data and perform similarity matching with customer characteristics. Based on the feature type, continuous values, enumerated values, multiple intervals, the corresponding relationship degree sum, and customer group size will be saved.

[0142] According to the grouping rule recommendation apparatus of the present invention, the matching module 40 is used for:

[0143] The selected pool's clustering rules are matched with the original rule items, original conditions, and original operators of the original rules in the recommendation pool. The rules that successfully match are used as recommendation rules. The clustering rules include clustering features and conditions, clustering operators, and clustering parameters.

[0144] Based on the consistent grouping rules of this system, the items are placed in the recommendation pool to await matching. The sum of the relationship degree and the customer group size will be weighted to form a new recommendation degree.

[0145] According to the grouping rule recommendation device of the present invention, the device further includes a display module, the display module being used for:

[0146] Based on the recommendation rules cited by the user, the recommended content corresponding to the recommendation rules is displayed to the user.

[0147] Figure 3 A schematic diagram of the physical structure of an electronic device is provided. This electronic device may include: a processor 310, a communication interface 320, a memory 330, and a communication bus 340. The processor 310, communication interface 320, and memory 330 communicate with each other via the communication bus 340. The processor 310 can invoke logical instructions stored in the memory 330 to execute a clustering rule recommendation method. This method includes:

[0148] S1. Preprocess the relational data in the recommended topics to obtain processed relational data; the relational data includes customer identifiers, relational targets, and relational degrees.

[0149] S2. After merging the user's feature wide table data and the processed relational data, group them by relational target and analyze the group characteristics;

[0150] S3. Add the user-selected rule item to the grouping feature to form a clustering rule;

[0151] S4. Based on the grouping rules, generate a selected pool, match the grouping rules of the selected pool with the original rules in the preset recommendation pool, and sort the matched recommendation rules according to their recommendation degree.

[0152] S5. Present the recommendation rules that meet the preset conditions for ranking the recommendation degree;

[0153] S6. Save the added rule after the user references the recommendation rule and add it to the feature wide table data, then return to S2.

[0154] Furthermore, the logical instructions in the aforementioned memory 330 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0155] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, wherein when the program instructions are executed by a computer, the computer is capable of executing the clustering rule recommendation method provided by the above methods, the method comprising:

[0156] S1. Preprocess the relational data in the recommended topics to obtain processed relational data; the relational data includes customer identifiers, relational targets, and relational degrees.

[0157] S2. After merging the user's feature wide table data and the processed relational data, group them by relational target and analyze the group characteristics;

[0158] S3. Add the user-selected rule item to the grouping feature to form a clustering rule;

[0159] S4. Based on the grouping rules, generate a selected pool, match the grouping rules of the selected pool with the original rules in the preset recommendation pool, and sort the matched recommendation rules according to their recommendation degree.

[0160] S5. Present the recommendation rules that meet the preset conditions for ranking the recommendation degree;

[0161] S6. Save the added rule after the user references the recommendation rule and add it to the feature wide table data, then return to S2.

[0162] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the aforementioned clustering rule recommendation methods, the method comprising:

[0163] S1. Preprocess the relational data in the recommended topics to obtain processed relational data; the relational data includes customer identifiers, relational targets, and relational degrees.

[0164] S2. After merging the user's feature wide table data and the processed relational data, group them by relational target and analyze the group characteristics;

[0165] S3. Add the user-selected rule item to the grouping feature to form a clustering rule;

[0166] S4. Based on the grouping rules, generate a selected pool, match the grouping rules of the selected pool with the original rules in the preset recommendation pool, and sort the matched recommendation rules according to their recommendation degree.

[0167] S5. Present the recommendation rules that meet the preset conditions for ranking the recommendation degree;

[0168] S6. Save the added rule after the user references the recommendation rule and add it to the feature wide table data, then return to S2.

[0169] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0170] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0171] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for recommending a grouping rule, characterized by, include: S1. Preprocess the relational data in the recommended topics to obtain the processed relational data; The relationship data includes customer identifiers, relationship objectives, and relationship degree; S2. After merging the user's feature wide table data and the processed relational data, group them by relational target and analyze the group characteristics; S3. Add the user-selected rule item to the grouping feature to form a clustering rule; S4. Based on the grouping rules, generate a selected pool, match the grouping rules of the selected pool with the original rules in the preset recommendation pool, and sort the matched recommendation rules according to their recommendation degree. S5. Present the recommendation rules that meet the preset conditions for ranking the recommendation degree; S6. Save the added rule after the user references the recommendation rule and add it to the feature wide table data, then return to S2; The method further includes: Based on the recommendation rules cited by the user, record the user's attention to the target of the reference relationship; The recommendation score is obtained based on the attention level and the relationship level. The analysis grouping features include: Cluster and bin the grouped relational data, and record the features, bin intervals, user scale, and sum of interval relation degrees; Alternatively, record features, enumerated values, customer size, and the sum of enumerated relation degrees; Alternatively, it can record characteristics, user scale, and the sum of relationship levels.

2. The method of claim 1, wherein, Before preprocessing the relational data in the recommended topics to obtain the processed relational data, the process includes: Create a recommendation topic, which includes: topic name, loss coefficient, and recommendation weight.

3. The method of claim 2, wherein, When the relational data in the recommended topics is preprocessed to obtain the processed relational data, it includes: Pre-grouping customer identifiers based on the data in the relation target column: If a customer identifier appears multiple times in relation to the relation target during pre-grouping, the relation degree will decrease infinitely. Let the loss coefficient be q, and the formula is: ； The relational degree in the relational data is normalized to a value from 0 to 100, using the following formula: ； Where X is the degree of relation between the user and the relation target, Xn represents the degree of relation between the customer identifier and the relation target appearing n times; q represents the loss coefficient, an represents the degree of relation of the current relation data; Xscale represents the normalized degree of relation; Xmin represents the minimum value among all degrees of relation, and Xmax represents the maximum value among all degrees of relation.

4. The method of claim 1, wherein, The step of matching the grouping rules of the selected pool with the original rules in the preset recommendation pool includes: The selected pool's clustering rules are matched with the original rule items, original conditions, and original operators of the original rules in the recommendation pool. The rules that successfully match are used as recommendation rules. The clustering rules include clustering features and conditions, clustering operators, and clustering parameters.

5. The clustering rule recommendation method according to any one of claims 1-4, characterized in that, After presenting the recommendation rules that rank the recommendations according to preset conditions, the process includes: Based on the recommendation rules cited by the user, the recommended content corresponding to the recommendation rules is displayed to the user.

6. A grouping rule recommendation device, characterized in that, include: The preprocessing module is used to preprocess the relational data in the recommended topics to obtain the processed relational data; The relationship data includes customer identifiers, relationship objectives, and relationship degree; The grouping module is used to merge the user's feature wide table data and the processed relational data, group them by relational target, and analyze the group characteristics. The clustering rule determination module is used to add the rule items selected by the user to the grouping features to form clustering rules; The matching module is used to generate a selected pool based on the grouping rules, match the grouping rules of the selected pool with the original rules in the preset recommendation pool, and sort the matched recommendation rules according to the degree of recommendation. The recommendation presentation module is used to present the recommendation rules that meet the preset conditions for recommendation ranking; The update module is used to save the added rules after the user references the recommendation rules and add them to the feature wide table data, and then return to the grouping module; The recording module is used to record the user's attention to the referenced relationship target based on the recommendation rules cited by the user; The determination module is used to obtain the recommendation score based on the attention level and the relationship level; The grouping module is specifically used to cluster and bin the grouped relational data, and record features, bin intervals, user scale, and the sum of interval relation degrees; or record features, enumerated values, customer scale, and the sum of enumerated relation degrees; or record features, user scale, and the sum of relation degrees.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the clustering rule recommendation method as described in any one of claims 1 to 5.

8. A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the grouping rule recommendation method as described in any one of claims 1 to 5.

Images