A Recommendation Method Based on Enterprise Service Bus

Abstract

The invention discloses a recommendation method based on an enterprise service bus, which belongs to the technical field of data recommendation. The method comprises: step one: determining an initial recommendation method, and performing user analysis based on the initial recommendation method to determine each optimized user; step two: identifying a user optimization category corresponding to each optimized user, performing recommendation analysis on each user optimization category, and determining each data supplement item corresponding to each user optimization category; step three: optimizing the initial recommendation method according to each data supplement item and the user optimization category to obtain a target recommendation method; and dynamically monitoring and adjusting the target recommendation method; step four: performing recommendation based on the target recommendation method; by performing recommendation based on an enterprise service bus and utilizing the enterprise service bus to support a variety of communication protocols and data formats, systems and services of different technology stacks can be easily integrated, and at the same time, its modular design makes the system easy to expand and upgrade to meet the ever-growing business needs of enterprises.

Description

Technical Field

[0001] The invention belongs to the technical field of data recommendation, and in particular is a recommendation method based on an enterprise service bus. Background Art

[0002] With the rapid development of information technology and increasingly fierce market competition, enterprises have put forward higher requirements for the efficiency and effectiveness of business development. In many fields such as e-commerce, finance, and retail, personalized recommendation systems have become an important means to improve user experience, increase user stickiness, and promote sales conversion. However, traditional recommendation methods are often limited to a single data source and a fixed recommendation strategy, which is difficult to meet diverse needs.

[0003] As a middleware platform for service integration and interaction between distributed systems, Enterprise Service Bus (ESB) provides new ideas for optimizing and upgrading recommendation systems. ESB provides flexible infrastructure to connect, route and mediate messages between different services, thus achieving service decoupling, reuse and flexible combination. Introducing ESB into the recommendation system can break down data barriers, achieve seamless integration between various systems within the enterprise, improve data circulation efficiency, and provide more comprehensive and accurate data support for recommendation algorithms.

[0004] Based on this, in order to solve the above recommendation problem, the present invention provides a recommendation method based on an enterprise service bus. Summary of the invention

[0005] In order to solve the problems existing in the above solutions, the present invention provides a recommendation method based on an enterprise service bus.

[0006] The purpose of the present invention can be achieved through the following technical solutions:

[0007] A recommended method based on an enterprise service bus includes:

[0008] Step 1: Determine the initial recommendation method, and conduct user analysis based on the initial recommendation method to determine each optimized user;

[0009] Furthermore, the method for determining the recommended method includes:

[0010] Identify each user information, classify each user according to each user information, obtain each user classification; and set the classification weight of each user classification;

[0011] Identify each recommendation method to be selected; obtain historical recommendation data of each recommendation method to be selected, divide the historical recommendation data according to each user category, and obtain classification analysis data corresponding to each user category;

[0012] Performing recommendation evaluation on each of the classification analysis data to obtain a recommendation evaluation result, wherein the recommendation evaluation result includes a qualified recommendation and an unqualified recommendation;

[0013] The classification recommendation value of the recommended qualified user classification is marked as 1;

[0014] Mark the category recommendation value of the user category that is not recommended as 0;

[0015] According to the formula Calculate the corresponding candidate value;

[0016] Where: PB is the value of the candidate method; j represents the corresponding user category, j=1, 2, ..., m, m is a positive integer; λj is the category weight; FTj is the category recommendation value;

[0017] The candidate recommendation methods are sorted in descending order according to the candidate method values ​​to obtain a first sequence, and an initial recommendation method is determined according to the first sequence.

[0018] Furthermore, the method of classifying each user according to each user information includes:

[0019] Evaluate the user's consumption level based on the user information; identify the user's gender, combine the user's gender and consumption level into the user's personal characteristics, and classify the users with the same personal characteristics into one category, which is marked as a unit category;

[0020] Mark each user in the unit class as a unit user, and calculate the recommendation similarity between each unit user in the unit class;

[0021] According to the recommendation similarity, the classification labels corresponding to the corresponding two unit user tags are marked, and the classification labels include different categories and the same category; when the recommendation similarity is greater than the threshold X1, the classification labels are the same category; when the recommendation similarity is not greater than the threshold X1, the classification labels are different categories;

[0022] The unit users in the unit class are reclassified according to the classification labels between the unit users to obtain the classification of each user.

[0023] Furthermore, the method for recommending and evaluating each classification analysis data includes:

[0024] Set recommendation criteria, establish a recommendation judgment model based on the recommendation criteria, analyze the classification analysis data through the recommendation judgment model, and obtain each recommendation judgment value;

[0025] According to the formula Calculate the corresponding recommended evaluation value;

[0026] Where: PA is the recommended assessment value; TP(si) is the recommended judgment value;

[0027] When the recommendation judgment value is greater than the threshold value X2, the evaluation result of the recommendation is evaluated as qualified;

[0028] When the recommendation judgment value is not greater than the threshold value X2, the recommendation evaluation effect is evaluated as unqualified.

[0029] Furthermore, the expression of the recommended judgment model is:

[0030] ;

[0031] In the formula: si is the input data, i=1, 2, ..., n, n is a positive integer; the input data is the classification analysis data; the output data is the recommended judgment value TP(si).

[0032] Furthermore, the method for determining the optimized user includes:

[0033] Identify the recommendation evaluation effect of the initial recommendation method on each user category, mark the user category with unqualified recommendation evaluation effect as a user optimization category, and mark each user in the user optimization category as an optimized user.

[0034] Step 2: Identify the user optimization categories corresponding to each optimized user, perform recommendation analysis on each of the user optimization categories, and determine each data supplement item corresponding to each of the user optimization categories;

[0035] Step 3: Optimize the initial recommendation method according to each data supplement item and user optimization classification to obtain the target recommendation method; and dynamically monitor and adjust the target recommendation method;

[0036] Furthermore, the method for dynamically monitoring and adjusting the target recommendation method includes:

[0037] Acquire the recommendation data of each user and the corresponding recommendation results in real time, integrate them into recommended material data, analyze the recommended material data through a preset recommendation judgment model, and obtain the recommendation judgment value corresponding to each recommended material data;

[0038] Set a unit period, count the recommended judgment values ​​according to the unit period, and obtain the recommended judgment values ​​corresponding to each unit period; mark the recommended judgment values ​​in the unit period as TPz, where z=1, 2, ..., c, and c is a positive integer;

[0039] According to the formula Calculate the unit evaluation value for each unit period;

[0040] Where: PC is the unit evaluation value; e is the natural constant;

[0041] Generate a recommended evaluation curve according to the unit evaluation values ​​corresponding to each unit period, where the horizontal axis of the recommended evaluation curve is the unit time and the vertical axis is the unit evaluation value;

[0042] A baseline is set in the recommended evaluation curve, and the unit evaluation value corresponding to the baseline is marked as the baseline value;

[0043] Identify the abnormal assessment segment [t1, t2] based on the baseline and recommended assessment curves;

[0044] Fit the recommended evaluation curve located in the abnormal evaluation segment to obtain the abnormal evaluation function, and mark the abnormal evaluation function as YU(t);

[0045] According to the formula Calculate the corresponding recommended monitoring values;

[0046] In the formula: MR is the recommended monitoring value; YR is the reference value;

[0047] Determine whether recommended adjustments are needed based on recommended monitoring values;

[0048] When it is determined that the recommended adjustment is not necessary, no corresponding operation is performed;

[0049] When it is determined that recommendation adjustment is necessary, the target recommendation method is adjusted.

[0050] Step 4: Make recommendations based on the target recommendation method.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] By making recommendations based on the enterprise service bus, which supports multiple communication protocols and data formats, it is possible to easily integrate systems and services with different technology stacks. At the same time, its modular design makes the system easy to expand and upgrade to meet the company's evolving business needs. As the hub for data exchange, the enterprise service bus can achieve data sharing and integration between various systems within the enterprise. Through unified data standards and interface specifications, it improves data consistency and accuracy, providing a high-quality data source for the recommendation algorithm. It also enables intelligent adjustment of the recommendation method to meet the recommendation needs of different users. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0054] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0055] 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, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] like Figure 1 As shown, a recommended method based on an enterprise service bus includes:

[0057] Step 1: Determine the initial recommendation method, and conduct user analysis based on the initial recommendation method to determine each optimized user;

[0058] In one embodiment, the initial recommendation method is selected and specified by the platform, and may also be a recommendation method previously applied.

[0059] In one embodiment, the method for determining the recommended method includes:

[0060] Identify the recommendation methods that are suitable for the platform's user recommendations or the platform's needs, and mark them as candidate recommendation methods;

[0061] Identify the corresponding user information on the platform, including gender, shopping history, historical recommendation results and other related information; classify each user according to the user information to obtain each user classification; and set the classification weight of each user classification. The classification weight is calculated and evaluated according to the number of users and consumption level, such as multiplying the number by the consumption amount corresponding to the consumption level, and then calculating its proportion in the whole, marked as classification weight, which can also be set by the platform;

[0062] Obtain historical recommendation data for each candidate recommendation method, identify user information corresponding to the historical recommendation data, match corresponding user categories according to the corresponding user information, classify the historical recommendation data according to each user category, obtain historical recommendation data corresponding to each user category, and mark them as classification analysis data;

[0063] Conduct recommendation evaluation on each classification analysis data to obtain recommendation evaluation results, which include qualified recommendation and unqualified recommendation;

[0064] The classification recommendation value of the recommended qualified user classification is marked as 1;

[0065] Mark the category recommendation value of the user category that is not recommended as 0;

[0066] According to the formula Calculate the corresponding candidate value;

[0067] Where: PB is the value of the candidate method; j represents the corresponding user category, j=1, 2, ..., m, m is a positive integer; λj is the category weight; FTj is the category recommendation value;

[0068] Sort the candidate recommendation methods in descending order according to the candidate method values ​​to obtain a first sequence, and determine the initial recommendation method according to the first sequence, that is, the platform selects the application according to the first sequence.

[0069] In one implementation, the method for classifying each user according to each user information is to classify them manually or by an existing method, and the classification standard is that the habits, recommendations, acceptance, etc. are the same or the differences are within an allowable range.

[0070] In one implementation, the method for classifying each user according to each user information includes:

[0071] Evaluate the user's consumption level based on the shopping records in the user information; for example, count the user's consumption amount on a monthly, quarterly or annual basis, and set the consumption amount range corresponding to different consumption levels according to the platform's needs, such as 0-500, 501-1000, 1001-5000, etc.;

[0072] Gender, consumption level and personal characteristics are combined, and users with the same personal characteristics are classified into one category, which is marked as a unit category;

[0073] Mark each user in the unit class as a unit user, calculate the recommendation similarity between each unit user in the unit class, that is, calculate the recommendation similarity based on the difference in recommendation effects of the same recommended content; mark the corresponding classification labels for the corresponding two unit users according to each recommendation similarity, and the classification labels include different classes and the same class; when the recommendation similarity is greater than the threshold X1, the classification label is the same class; when the recommendation similarity is not greater than the threshold X1, the classification label is different class;

[0074] The unit users in the unit class are reclassified according to the classification labels between the unit users, and the corresponding classifications are marked as user classifications.

[0075] Recommended methods for evaluating the data for each category analysis include:

[0076] Set recommendation criteria, such as how long a user browses, makes a purchase, clicks to open, etc. to be considered as meeting the recommendation criteria; set them according to the actual needs of the platform; establish a recommendation judgment model based on the recommendation criteria, which is used to analyze the classification analysis data to determine whether the classification analysis data meets the recommendation criteria by comparing the corresponding features; the expression of the recommendation judgment model is: ; Where: si is the input data, i=1, 2, ..., n, n is a positive integer; the input data is the classification analysis data; the output data is the recommended judgment value TP (si);

[0077] Analyze the classification analysis data through the recommendation judgment model to obtain each recommendation judgment value;

[0078] According to the formula Calculate the corresponding recommended evaluation value;

[0079] Where: PA is the recommended assessment value; TP(si) is the recommended judgment value;

[0080] When the recommendation judgment value is greater than the threshold value X2, the evaluation result of the recommendation is evaluated as qualified;

[0081] When the recommendation judgment value is not greater than the threshold value X2, the recommendation evaluation effect is evaluated as unqualified.

[0082] Methods for optimizing users include:

[0083] Identify the recommendation evaluation effect of the initial recommendation method on each user category, mark the user category with unqualified recommendation evaluation effect as a user optimization category, and mark each user in the user optimization category as an optimized user.

[0084] Step 2: Identify the user optimization category corresponding to each optimized user, perform recommendation analysis on the user optimization category, and determine which data to supplement for recommendation analysis to improve the recommendation effect so that its recommendation evaluation effect can be qualified. This step is generally set by the platform's professionals to form the user optimization category corresponding to each data supplement item; in other implementations, analysis can also be performed based on existing artificial intelligence technology to determine each data supplement item;

[0085] Step 3: Optimize the initial recommendation method according to each data supplement item and user optimization classification to obtain the target recommendation method; and dynamically monitor and adjust the target recommendation method;

[0086] That is, the corresponding recommendation analysis data is supplemented by each data supplement item corresponding to the user optimization classification, and the recommendation optimization is performed in combination with the collected data of each data supplement item, which is equivalent to adding more restrictions for recommendation analysis.

[0087] The method for dynamically monitoring and adjusting the target recommendation method includes:

[0088] Acquire the recommendation data of each user and the corresponding recommendation results in real time, integrate them into recommended material data, analyze the recommended material data through a preset recommendation judgment model, and obtain the recommendation judgment value corresponding to each recommended material data;

[0089] Set a unit time period, such as one day, three days, one week, etc. as a unit time period, and set it specifically according to the user's online time, etc.; count the recommended judgment values ​​according to the unit time period to obtain the recommended judgment values ​​corresponding to each unit time period; mark the recommended judgment values ​​in the unit time period as TPz, z=1, 2, ..., c, c is a positive integer;

[0090] According to the formula Calculate the unit evaluation value for each unit period;

[0091] Where: PC is the unit evaluation value; e is the natural constant;

[0092] Generate a recommended evaluation curve based on the unit evaluation values ​​corresponding to each unit period, with the horizontal axis being the corresponding unit time and the vertical axis being the unit evaluation value;

[0093] A baseline is set in the recommendation evaluation curve. The baseline is the unit evaluation value of the expected user that is not lower than a certain value. It is set according to the expected recommendation target, such as 80% expected pass rate. The unit evaluation value corresponding to the baseline is marked as the baseline value.

[0094] Identify an abnormal assessment segment [t1, t2] according to the baseline and the recommended assessment curve, where the abnormal assessment segment is a time period of the recommended assessment curve that is below the baseline;

[0095] Fit the recommended evaluation curve located in the abnormal evaluation segment to obtain the abnormal evaluation function, and mark the abnormal evaluation function as YU(t);

[0096] According to the formula Calculate the corresponding recommended monitoring values;

[0097] In the formula: MR is the recommended monitoring value; YR is the reference value;

[0098] Determine whether a recommended adjustment is needed based on the recommended monitoring value, that is, compare the recommended judgment value with the preset value to determine whether a recommended adjustment is needed;

[0099] When it is determined that the recommended adjustment is not necessary, no corresponding operation is performed;

[0100] When it is determined that recommendation adjustment is necessary, the target recommendation method is adjusted; the user can be re-considered as an optimized user and the target recommendation method can be re-formulated, or data supplementary items can be re-added to adjust the recommendation restrictions; at the same time, adjustments can be made based on existing technologies.

[0101] Step 4: Make recommendations to corresponding users based on the target recommendation method.

[0102] The above formulas are all calculated by removing dimensions and taking numerical values. The formula is a formula that is closest to the actual situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and preset thresholds in the formula are set by technical personnel in this field according to actual conditions or obtained by simulating a large amount of data.

[0103] 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 recommendation method based on an enterprise service bus, characterized in that: Methods include: Step 1: Determine the initial recommendation method, and conduct user analysis based on the initial recommendation method to determine each optimized user; Step 2: Identify the user optimization categories corresponding to each optimized user, perform recommendation analysis on each of the user optimization categories, and determine each data supplement item corresponding to each of the user optimization categories; Step 3: Optimize the initial recommendation method according to each data supplement item and user optimization classification to obtain the target recommendation method; and dynamically monitor and adjust the target recommendation method; Step 4: making recommendations based on the target recommendation method; The method for dynamically monitoring and adjusting the target recommendation method includes: Acquire the recommendation data of each user and the corresponding recommendation results in real time, integrate them into recommended material data, analyze the recommended material data through a preset recommendation judgment model, and obtain the recommendation judgment value corresponding to each recommended material data; Set a unit period, count the recommended judgment values ​​according to the unit period, and obtain the recommended judgment values ​​corresponding to each unit period; mark the recommended judgment values ​​in the unit period as TPz, where z=1, 2, ..., c, and c is a positive integer; According to the formula Calculate the unit evaluation value for each unit period; Where: PC is the unit evaluation value; e is the natural constant; Generate a recommended evaluation curve according to the unit evaluation values ​​corresponding to each unit period, where the horizontal axis of the recommended evaluation curve is the unit time and the vertical axis is the unit evaluation value; A baseline is set in the recommended evaluation curve, and the unit evaluation value corresponding to the baseline is marked as the baseline value; Identify the abnormal assessment segment [t1, t2] based on the baseline and recommended assessment curves; Fit the recommended evaluation curve located in the abnormal evaluation segment to obtain the abnormal evaluation function, and mark the abnormal evaluation function as YU(t); According to the formula Calculate the corresponding recommended monitoring values; In the formula: MR is the recommended monitoring value; YR is the reference value; Determine whether recommended adjustments are needed based on recommended monitoring values; When it is determined that the recommended adjustment is not necessary, no corresponding operation is performed; When it is determined that recommendation adjustment is necessary, the target recommendation method is adjusted.

2. The recommendation method based on enterprise service bus according to claim 1, characterized in that: Methods for determining recommended approaches include: Identify each user information, classify each user according to each user information, obtain each user classification; and set the classification weight of each user classification; Identify each recommendation method to be selected; obtain historical recommendation data of each recommendation method to be selected, divide the historical recommendation data according to each user category, and obtain classification analysis data corresponding to each user category; Performing recommendation evaluation on each of the classification analysis data to obtain a recommendation evaluation result, wherein the recommendation evaluation result includes a qualified recommendation and an unqualified recommendation; The classification recommendation value of the recommended qualified user classification is marked as 1; Mark the category recommendation value of the user category that is not recommended as 0; According to the formula Calculate the corresponding candidate value; Where: PB is the value of the candidate method; j represents the corresponding user category, j=1, 2, ..., m, m is a positive integer; λj is the category weight; FTj is the category recommendation value; The candidate recommendation methods are sorted in descending order according to the candidate method values ​​to obtain a first sequence, and an initial recommendation method is determined according to the first sequence.

3. The recommendation method based on enterprise service bus according to claim 2, characterized in that: Methods for classifying each user according to each user information include: Evaluate the user's consumption level based on the user information; identify the user's gender, combine the user's gender and consumption level into the user's personal characteristics, and classify the users with the same personal characteristics into one category, which is marked as a unit category; Mark each user in the unit class as a unit user, and calculate the recommendation similarity between each unit user in the unit class; According to the recommendation similarity, the classification labels corresponding to the corresponding two unit user tags are marked, and the classification labels include different categories and the same category; when the recommendation similarity is greater than the threshold X1, the classification labels are the same category; when the recommendation similarity is not greater than the threshold X1, the classification labels are different categories; The unit users in the unit class are reclassified according to the classification labels between the unit users to obtain the classification of each user.

4. The recommendation method based on enterprise service bus according to claim 2, characterized in that: Recommended methods for evaluating the data for each category analysis include: Set recommendation criteria, establish a recommendation judgment model based on the recommendation criteria, analyze the classification analysis data through the recommendation judgment model, and obtain each recommendation judgment value; According to the formula Calculate the corresponding recommended evaluation value; Where: PA is the recommended evaluation value; TP(si) is the recommended judgment value, i=1, 2, ..., n, n is a positive integer; When the recommendation judgment value is greater than the threshold value X2, the evaluation result of the recommendation is evaluated as qualified; When the recommendation judgment value is not greater than the threshold value X2, the recommendation evaluation effect is evaluated as unqualified.

5. The recommendation method based on enterprise service bus according to claim 4, characterized in that: The expression of the recommended judgment model is: ； In the formula: si is the input data, i=1, 2, ..., n, n is a positive integer; the input data is the classification analysis data; the output data is the recommended judgment value TP(si).

6. The recommendation method based on enterprise service bus according to claim 2, characterized in that: Methods for optimizing users include: Identify the recommendation evaluation effect of the initial recommendation method on each user category, mark the user category with unqualified recommendation evaluation effect as a user optimization category, and mark each user in the user optimization category as an optimized user.

Images