Recommendation method and system for reversely matching sales products by using density clustering algorithm

Abstract

The invention discloses a recommendation method and system for reversely matching sales products by using a density clustering algorithm, and belongs to the technical field of information processing. The method comprises the following steps: acquiring a historical cluster; obtaining customer information; performing recommendation matching according to the historical cluster and the customer information to obtain a recommendation result; and screening out recommended sales products from the recommendation result. According to multi-dimensional static and dynamic information sets such as comprehensive information of customers, existing assets of the customers and real-time retrieval condition information of the customers, target sales products needed by the customers are calculated and analyzed in real time. The method has the advantages that static past historical information of the customer user and real-time actual needs of the current session are considered, comprehensiveness is high, real needs of the customer can be met as much as possible, customer satisfaction is achieved, customer experience can be improved for a long time, customer stickiness is improved, and the long-term goals of cost reduction and efficiency improvement of an operator are achieved.

Description

Technical Field

[0001] The present invention relates to the field of information processing technology, and in particular to a recommendation method and system for reverse matching sales products using a density clustering algorithm. Background Art

[0002] With the vigorous development of e-commerce and mobile Internet, in order to meet the ever-increasing demands of customers for faster, more accurate and more refined search results on sales platforms, the industry has a variety of query recommendation solutions, which enable customers to continuously optimize search results during the product search process, and use a variety of algorithms to combine information from the pre-search, search, and post-search stages to recommend products that customers really need.

[0003] In the industry and business field where telecom operators' CRM is located, this demand is similar to that of e-commerce recommendation solutions, and there is an increasing need for this. The first reason is that operators have accumulated thousands of sales products over the years, and the types and quantities of sales products required by customers are increasing year by year. In the process of handling business, there is an urgent need for a business system to recommend to customers the sales products they really need; second, due to the competition among the three major operators and the group's annual requirement for shorter counter time, and considering the increase in manual business learning and acceptance costs, under the general background of reducing costs and increasing efficiency, there is also an urgent need for a recommendation system to assist sales staff in completing intelligent, fast and accurate recommendations and acceptance of sales products.

[0004] With the development and maturity of big data and artificial intelligence technologies, it is feasible to meet this demand in the operator industry. Currently, the more similar patents include the following two solutions submitted by two organizations, Northwestern University and Sankuai Online:

[0005] 1. A method for obtaining a product recommendation sequence and product recommendation (authorization announcement number CN109064285B, application publication number CN109064285A). The key points are: first, obtaining product categories through clustering; second, calculating the distance between each product to be recommended and all products; third, the query is based on the evaluation score of the product to be recommended;

[0006] First, the historical classification is obtained by analyzing the customer's historical products, which is heavily dependent on historical orders and is more suitable for the "heavy product" scenario of e-commerce; second, it relies on the user's predicted score for each product, and the recommended product ranking is given by calculating the distance between the indicator and the product to be recommended. This process has no evaluation basis when the operator accepts a large number of user behaviors; third, the evaluation dimension of the entire process is relatively monotonous. For example, it only relies on history, does not take into account the actual real-time search needs of the time, and does not take into account the customer's historical comprehensive instance information.

[0007] 2. Product recommendation model training method, product recommendation method and device (authorization announcement number CN112598467A, application number 202011537233.7. Its main features are: 1. Feature vectors are extracted from historical orders and merchant information, and finally a set of feature vectors are defined for each product. 2. The key is to use "attribute frequency" to count preference feature vectors.

[0008] 3. User preference feature vector: Analyze user historical behavior.

[0009] First, the analysis is conducted around the characteristic vector of each product sold, but its scope is limited to the attribute set of the product itself. It does not take into account real-time search needs and does not integrate users' existing product instances, which does not meet the three-dimensional information required by operators. Second, the "attribute frequency" statistical preference feature is used in the statistical process, and more advanced clustering algorithms are not used, resulting in the algorithm's accuracy and practicality being unable to meet the complex situations of operators. Summary of the invention

[0010] The object of the present invention is to provide a method and system for recommending reverse matching of sales items by using a density clustering algorithm.

[0011] In order to solve the above technical problems, the present invention provides a method for recommending reverse matching sales products using a density clustering algorithm, comprising the following steps:

[0012] Get historical clustering;

[0013] Obtain customer information;

[0014] Perform recommendation matching based on historical clustering and customer information to obtain recommendation results;

[0015] Filter recommended sales items from the recommended results.

[0016] Preferably, obtaining historical clustering specifically includes the following steps:

[0017] Get sales product instance data;

[0018] Acquire the sales product ordering process information from the sales product instance data as the original data set;

[0019] The original data set is classified based on the density clustering algorithm to obtain historical clusters.

[0020] Preferably, the customer information includes basic customer information, user business actions and user behavior dynamics.

[0021] Preferably, recommendation matching is performed based on historical clustering and customer information to obtain recommendation results, which specifically includes the following steps:

[0022] Perform the first match between the user's basic information and historical clustering to obtain the first recommendation result;

[0023] Perform a second match between the user's business action and the first recommendation result to obtain a second recommendation result;

[0024] The user behavior dynamics are matched with the second recommendation result for a third time to obtain the third recommendation result.

[0025] Preferably, selecting recommended sales products from the recommendation results specifically includes the following steps:

[0026] The recommended sales items are selected from the third recommendation results and sorted.

[0027] Preferably, the basic user information includes subscription package information, information about the user's existing sales items or product instances, and session process site information;

[0028] The subscription package information includes customer portrait, historical contact information and the user's current ARPU value;

[0029] The information of the user's existing sales items or product instances includes package instances, optional package sales item instances, product instances and attribute information;

[0030] The on-site information of the session process includes search records and the length of time the search object stays.

[0031] Preferably, the user service action includes a meal instance, an optional accessory package instance, a product instance and an attribute instance.

[0032] Preferably, the user behavior dynamics include actual user needs.

[0033] Preferably, the first matching, the second matching and the third matching are all matched using a similarity calculation method;

[0034] The similarity calculation is a matrix vector cosine algorithm.

[0035] The present invention also provides a recommendation system for reverse matching sales products using a density clustering algorithm, comprising:

[0036] A first acquisition module is used to acquire historical clustering;

[0037] The second acquisition module is used to acquire customer information;

[0038] The matching module is used to perform recommendation matching based on historical clustering and customer information to obtain recommendation results;

[0039] The recommendation module is used to filter out recommended sales items from the recommendation results.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] In the face of increasingly complex sales products of telecom operators and the increasing requirements of customers for efficient recommendation and retrieval of sales products, the present invention provides a method for real-time calculation and analysis of target sales products required by customers based on multi-dimensional static and dynamic information sets such as comprehensive information of customers themselves, existing assets of customers, and real-time retrieval condition information of customers. In the process, a density clustering analysis algorithm is used to precipitate the feature clustering of each sales product. During the query process of customer users, all key features are obtained in real time, and then the clustering set is matched. According to the ratio threshold set in advance, the hit cluster can be matched, and then the hit cluster can be reversely inferred to be which sales product can be recommended. The overall effect takes into account both the static past historical information of customer users and the real-time actual needs of the current session. It is highly comprehensive and can meet the real needs of customers as much as possible, achieve customer satisfaction, improve customer experience in the long term, thereby improving customer stickiness, and achieving the long-term goal of operators to reduce costs and increase efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The specific implementation modes of the present invention are further described in detail below with reference to the accompanying drawings.

[0043] Figure 1 It is a flow chart of a method for recommending reverse matching of sales products using a density clustering algorithm according to the present invention;

[0044] Figure 2 It is a schematic diagram for obtaining historical clustering;

[0045] Figure 3 is a schematic diagram of the dynamic matching process of query recommendation;

[0046] Figure 4 is a schematic diagram of the recommendation module of Example 1. DETAILED DESCRIPTION

[0047] Many specific details are set forth in the following description to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the connotation of the present invention, so the present invention is not limited to the specific implementation disclosed below.

[0048] The terms used in one or more embodiments of this specification are only for the purpose of describing specific embodiments, and are not intended to limit one or more embodiments of this specification. The singular forms of "a", "said" and "the" used in one or more embodiments of this specification and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0049] It should be understood that although the terms first, second, etc. may be used to describe various information in one or more embodiments of this specification, this information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of one or more embodiments of this specification, the first may also be referred to as the second, and similarly, the second may also be referred to as the first. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0050] The present invention is further described in detail below in conjunction with the accompanying drawings:

[0051] The present invention provides a method for recommending reverse matching of sales products using a density clustering algorithm, comprising the following steps:

[0052] Get historical clustering;

[0053] Obtain customer information;

[0054] Perform recommendation matching based on historical clustering and customer information to obtain recommendation results;

[0055] Filter recommended sales items from the recommended results.

[0056] Preferably, obtaining historical clustering specifically includes the following steps:

[0057] Get sales product instance data;

[0058] Acquire the sales product ordering process information from the sales product instance data as the original data set;

[0059] The original data set is classified based on the density clustering algorithm to obtain historical clusters.

[0060] Preferably, the customer information includes basic customer information, user business actions and user behavior dynamics.

[0061] Preferably, recommendation matching is performed based on historical clustering and customer information to obtain recommendation results, which specifically includes the following steps:

[0062] Perform the first match between the user's basic information and historical clustering to obtain the first recommendation result;

[0063] Perform a second match between the user's business action and the first recommendation result to obtain a second recommendation result;

[0064] The user behavior dynamics are matched with the second recommendation result for a third time to obtain the third recommendation result.

[0065] Preferably, selecting recommended sales products from the recommendation results specifically includes the following steps:

[0066] The recommended sales items are selected from the third recommendation results and sorted.

[0067] Preferably, the basic user information includes subscription package information, information about the user's existing sales items or product instances, and session process site information;

[0068] The subscription package information includes customer portrait, historical contact information and the user's current ARPU value;

[0069] The information of the user's existing sales items or product instances includes package instances, optional package sales item instances, product instances and attribute information;

[0070] The on-site information of the session process includes search records and the length of time the search object stays.

[0071] Preferably, the user service action includes a meal instance, an optional accessory package instance, a product instance and an attribute instance.

[0072] Preferably, the user behavior dynamics include actual user needs.

[0073] Preferably, the first matching, the second matching and the third matching are all matched using a similarity calculation method;

[0074] The similarity calculation is a matrix vector cosine algorithm.

[0075] The present invention also provides a recommendation system for reverse matching sales products using a density clustering algorithm, comprising:

[0076] A first acquisition module is used to acquire historical clustering;

[0077] The second acquisition module is used to acquire customer information;

[0078] The matching module is used to perform recommendation matching based on historical clustering and customer information to obtain recommendation results;

[0079] The recommendation module is used to filter out recommended sales items from the recommendation results.

[0080] The present invention forms comprehensive information about customers by constructing customer portraits, customer passive or active contact tracks with marketing or business operations, customer ARPU values, etc., which is the first source of data samples.

[0081] The present invention obtains a series of instance data around the user's main product instance, including sales product instances, subsidiary sales product instances and corresponding product instances, to form the second data sample source.

[0082] The present invention forms a third source of data samples by recording trace information of the sales product ordering process, such as search objects and stay duration.

[0083] The present invention regularly performs density cluster analysis on the above samples to form cluster classification clusters around the specifications of each sales product. It has good feature convergence characteristics and can achieve good hit accuracy in the query recommendation process, truly realizing the sales products required by customers.

[0084] In order to better illustrate the technical effect of the present invention, the present invention provides the following specific embodiments to illustrate the above technical process:

[0085] Embodiment 1: A method for recommending reverse matching of sales products using a density clustering algorithm, such as Figure 1 As shown, the main ideas are as follows:

[0086] 1. Get all historical clusters, such as Figure 2 As shown;

[0087] 1.1. All sales product ordering process information is obtained from all recorded sales product instance data to form an original data set, which includes three categories. The first category is the customer profile of users who ordered the package, historical contact information (including marketing contact track, location, etc.), and the user's ARPU value at that time; the second category is the information of the user's own sales products or product instances, which are commonly package instances, optional package sales product instances, product instances, and their key attribute information; the third category is the on-site information retained by the user during the session when ordering the package, such as search records, search object stay time, etc.

[0088] 1.2. Use the density clustering algorithm to perform unsupervised learning classification on the sales product acceptance data set, and perform preliminary clustering classification. According to 1, the ordering data set of each package specification in the province has been obtained. Use the DBSCAN algorithm to carry out machine learning on each package data set one by one to obtain K-class cluster divisions, that is, which similar groups the data set can be classified into. Common algorithms include DBSCAN, MDCA, etc. The data classification is multi-dimensional, including the three types of multi-dimensional space in step 1.1.

[0089] 1.3. Regularly optimize the package data of the whole province and keep the clustering classification of the data set up to date.

[0090] 2. Dynamic matching process of query recommendation, such as Figure 3 As shown;

[0091] 2.1. For example, when a customer visits a counter at a business hall, after the customer is located, the system obtains the customer's basic information, i.e., the first type of information, including the customer's profile, historical contact trajectory, static belonging area, recent ARPU value, etc. Based on this information, it can be matched to the cluster classification, matching the first batch of recommended sales products with a relatively wide range (i.e., the first recommendation result).

[0092] 2.2. When the number user is further entered, further business actions are carried out around the user, such as changing the package or changing the sales product, then all instance information centered on the user is obtained, such as package instances, optional package instances, product instances and their attribute instances, etc. Based on this information, secondary matching is performed in the cluster to obtain a more accurate second batch of recommended sales products based on the first batch range (that is, the second recommendation result).

[0093] 2.3. Dynamic intervention of user behavior. When the user puts forward the actual demand, according to the search sales items informed to the salesperson or self-entered, the system can carry out streaming calculation, continue to match cluster classification, and recommend the third batch of recommended sales products (i.e. the third recommendation result) in real time based on the minimum range of the second batch of sales products.

[0094] 2.4. In the three steps of dynamic matching mentioned above, the matching degree of clustering and customer user's real-time information factors (that is, obtaining three categories of more than 10 factors of the original data set of historical clustering) must be calculated. Here, the similarity calculation method can be used. Common algorithms include matrix vector cosine algorithm. The key point is how much the current customer user information is considered to hit the cluster in a certain clustering factor, so as to trigger the sales product specifications corresponding to the cluster classification. Set the ratio threshold based on experience.

[0095] 3. Provide a list of recommended sales items.

[0096] 3.1. Based on the matching results of step 2, sort all the selected recommended sales products according to the hit ratio thresholds of 2..4 above, from high to low, to obtain a list of recommended sales products in order of priority.

[0097] The recommendation method of reverse matching sales products using density clustering algorithm in this embodiment has the following advantages:

[0098] 1. Product recommendation is different in different industries, and the complexity and problem-solving ideas are slightly different. The industry solved by the present invention is the inquiry and acceptance of sales products in CRM of telecom operators. Sales products involve rich dimensions and factors, and it is necessary to build a multi-dimensional cluster analysis model.

[0099] 2. Select the density clustering algorithm, which is different from the traditional K-means. The common K-means algorithm is not applicable to this scenario in the following two points: ① It is only applicable to convex sample sets. ② It needs to specify the number of clusters, while density clustering only requires the neighborhood radius and density threshold.

[0100] 3. When querying for recommendations, reversely utilize the results of the clustering algorithm to determine to which origin the sales product originally belongs by hitting the classified clusters.

[0101] 4. In the process of judging whether the sales product is hit, the similarity calculation method is used to match the clustering clusters. Common algorithms include matrix vector cosine algorithm. The key point is how much the current customer user information is considered to hit the cluster in a certain clustering factor, so as to trigger the sales product specifications corresponding to the cluster classification. Set the ratio threshold based on experience.

[0102] Example 1: Shanghai Telecom sales portal sales product shopping cart query recommendation module, such as Figure 4 shown.

[0103] In the several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the modules, modules or units is only a logical function division, and there may be other division methods in actual implementation, for example, multiple units, modules or components can be combined or integrated into another device, or some features can be ignored or not executed.

[0104] The units may or may not be physically separated, and the components displayed as units may be one physical unit or multiple physical units, that is, they may be located in one place or distributed in multiple different places. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0105] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0106] In particular, according to the embodiments disclosed in the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present invention include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part, and / or installed from a removable medium. When the computer program is executed by the central processing unit (CPU), the above-mentioned functions defined in the method of the present invention are executed. It should be noted that the above-mentioned computer-readable medium of the present invention can be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium can be, for example, but not limited to, a system, device or device of an electrical, magnetic, optical, electromagnetic, infrared segment, or semiconductor, or any combination of the above.

[0107] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0108] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for recommending reverse matching of sales products using a density clustering algorithm, characterized in that: The following steps are involved: Get historical clustering; Obtain customer information; Perform recommendation matching based on historical clustering and customer information to obtain recommendation results; Filter recommended sales items from the recommended results.

2. The method for recommending reverse matching sales products using a density clustering algorithm according to claim 1, characterized in that: Obtaining historical clustering includes the following steps: Get sales product instance data; Acquire the sales product ordering process information from the sales product instance data as the original data set; The original data set is classified based on the density clustering algorithm to obtain historical clusters.

3. The method for recommending reverse matching sales products using a density clustering algorithm according to claim 2, characterized in that: The customer information includes basic customer information, user business actions and user behavior dynamics.

4. The method for recommending reverse matching sales products using a density clustering algorithm according to claim 3, characterized in that: Recommendation matching is performed based on historical clustering and customer information to obtain recommendation results, which specifically includes the following steps: Perform the first match between the user's basic information and historical clustering to obtain the first recommendation result; Perform a second match between the user's business action and the first recommendation result to obtain a second recommendation result; The user behavior dynamics are matched with the second recommendation result for a third time to obtain the third recommendation result.

5. The method for recommending reverse matching sales products using a density clustering algorithm according to claim 4, characterized in that: Filtering recommended sales products from the recommendation results specifically includes the following steps: The recommended sales items are selected from the third recommendation results and sorted.

6. The method for recommending reverse matching sales products using a density clustering algorithm according to claim 5, characterized in that: The basic user information includes subscription package information, information about the user's existing sales items or product instances, and session process information; The subscription package information includes customer portrait, historical contact information and the user's current ARPU value; The information of the user's existing sales items or product instances includes package instances, optional package sales item instances, product instances and attribute information; The on-site information of the session process includes search records and the length of time the search object stays.

7. The method for recommending reverse matching sales products using a density clustering algorithm according to claim 6, characterized in that: The user business action includes a meal instance, an optional accessory package instance, a product instance, and an attribute instance.

8. The method for recommending reverse matching sales products using a density clustering algorithm according to claim 7, characterized in that: The user behavior dynamics include actual user needs.

9. The method for recommending reverse matching sales products using a density clustering algorithm according to claim 8, characterized in that: The first matching, the second matching and the third matching are all matched using a similarity calculation method; The similarity calculation is a matrix vector cosine algorithm.

10. A recommendation system for reverse matching sales products using a density clustering algorithm, used to implement the recommendation method for reverse matching sales products using a density clustering algorithm as claimed in any one of claims 1 to 9, characterized in that: include: A first acquisition module is used to acquire historical clustering; The second acquisition module is used to acquire customer information; The matching module is used to perform recommendation matching based on historical clustering and customer information to obtain recommendation results; The recommendation module is used to filter out recommended sales items from the recommendation results.

Images