An intelligent big data analysis system

Abstract

The embodiment of the present specification provides an intelligent big data analysis system, which belongs to the field of data processing, and includes a graph establishment module, which is used to establish a commodity relationship graph of multiple commodities sold on an e-commerce platform; an information acquisition module, which is used to obtain purchase records of multiple customers on the e-commerce platform; a customer clustering module, which is used to cluster multiple customers and generate multiple customer clusters; a cluster screening module, which is used to determine at least one target customer cluster from multiple customer clusters based on relevant information of promotional commodities; an information screening module, which is used to obtain purchase records and search records related to promotional commodities of each customer included in the target customer cluster; a secondary clustering module, which is used to perform secondary clustering on at least one customer included in the customer cluster to determine multiple customer groups; a customer screening module, which is used to determine whether a customer group is a target customer group; and a message push module, which is used to send push messages to the target customer group, which has the advantage of improving the order conversion rate of promotional products on the e-commerce platform.

Description

Technical Field

[0001] This specification relates to the field of data processing, and in particular to an intelligent big data analysis system. Background Art

[0002] In the field of e-commerce, most consumers like and are accustomed to purchasing goods on e-commerce platforms. In the current process of purchasing goods, customers need to open the e-commerce platform and carefully select and purchase goods from a large number of goods. They usually need to compare prices and quality, so they spend a lot of time shopping on the e-commerce platform. E-commerce platforms sometimes use discounts and promotions to increase sales. Existing e-commerce platforms lack a suitable method for actively recommending promotional goods, and customers often cannot obtain discount promotion information for the goods they want in the first place. The existing technology lacks a suitable method for actively recommending promotional goods, resulting in a single recommendation of promotional goods, a low optimization rate, and no increase in the order rate of promotional goods.

[0003] Therefore, it is necessary to provide an intelligent big data analysis system that has the advantage of recommending promotional products to different customers in a targeted manner and improving the order conversion rate of promotional products on e-commerce platforms. Summary of the invention

[0004] One of the embodiments of the present specification provides an intelligent big data analysis system, including: a graph establishment module, used to establish a commodity relationship graph of multiple commodities sold on an e-commerce platform; an information acquisition module, used to obtain purchase records of multiple customers of the e-commerce platform; a customer clustering module, used to cluster the multiple customers based on the commodity relationship graph of the multiple commodities sold on the e-commerce platform and the purchase records of the multiple customers of the e-commerce platform to generate multiple customer clusters, wherein each of the customer clusters includes at least one of the customers; a cluster screening module, used to determine at least one target customer cluster from the multiple customer clusters based on relevant information of promotional commodities; an information screening module, used to obtain, for each of the target customer clusters, the information of each customer included in the target customer cluster; purchase records and search records related to the promotional goods; a secondary clustering module, for, for each target customer cluster, based on the purchase records and search records related to the promotional goods of each customer included in the target customer cluster, to perform secondary clustering on at least one of the customers included in the customer cluster to determine a plurality of customer groups, wherein each of the customer groups includes at least one of the customers; a customer screening module, for, for each of the customer groups, based on the purchase records and search records related to the promotional goods of the cluster center of the customer group, to determine whether the customer group is a target customer group; a message push module, for generating a push message related to the promotional goods, and sending the push message to a user terminal used by the customers included in the target customer group.

[0005] In some embodiments, establishing a product relationship map of multiple products sold on an e-commerce platform includes: obtaining relevant information about the products sold on the e-commerce platform, wherein the relevant information about the products sold on the e-commerce platform includes at least the name, structure, function and application scenario of the products sold on the e-commerce platform; and establishing the product relationship map based on the relevant information about the products sold on the e-commerce platform.

[0006] In some embodiments, the multiple customers are clustered based on the commodity relationship graph of the multiple commodities sold on the e-commerce platform and the purchase records of the multiple customers of the e-commerce platform to generate multiple customer clusters, including: determining multiple major commodity categories based on the commodity relationship graph of the multiple commodities sold on the e-commerce platform; for each customer of the e-commerce platform, determining the major commodity category to which the purchased commodity of the customer belongs based on the purchase record of the customer, and generating a major commodity category label; clustering the multiple customers based on the major commodity category labels of the multiple customers to generate multiple customer clusters.

[0007] In some embodiments, the determining of at least one target customer cluster from the multiple customer clusters based on the relevant information of the promotional products includes: determining the major product category of the promotional products based on the relevant information of the promotional products and the product relationship map; for each of the customer clusters, judging whether the customer cluster is the target customer cluster based on the major product category of the promotional products and the major product category label corresponding to the customer cluster.

[0008] In some embodiments, based on the purchase records and search records related to the promotional goods of each customer included in the target customer cluster, secondary clustering is performed on at least one of the customers included in the customer cluster to determine multiple customer groups, including: for each customer included in the target customer cluster, using a trend prediction model to determine the customer's consumption trend value for the major category of the promotional goods based on the customer's purchase records and search records related to the promotional goods; based on the consumption trend value of each customer included in the target customer cluster for the major category of the promotional goods, secondary clustering is performed on the customers included in the customer cluster to determine the multiple customer groups.

[0009] In some embodiments, obtaining the purchase record and search record related to the promotional product of each customer included in the target customer cluster includes: for each customer included in the target customer cluster, obtaining the customer's purchase record and search record for the promotional product in the first target historical time period; based on the product relationship map, determining the related products of the promotional product; for each customer included in the target customer cluster, obtaining the customer's purchase record and search record for the related products of the promotional product in the first target historical time period.

[0010] In some embodiments, obtaining the customer's search records for the promotional product during the first target historical time period includes: obtaining all search records of the customer during the first target historical time period; for each of the search records, determining the customer's stay time on the search results page, and judging whether the search record is a valid search record based on the customer's stay time on the search results page; for each of the valid search records, determining the product corresponding to the valid search record based on the search results of the valid search record, and judging whether the valid search record is the customer's search record for the promotional product during the first target historical time period based on the product corresponding to the valid search record.

[0011] In some embodiments, the determining whether the customer group is the target customer group based on the purchase records and search records related to the promotional item by the cluster center of the customer group includes: predicting the consumption probability of the cluster center of the customer group for the promotional item based on the service life and shelf life of the promotional item and the purchase records and search records related to the promotional item by the cluster center of the customer group in the first target historical time period through a consumption prediction model; and determining whether the customer group is the target customer group based on the consumption probability of the cluster center of the customer group for the promotional item predicted by the consumption prediction model.

[0012] In some embodiments, it also includes: for each customer included in each of the target customer groups, obtaining the customer's purchase record in the second target time period, and determining the customer's delivery address based on the customer's purchase record in the second target time period; based on the delivery address of each customer included in all of the target customer groups, performing a third clustering on the customers included in all of the target customer groups to determine at least one target customer cluster; based on the delivery address of the cluster center of the target customer cluster, determining the distribution quantity of the promotional goods in each distribution warehouse.

[0013] The intelligent big data analysis system provided in this specification has at least the following beneficial effects compared with the prior art:

[0014] (1) Based on the relevant information of the promotional products, at least one target customer cluster is determined from multiple customer clusters, and potential customers of the promotional products among multiple customers of the e-commerce platform are screened for the first time. On this basis, based on the purchase records and search records related to the promotional products of the cluster center of the customer group, it is determined whether the customer group is the target customer group, and the potential customers of the promotional products among multiple customers of the e-commerce platform are screened for the second time, so as to more accurately determine the customers who need to be recommended, generate push messages related to the promotional products, and send the push messages to the user terminals used by the customers included in the target customer group, thereby improving the order conversion rate of the promotional products of the e-commerce platform, and the overall data processing volume is smaller and the efficiency is higher;

[0015] (2) By determining the distribution quantity of promotional goods in each distribution warehouse based on the distribution address of the cluster center of the target customer cluster, the corresponding amount of promotional goods can be allocated to each distribution warehouse in advance. During the sales period of the promotional goods, it is ensured that the distribution warehouses in the areas where most of the customers who place orders are located have spot goods, thereby improving distribution efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] This specification will be further described in the form of exemplary embodiments, which will be described in detail by the accompanying drawings. These embodiments are not restrictive, and in these embodiments, the same number represents the same structure, wherein:

[0017] Figure 1 It is an application scenario diagram of the intelligent big data analysis system shown in some embodiments of this specification;

[0018] Figure 2 is a module diagram of an intelligent big data analysis system according to some embodiments of this specification;

[0019] Figure 3 is an exemplary flow chart of a method for recommending promotional products on an e-commerce platform using an intelligent big data analysis system according to some embodiments of this specification;

[0020] Figure 4 It is a schematic diagram of a commodity relationship map shown in some embodiments of this specification.

[0021] In the figure, 110, processing device; 120, network; 130, user terminal; 140, storage device; 410, commodity node; 420, commodity category. DETAILED DESCRIPTION

[0022] In order to more clearly illustrate the technical solutions of the embodiments of this specification, the following is a brief introduction to the drawings required for the description of the embodiments. Obviously, the drawings described below are only some examples or embodiments of this specification. For ordinary technicians in this field, this specification can also be applied to other similar scenarios based on these drawings without creative work. Unless it is obvious from the language environment or otherwise explained, the same reference numerals in the figures represent the same structure or operation.

[0023] It should be understood that the "system", "device", "unit" and / or "module" used herein are a method for distinguishing different components, elements, parts, portions or assemblies at different levels. However, if other words can achieve the same purpose, the words can be replaced by other expressions.

[0024] As shown in this specification and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "comprise" and "include" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0025] Flowcharts are used in this specification to illustrate the operations performed by the system according to the embodiments of this specification. It should be understood that the preceding or following operations are not necessarily performed precisely in order. Instead, the steps may be processed in reverse order or simultaneously. At the same time, other operations may also be added to these processes, or one or more operations may be removed from these processes.

[0026] Figure 1 This is an application scenario diagram of the intelligent big data analysis system shown in some embodiments of this specification. Figure 1 As shown, the intelligent big data analysis system may include a processing device 110, a network 120, a user terminal 130, and a storage device 140. The intelligent big data analysis system may recommend promotional products on an e-commerce platform by implementing the methods and / or processes disclosed in this specification.

[0027] The processing device 110 can be used to process data and / or information from at least one component of the intelligent big data analysis system or an external data source (e.g., a cloud data center). The processing device 110 can access data and / or information from the user terminal 130 and the storage device 140 through the network 120. The processing device 110 can directly connect to the storage device 140 to access information and / or data. For example, the processing device 110 can obtain purchase records of multiple customers of an e-commerce platform from the storage device 140. The processing device 110 can process the obtained data and / or information. For example, the processing device 110 may establish a commodity relationship map of multiple commodities sold on the e-commerce platform, obtain purchase records of multiple customers on the e-commerce platform, cluster multiple customers based on the commodity relationship map of multiple commodities sold on the e-commerce platform and the purchase records of multiple customers on the e-commerce platform, generate multiple customer clusters, wherein each customer cluster includes at least one customer, determine at least one target customer cluster from the multiple customer clusters based on the relevant information of the promotional commodity, obtain the purchase record and search record related to the promotional commodity of each customer included in the target customer cluster for each target customer cluster, perform secondary clustering on at least one customer included in the customer cluster based on the purchase record and search record related to the promotional commodity of each customer included in the target customer cluster for each target customer cluster, determine multiple customer groups, wherein each customer group includes at least one customer, determine whether the customer group is the target customer group based on the purchase record and search record related to the promotional commodity of the cluster center of the customer group, generate a push message related to the promotional commodity, and send the push message to the user terminal 130 used by the customer included in the target customer group. In some embodiments, the processing device 110 may be a single server or a server group. The processing device 110 may be local or remote.

[0028] The network 120 may include any suitable network that provides information and / or data exchange that can facilitate the intelligent big data analysis system. In some embodiments, one or more components of the intelligent big data analysis system (e.g., processing device 110, user terminal 130, storage device 140) may exchange information and / or data through the network 120. In some embodiments, the network 120 may be any one or more of a wired network or a wireless network. In some embodiments, the network 120 may include one or more network access points. For example, the network 120 may include a wired or wireless network access point, such as a base station and / or a network exchange point, through which one or more components of the intelligent big data analysis system may be connected to the network 120 to exchange data and / or information.

[0029] The user terminal 130 refers to one or more terminals or software used by customers of the e-commerce platform. In some embodiments, the user terminal 130 may include but is not limited to a smart phone, a tablet computer, a laptop computer, a desktop computer, etc. In some embodiments, the user terminal 130 may interact with other components in the intelligent big data analysis system through the network 120. For example, the processing device 110 may send a generated push message related to the promotional product to the user terminal 130 used by the customers included in the target customer group through the network.

[0030] The storage device 140 can be used to store data, instructions and / or any other information. In some embodiments, the storage device 140 can store data from the processing device 110, the user terminal 130, and the storage device 140. For example, the storage device 140 can store the purchase records of multiple customers of the e-commerce platform. For another example, the storage device 140 can store a trained machine learning model. In some embodiments, the storage device 140 may include a large-capacity memory, a removable memory, a volatile read-write memory, a read-only memory (ROM), etc. or any combination thereof. Exemplary large-capacity memory may include a disk, an optical disk, a solid-state disk, etc. Exemplary removable memory may include a flash drive, a floppy disk, an optical disk, a memory card, a compressed disk, a tape, etc. Exemplary volatile read-write memory may include a random access memory (RAM). Exemplary RAM may include dynamic random access memory (DRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), static random access memory (SRAM), thyristor random access memory (T-RAM) and zero capacitance random access memory (Z-RAM), etc. Exemplary ROMs may include mask read-only memory (MROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), compact disk read-only memory (CD-ROM), and digital versatile disk read-only memory, etc. In some embodiments, the storage device 140 may be executed on a cloud platform. By way of example only, the cloud platform may include a private cloud, a public cloud, a hybrid cloud, a community cloud, a distributed cloud, an internal cloud, a multi-layer cloud, etc., or any combination thereof.

[0031] It should be noted that the intelligent big data analysis system is provided for illustrative purposes only and is not intended to limit the scope of this specification. For those of ordinary skill in the art, various modifications or variations can be made based on the description of this specification. For example, the intelligent big data analysis system may also include a database. However, these variations and modifications will not deviate from the scope of this specification.

[0032] Figure 2 is a module diagram of an intelligent big data analysis system according to some embodiments of this specification. Figure 2As shown, the intelligent big data analysis system may include a graph building module, an information acquisition module, a customer clustering module, a cluster screening module, an information screening module, a secondary clustering module, a customer screening module, a message push module and a distribution optimization module.

[0033] The graph building module can be used to build a commodity relationship graph of multiple commodities sold on the e-commerce platform. In some embodiments, the graph building module can also be used to obtain relevant information of commodities sold on the e-commerce platform, wherein the relevant information of commodities sold on the e-commerce platform at least includes the name, structure, function and application scenario of the commodities sold on the e-commerce platform; and build a commodity relationship graph based on the relevant information of commodities sold on the e-commerce platform.

[0034] The information acquisition module can be used to obtain the purchase records of multiple customers on the e-commerce platform.

[0035] The customer clustering module can be used to cluster multiple customers based on the commodity relationship graph of multiple commodities sold on the e-commerce platform and the purchase records of multiple customers on the e-commerce platform to generate multiple customer clusters, wherein each customer cluster includes at least one customer. In some embodiments, the customer clustering module can also be used to determine multiple commodity categories 420 based on the commodity relationship graph of multiple commodities sold on the e-commerce platform; for each customer on the e-commerce platform, determine the commodity category 420 to which the customer's purchased commodity belongs based on the customer's purchase record, and generate a commodity category label; based on the commodity category labels of multiple customers, cluster multiple customers to generate multiple customer clusters.

[0036] The cluster screening module can be used to determine at least one target customer cluster from multiple customer clusters based on the relevant information of the promotional products. In some embodiments, the cluster screening module can also be used to determine the product category 420 of the promotional products based on the relevant information of the promotional products and the product relationship map; for each customer cluster, based on the product category 420 of the promotional products and the product category label corresponding to the customer cluster, determine whether the customer cluster is a target customer cluster.

[0037] The information screening module can be used to obtain, for each target customer cluster, purchase records and search records related to promotional products for each customer included in the target customer cluster. In some embodiments, the information screening module can also be used to obtain, for each customer included in the target customer cluster, the customer's purchase records and search records for promotional products in the first target historical time period; determine the associated products of the promotional products based on the product relationship map; for each customer included in the target customer cluster, obtain the customer's purchase records and search records for the associated products of the promotional products in the first target historical time period. In some embodiments, the information screening module can also be used to obtain all search records of the customer in the first target historical time period; for each search record, determine the customer's stay time on the search result page, and determine whether the search record is a valid search record based on the customer's stay time on the search result page; for each valid search record, determine the product corresponding to the valid search record based on the search results of the valid search record, and determine whether the valid search record is the customer's search record for the promotional product in the first target historical time period based on the product corresponding to the valid search record.

[0038] The secondary clustering module can be used to perform secondary clustering on at least one customer included in each target customer cluster based on the purchase records and search records related to promotional commodities of each customer included in the target customer cluster, and determine multiple customer groups, wherein each customer group includes at least one customer. In some embodiments, the secondary clustering module can also be used to determine the consumption trend value of the customer for the major category 420 of promotional commodities for each customer included in the target customer cluster through a trend prediction model based on the purchase records and search records related to promotional commodities of the customer; perform secondary clustering on the customers included in the customer cluster based on the consumption trend value of the major category 420 of promotional commodities for each customer included in the target customer cluster, and determine multiple customer groups.

[0039] The customer screening module can be used to determine whether each customer group is a target customer group based on the purchase records and search records of the cluster center of the customer group related to the promotional product. In some embodiments, the customer screening module can also be used to predict the probability of the cluster center of the customer group consuming the promotional product based on the service life and shelf life of the promotional product and the purchase records and search records of the cluster center of the customer group related to the promotional product in the first target historical time period through a consumption prediction model; and determine whether the customer group is a target customer group based on the consumption probability of the cluster center of the customer group consuming the promotional product predicted by the consumption prediction model.

[0040] The message push module can be used to generate push messages related to promotional products, and send the push messages to user terminals used by customers included in the target customer group.

[0041] The distribution optimization module can be used to: for each customer included in each target customer group, obtain the customer's purchase record in the second target time period, and determine the customer's delivery address based on the customer's purchase record in the second target time period; based on the delivery address of each customer included in all target customer groups, perform a third clustering on the customers included in all target customer groups to determine at least one target customer cluster; based on the delivery address of the cluster center of the target customer cluster, determine the distribution quantity of promotional goods in each distribution warehouse.

[0042] It should be noted that the above description of the intelligent big data analysis system and its modules is only for convenience of description and does not limit this specification to the scope of the embodiments. It is understandable that for those skilled in the art, after understanding the principle of the system, it is possible to arbitrarily combine the modules or form a subsystem to connect with other modules without deviating from this principle. In some embodiments, Figure 2 The graph establishment module, information acquisition module, customer clustering module, cluster screening module, information screening module, secondary clustering module, customer screening module, message push module and distribution optimization module disclosed in the specification can be different modules in a system, or a module can realize the functions of two or more modules mentioned above. For example, each module can share a storage module, or each module can have its own storage module. Such variations are all within the protection scope of this specification.

[0043] Figure 3 This is an exemplary flow chart of a method for recommending promotional products on an e-commerce platform using an intelligent big data analysis system according to some embodiments of this specification. Figure 3 As shown, the method for recommending promotional products on an e-commerce platform using an intelligent big data analysis system may include the following steps.

[0044] Step 310: Create a commodity relationship graph of multiple commodities sold on the e-commerce platform. In some embodiments, step 310 is performed by a graph creation module.

[0045] Figure 4 is a schematic diagram of a commodity relationship map according to some embodiments of this specification, such as Figure 4 As shown, the commodity relationship graph may be composed of commodity nodes 410 of multiple commodities sold on the e-commerce platform.

[0046] In some embodiments, the graph building module building a commodity relationship graph of multiple commodities sold on the e-commerce platform may include the following steps:

[0047] Obtaining relevant information about the goods sold on the e-commerce platform, wherein the relevant information about the goods sold on the e-commerce platform at least includes the name, picture, structure, function, application scenario and supplier information of the goods sold on the e-commerce platform, wherein the supplier information may include the supplier name and the supplier's product range, etc.;

[0048] A product relationship map is established based on the relevant information of the products sold on the e-commerce platform.

[0049] Specifically, for any two products sold on the e-commerce platform, the graph establishment module can calculate the correlation between the two products based on the names, pictures, structures, functions, application scenarios and supplier information of the two products, and then establish a product relationship graph.

[0050] For example, the graph establishment module can extract a name keyword pair from the names of the two commodities based on the name keyword extraction model, wherein the name keyword pair can include two name keyword sequences extracted from the names of the two commodities respectively, and a name keyword sequence can correspond to at least one name keyword extracted from the name of a commodity by the name keyword extraction model, and use word2vec to obtain the name representation vector of each name keyword, calculate the cosine similarity between any name keyword in one name keyword sequence and any name keyword in another name keyword sequence, and take the maximum cosine similarity as the name similarity of the two commodities. The name keyword extraction model can include but is not limited to a neural network (NN), a convolutional neural network (CNN), a deep neural network (DNN), a recurrent neural network (RNN), etc. or any combination thereof.

[0051] As an example only, the similarity of the names of two products can be determined based on the following formula:

[0052] S (name，(q，k)) =max(S (i，j) )

[0053] Among them, S (name，(q，k)) is the similarity between the names of the qth item and the kth item, S (i，j) is the cosine similarity between the i-th name keyword in the name keyword sequence corresponding to the q-th product and the j-th name keyword in the name keyword sequence corresponding to the k-th product, i=1, 2...n, where n is the total number of name keywords included in the name keyword sequence corresponding to the q-th product, j=1, 2...m, m is the total number of name keywords included in the name keyword sequence corresponding to the k-th product, and max() is the maximum value operation.

[0054] Furthermore, the graph building module can calculate the appearance similarity of the two commodities based on the pictures of the two commodities through the commodity model. The commodity model may include but is not limited to a neural network (NN), a convolutional neural network (CNN), a deep neural network (DNN), a recurrent neural network (RNN), etc. or any combination thereof.

[0055] Furthermore, the graph establishment module can extract a pair of functional keywords from the functional descriptions of the two commodities based on the functional keyword extraction model, wherein the pair of functional keywords can include two functional keyword sequences extracted from the functions of the two commodities respectively, and one functional keyword sequence can correspond to at least one functional keyword extracted from the function of one commodity by the functional keyword extraction model, and word2vec is used to obtain a functional representation vector of each functional keyword, and the cosine similarity between any functional keyword in one functional keyword sequence and any functional keyword in another functional keyword sequence is calculated, and the maximum value of the cosine similarity is used as the functional similarity of the two commodities. The functional keyword extraction model can include but is not limited to a neural network (NN), a convolutional neural network (CNN), a deep neural network (DNN), a recurrent neural network (RNN), etc. or any combination thereof.

[0056] Furthermore, the graph establishment module can extract application scenario keyword pairs from the application scenario descriptions of the two commodities based on the application scenario keyword extraction model, wherein the application scenario keyword pairs can include two application scenario keyword sequences extracted from the application scenarios of the two commodities respectively, and one application scenario keyword sequence can correspond to at least one application scenario keyword extracted from the application scenario of one commodity by the application scenario keyword extraction model, and use word2vec to obtain the application scenario representation vector of each application scenario keyword, calculate the cosine similarity between any one application scenario keyword in one application scenario keyword sequence and any one application scenario keyword in another application scenario keyword sequence, and take the maximum cosine similarity as the application scenario similarity of the two commodities. The application scenario keyword extraction model can include but is not limited to a neural network (NN), a convolutional neural network (CNN), a deep neural network (DNN), a recurrent neural network (RNN), etc. or any combination thereof.

[0057] Moreover, for each supplier, the graph building module can build a corresponding supplier portrait based on the supplier information. For any two items sold on the e-commerce platform, the supplier portraits corresponding to the two items can be called, and the cosine similarity between the two supplier portraits can be calculated as the supplier similarity corresponding to the two items.

[0058] As an example only, the graph building module can calculate the relevance of two products based on the following formula according to the similarity of the names, appearances, functions, application scenarios and suppliers of the two products:

[0059] S (q，k) =a1S (name，(q，k)) +a2S (shape，(q，k)) +a3S (function，(q，k)) +a4S (scene，(q，k)) +a5S (supplier，(q，k))

[0060] Among them, S (q，k) is the correlation between the qth item and the kth item, S (shape，(q，k)) is the appearance similarity between the qth item and the kth item, S (function，(q，k)) is the functional similarity between the qth item and the kth item, S (scene，(q，k)) is the similarity of the application scenarios of the qth item and the kth item, S (supplier，(q，k)) is the similarity between the suppliers of the qth item and the kth item, and a1, a2, a3, a4 and a5 are all preset weights.

[0061] It can be understood that when the correlation between the qth item and the kth item is greater than the preset correlation threshold, there is a correlation between the qth item and the kth item, and the two nodes representing the qth item and the kth item in the item relationship graph can be connected by an edge, and the greater the correlation between the qth item and the kth item, the shorter the edge between the two nodes representing the qth item and the kth item.

[0062] Step 320, obtaining purchase records of multiple customers of the e-commerce platform. In some embodiments, step 320 is performed by an information acquisition module.

[0063] In some embodiments, the information acquisition module may acquire purchase records of multiple customers of the e-commerce platform from the user terminal 130 , the storage device 140 , or an external data source.

[0064] In some embodiments, in order to reduce the amount of data acquisition, the information acquisition module can acquire the purchase records of multiple customers of the e-commerce platform in the first target historical time period. For example, the information acquisition module can acquire the purchase records of multiple customers of the e-commerce platform in the last three years.

[0065] Step 330: Cluster multiple customers based on the commodity relationship graph of multiple commodities sold on the e-commerce platform and the purchase records of multiple customers on the e-commerce platform to generate multiple customer clusters. In some embodiments, step 330 is performed by a customer clustering module.

[0066] Each customer cluster includes at least one customer.

[0067] In some embodiments, the customer clustering module clusters multiple customers based on the commodity relationship graph of multiple commodities sold on the e-commerce platform and the purchase records of multiple customers on the e-commerce platform to generate multiple customer clusters, which may include:

[0068] Based on a commodity relationship graph of multiple commodities sold on the e-commerce platform, multiple commodity categories are determined 420;

[0069] For each customer of the e-commerce platform, the commodity category 420 to which the customer's purchased commodity belongs is determined based on the customer's purchase record, and a commodity category label is generated;

[0070] Based on the commodity category labels of multiple customers, multiple customers are clustered to generate multiple customer clusters.

[0071] In some embodiments, Figure 4 As shown, the customer clustering module may group multiple commodities that are directly or indirectly connected in the commodity relationship graph into a commodity category 420 .

[0072] In some embodiments, for each customer of the e-commerce platform, the customer clustering module can determine the commodity category 420 to which the customer's purchased goods belong based on the customer's purchase record, and generate a commodity category label. When the commodity category 420 to which the purchased goods corresponding to the next purchase record of the customer belong is repeated with the already generated commodity category label, the commodity category 420 to which the purchased goods corresponding to the purchase record corresponding to the purchase record may not be generated, and the commodity category 420 to which the purchased goods corresponding to the next purchase record belong may be directly determined. It can be understood that a customer may correspond to one or more commodity category labels.

[0073] In some embodiments, the customer clustering module can calculate the similarity of the commodity category labels of any two customers by using a cosine similarity algorithm. For example, for any two customers, the customer clustering module can generate a label sequence for each customer, wherein the label sequence can include multiple commodity category labels determined according to the purchase records of the customer, and the customer clustering module can calculate the similarity of the label sequences of the two customers by using a cosine similarity algorithm.

[0074] In some embodiments, the customer clustering module may cluster multiple customers based on the commodity category tags of multiple customers according to the K-Means algorithm to generate multiple customer clusters. For example, the customer clustering module may cluster multiple customers based on the similarity of the tag sequences of any two customers according to the K-Means algorithm to generate multiple customer clusters.

[0075] Step 340, based on the relevant information of the promotional product, determine at least one target customer cluster from the multiple customer clusters. In some embodiments, step 340 is performed by a cluster screening module.

[0076] In some embodiments, the cluster screening module determines at least one target customer cluster from multiple customer clusters based on the relevant information of the promotional product, which may include:

[0077] Determine the product category of the promotional product based on the relevant information of the promotional product and the product relationship map 420;

[0078] For each customer cluster, based on the commodity category 420 of the promotional commodity and the commodity category label corresponding to the customer cluster, it is determined whether the customer cluster is a target customer cluster.

[0079] In some embodiments, the cluster screening module may determine the commodity category 420 corresponding to the commodity node 410 according to the commodity node 410 corresponding to the promotional commodity in the commodity relationship graph, thereby determining the commodity category 420 of the promotional commodity.

[0080] In some embodiments, for each customer cluster, if the product category 420 of the promotional product and the product category label corresponding to the customer cluster are consistent, the cluster screening module may determine that the customer cluster is a target customer cluster.

[0081] Step 350: For each target customer cluster, obtain the purchase record and search record related to the promotional product of each customer included in the target customer cluster. In some embodiments, step 350 is performed by an information screening module.

[0082] In some embodiments, the information screening module obtains the purchase records and search records related to the promotional products of each customer included in the target customer cluster, which may include:

[0083] For each customer included in the target customer cluster, obtain the customer's purchase record and search record of the promotional product in the first target historical time period;

[0084] Determine the related products of the promotional products based on the product relationship map;

[0085] For each customer included in the target customer cluster, the purchase record and search record of the customer for the associated products of the promotional product in the first target historical time period are obtained.

[0086] In some embodiments, the information screening module may determine the product node 410 corresponding to the promotional product in the product relationship graph, and directly connect the product node 410 corresponding to the promotional product in the product relationship graph to the product corresponding to the product node 410 as an associated product of the promotional product.

[0087] In some embodiments, the information screening module obtains the customer's search record for the promotional product in the first target historical time period, which may include:

[0088] Obtain all search records of customers in the first target historical time period;

[0089] For each search record, determine the time the customer stays on the search results page, and based on the time the customer stays on the search results page, determine whether the search record is a valid search record;

[0090] For each valid search record, the product corresponding to the valid search record is determined based on the search results of the valid search record, and whether the valid search record is a search record for promotional products by the customer in the first target historical time period is determined based on the product corresponding to the valid search record.

[0091] In some embodiments, for each search record, when the customer stays on the search result page corresponding to the search record for longer than a stay time threshold (e.g., 1 minute), the information screening module may determine that the search record is a valid search record. When the product corresponding to the valid search record is a promotional product or an associated product of a promotional product, the information screening module may determine that the valid search record is a search record of the customer for the promotional product in the first target historical time period.

[0092] Step 360: for each target customer cluster, based on the purchase records and search records related to the promotional products of each customer included in the target customer cluster, perform secondary clustering on at least one customer included in the customer cluster to determine multiple customer groups. In some embodiments, step 360 is performed by a secondary clustering module.

[0093] Each customer group includes at least one customer.

[0094] In some embodiments, the secondary clustering module performs secondary clustering on at least one customer included in the customer cluster based on the purchase records and search records related to the promotional product of each customer included in the target customer cluster to determine multiple customer groups, which may include:

[0095] For each customer included in the target customer cluster, the consumption trend value of the customer for the promotional product category 420 is determined based on the customer's purchase record and search record related to the promotional product through the trend prediction model;

[0096] Based on the consumption trend value of each customer included in the target customer cluster for the major category 420 of promotional merchandise, the customers included in the customer cluster are secondary clustered to determine a plurality of customer groups.

[0097] In some embodiments, the input of the trend prediction model is the purchase record and search record of the customer related to the promotional goods, and the output of the trend prediction model is the consumption trend value of the customer for the promotional goods category 420. The consumption trend value can represent the customer's consumption demand for the goods in the promotional goods category 420. The larger the consumption trend value, the greater the customer's consumption demand for the goods in the promotional goods category 420. The trend prediction model can include but is not limited to a neural network (NN), a convolutional neural network (CNN), a deep neural network (DNN), a recurrent neural network (RNN), etc. or any combination thereof. For example, the trend prediction model can be a model formed by combining a convolutional neural network and a deep neural network.

[0098] In some embodiments, the secondary clustering module may perform secondary clustering on the customers included in the customer cluster based on the consumption trend value of each customer included in the target customer cluster for the major category 420 of promotional goods according to the K-Means algorithm to determine multiple customer groups.

[0099] Step 370: for each customer group, based on the purchase records and search records related to the promotional products of the cluster center of the customer group, determine whether the customer group is a target customer group. In some embodiments, step 370 is performed by a customer screening module.

[0100] In some embodiments, the customer screening module determines whether the customer group is a target customer group based on the purchase records and search records related to the promotional products of the cluster center of the customer group, which may include:

[0101] Predicting the consumption probability of the cluster center of the customer group for the promotional product based on the service life and shelf life of the promotional product and the purchase record and search record related to the promotional product of the cluster center of the customer group in the first target historical time period through the consumption prediction model;

[0102] Based on the consumption probability of the cluster center of the customer group predicted by the consumption prediction model for the promotional products, it is determined whether the customer group is the target customer group.

[0103] Understandably, the lifespan and shelf life of a product can influence a customer’s desire to purchase the product again.

[0104] In some embodiments, the input of the consumption prediction model is the service life and shelf life of the promotional product and the purchase record and search record related to the promotional product of the cluster center of the customer group in the first target historical time period, and the output of the trend prediction model is the consumption probability of the cluster center of the customer group for the promotional product. Among them, the consumption probability of the customer for the promotional product can represent the possibility of the customer consuming the promotional product. The greater the consumption probability of the promotional product, the greater the possibility of the customer consuming the promotional product. The consumption prediction model can include but is not limited to a neural network (NN), a convolutional neural network (CNN), a deep neural network (DNN), a recurrent neural network (RNN), etc. or any combination thereof. For example, the consumption prediction model can be a model formed by combining a convolutional neural network and a deep neural network.

[0105] In some embodiments, if the consumption probability of the cluster center of the customer group predicted by the consumption prediction model for the promotional product is greater than the consumption probability threshold, the customer screening module may determine that the customer group is a target customer group.

[0106] Step 380: Generate a push message related to the promotional product, and send the push message to the user terminals used by the customers included in the target customer group. In some embodiments, step 380 is performed by a message push module.

[0107] In some embodiments, the e-commerce platform promotional product recommendation method applied by the intelligent big data analysis system may also include:

[0108] For each customer included in each target customer group, obtaining the customer's purchase record in the second target time period, and determining the customer's delivery address based on the customer's purchase record in the second target time period;

[0109] Based on the delivery address of each customer included in all target customer groups, performing a third clustering on the customers included in all target customer groups to determine at least one target customer cluster;

[0110] Based on the delivery address of the cluster center of the target customer cluster, the distribution quantity of the promotional goods in each distribution warehouse is determined.

[0111] In some embodiments, for any two customers included in all target customer groups, the delivery optimization module can determine the distance between the delivery addresses of the two customers. The delivery optimization module can perform a third clustering on the customers included in all target customer groups based on the distance between the delivery addresses of any two customers included in all target customer groups according to the K-Means algorithm to determine at least one target customer cluster.

[0112] In some embodiments, for any target customer cluster, the distribution optimization module can use a data prediction model to predict the demand for promotional products in the target customer cluster based on the service life and shelf life of the promotional products and the purchase and search records of the promotional products by the customers included in the target customer cluster, and the purchase and search records of the promotional products related to the promotional products. The data prediction model can include but is not limited to a neural network (NN), a convolutional neural network (CNN), a deep neural network (DNN), a recurrent neural network (RNN), etc. or any combination thereof. For example, the data prediction model can be a model formed by combining a convolutional neural network and a deep neural network.

[0113] In some embodiments, the distribution optimization module may determine the distribution quantity of the promotional merchandise in each distribution warehouse according to the predicted demand quantity of the target customer cluster for the promotional merchandise.

[0114] In some embodiments, the distribution quantity of promotional goods in each distribution warehouse is determined based on the delivery address of the cluster center of the target customer cluster, and the corresponding amount of promotional goods can be allocated to each distribution warehouse in advance. During the sales period of the promotional goods, it is ensured that the distribution warehouses in the areas where most of the customers who place orders are located have spot stocks, thereby improving delivery efficiency.

[0115] It should be noted that the above description of the method for recommending promotional products on an e-commerce platform applied by an intelligent big data analysis system is only for example and explanation, and does not limit the scope of application of this specification. For those skilled in the art, various modifications and changes can be made to the method for recommending promotional products on an e-commerce platform applied by an intelligent big data analysis system under the guidance of this specification. However, these modifications and changes are still within the scope of this specification.

[0116] The basic concepts have been described above. Obviously, for those skilled in the art, the above detailed disclosure is only for example and does not constitute a limitation of this specification. Although not explicitly stated here, those skilled in the art may make various modifications, improvements and corrections to this specification. Such modifications, improvements and corrections are suggested in this specification, so such modifications, improvements and corrections still belong to the spirit and scope of the exemplary embodiments of this specification.

[0117] At the same time, this specification uses specific words to describe the embodiments of this specification. For example, "one embodiment", "an embodiment", and / or "some embodiments" refer to a certain feature, structure or characteristic related to at least one embodiment of this specification. Therefore, it should be emphasized and noted that "one embodiment" or "an embodiment" or "an alternative embodiment" mentioned twice or more in different positions in this specification does not necessarily refer to the same embodiment. In addition, certain features, structures or characteristics in one or more embodiments of this specification can be appropriately combined.

[0118] In addition, unless explicitly stated in the claims, the order of the processing elements and sequences described in this specification, the use of alphanumeric characters, or the use of other names are not intended to limit the order of the processes and methods of this specification. Although the above disclosure discusses some invention embodiments that are currently considered useful through various examples, it should be understood that such details are only for illustrative purposes, and the attached claims are not limited to the disclosed embodiments. On the contrary, the claims are intended to cover all modifications and equivalent combinations that are consistent with the essence and scope of the embodiments of this specification. For example, although the system components described above can be implemented by hardware devices, they can also be implemented only by software solutions, such as installing the described system on an existing server or mobile device.

[0119] Similarly, it should be noted that in order to simplify the description disclosed in this specification and thus help understand one or more embodiments of the invention, in the above description of the embodiments of this specification, multiple features are sometimes combined into one embodiment, figure or description thereof. However, this disclosure method does not mean that the features required by the subject matter of this specification are more than the features mentioned in the claims. In fact, the features of the embodiments are less than all the features of the single embodiment disclosed above.

[0120] In some embodiments, numbers describing the number of components and attributes are used. It should be understood that such numbers used in the description of the embodiments are modified by the modifiers "about", "approximately" or "substantially" in some examples. Unless otherwise specified, "about", "approximately" or "substantially" indicate that the numbers are allowed to vary by ±20%. Accordingly, in some embodiments, the numerical parameters used in the specification and claims are approximate values, which may change according to the required features of individual embodiments. In some embodiments, the numerical parameters should take into account the specified significant digits and adopt the general method of retaining digits. Although the numerical domains and parameters used to confirm the breadth of their range in some embodiments of this specification are approximate values, in specific embodiments, the setting of such numerical values ​​is as accurate as possible within the feasible range.

[0121] Finally, it should be understood that the embodiments described in this specification are only used to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of this specification. Therefore, as an example and not a limitation, alternative configurations of the embodiments of this specification may be considered consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to the embodiments explicitly introduced and described in this specification.

Claims

1. An intelligent big data analysis system, characterized in that: include: A graph building module, used to build a commodity relationship graph of multiple commodities sold on the e-commerce platform; An information acquisition module, used to acquire purchase records of multiple customers on the e-commerce platform; A customer clustering module, configured to cluster the multiple customers based on a commodity relationship graph of the multiple commodities sold on the e-commerce platform and purchase records of the multiple customers on the e-commerce platform, so as to generate multiple customer clusters, wherein each of the customer clusters includes at least one of the customers; A cluster screening module, configured to determine at least one target customer cluster from the plurality of customer clusters based on relevant information of the promotional products; An information screening module, for obtaining, for each target customer cluster, purchase records and search records related to the promotional product of each customer included in the target customer cluster; A secondary clustering module, configured to perform secondary clustering on at least one of the customers included in each target customer cluster based on the purchase records and search records related to the promotional product of each customer included in the target customer cluster, to determine a plurality of customer groups, wherein each of the customer groups includes at least one of the customers; A customer screening module, for determining, for each customer group, whether the customer group is a target customer group based on the purchase records and search records related to the promotional product at the cluster center of the customer group; A message push module, used for generating a push message related to the promotional product, and sending the push message to a user terminal used by customers included in the target customer group; The establishment of a commodity relationship map of multiple commodities sold on the e-commerce platform includes: Obtaining relevant information about the goods sold on the e-commerce platform, wherein the relevant information about the goods sold on the e-commerce platform at least includes the name, structure, function and application scenario of the goods sold on the e-commerce platform; For any two commodities sold on the e-commerce platform, the association between the two commodities is calculated based on the names, pictures, structures, functions, application scenarios and supplier information of the two commodities, and then a commodity relationship map is established. Specifically, name keyword pairs are extracted from the names of the two commodities based on a name keyword extraction model, wherein the name keyword pairs include two name keyword sequences respectively extracted from the names of the two commodities, and one name keyword sequence corresponds to at least one name keyword extracted from the name of a commodity by the name keyword extraction model. Word2vec is used to obtain the name representation vector of each name keyword, and the cosine similarity between any name keyword in one name keyword sequence and any name keyword in another name keyword sequence is calculated, and the maximum value of the cosine similarity is taken as the name similarity of the two commodities; Determine the similarity of the names of two products based on the following formula: in, is the similarity between the names of the qth item and the kth item, is the cosine similarity between the i-th name keyword in the name keyword sequence corresponding to the q-th product and the j-th name keyword in the name keyword sequence corresponding to the k-th product, i=1, 2…n, where n is the total number of name keywords included in the name keyword sequence corresponding to the q-th product, j=1, 2…m, where m is the total number of name keywords included in the name keyword sequence corresponding to the k-th product, To obtain the maximum value operation; Calculate the appearance similarity of two products based on the pictures of the two products through the product model; Extracting a pair of functional keywords from the functional descriptions of the two commodities based on a functional keyword extraction model, wherein the pair of functional keywords includes two functional keyword sequences extracted from the functions of the two commodities respectively, and one functional keyword sequence corresponds to at least one functional keyword extracted from the function of one commodity by the functional keyword extraction model, using word2vec to obtain a functional representation vector of each functional keyword, calculating the cosine similarity between any functional keyword in one functional keyword sequence and any functional keyword in another functional keyword sequence, and taking the maximum value of the cosine similarity as the functional similarity of the two commodities; Extracting application scenario keyword pairs from the application scenario descriptions of the two commodities based on the application scenario keyword extraction model, wherein the application scenario keyword pairs include two application scenario keyword sequences extracted from the application scenarios of the two commodities respectively, one application scenario keyword sequence corresponds to at least one application scenario keyword extracted from the application scenario of one commodity by the application scenario keyword extraction model, using word2vec to obtain an application scenario representation vector of each application scenario keyword, calculating the cosine similarity between any application scenario keyword in one application scenario keyword sequence and any application scenario keyword in another application scenario keyword sequence, and taking the maximum value of the cosine similarity as the application scenario similarity of the two commodities; For each supplier, a corresponding supplier portrait is created based on the supplier information. For any two items sold on the e-commerce platform, the supplier portraits corresponding to the two items are called, and the cosine similarity between the two supplier portraits is calculated as the supplier similarity corresponding to the two items. The correlation between two products is calculated based on the following formula according to their name similarity, appearance similarity, function similarity, application scenario similarity and supplier similarity: in, is the correlation between the qth item and the kth item, is the appearance similarity between the qth item and the kth item, is the functional similarity between the qth item and the kth item, is the similarity of application scenarios between the qth item and the kth item, is the similarity between the suppliers of the qth item and the kth item, , , , and All are preset weights; When the correlation between any two commodities is greater than a preset correlation threshold, there is a correlation between the two commodities, and two nodes representing the two commodities in the commodity relationship graph are connected by an edge, and the greater the correlation between the two commodities, the shorter the edge between the two nodes representing the two commodities; The clustering of the multiple customers based on the commodity relationship graph of the multiple commodities sold on the e-commerce platform and the purchase records of the multiple customers on the e-commerce platform to generate multiple customer clusters includes: Determine multiple commodity categories based on a commodity relationship graph of multiple commodities sold on the e-commerce platform; For each customer of the e-commerce platform, determine the commodity category to which the purchased commodity belongs based on the purchase record of the customer, and generate a commodity category label; Clustering the multiple customers based on the commodity category labels of the multiple customers to generate multiple customer clusters; Specifically, multiple commodities that are directly or indirectly connected in the commodity relationship graph are regarded as a commodity category. For each customer of the e-commerce platform, the commodity category to which the customer's purchased commodity belongs is determined based on the customer's purchase record, and a commodity category label is generated. When the commodity category to which the purchased commodity corresponding to the customer's next purchase record belongs is repeated with the generated commodity category label, the commodity category corresponding to the purchase record is not generated, and the commodity category to which the purchased commodity corresponding to the next purchase record belongs is directly determined; For any two customers, generate a label sequence for each customer, wherein the label sequence includes multiple commodity category labels determined according to the purchase records of the customer, and calculate the similarity of the label sequences of the two customers by using a cosine similarity algorithm; According to the K-Means algorithm, multiple customers are clustered based on the similarity of the label sequences of any two customers to generate multiple customer clusters; The step of determining at least one target customer cluster from the plurality of customer clusters based on the relevant information of the promotional product comprises: Determine the major category of the promotional product based on the relevant information of the promotional product and the product relationship map; For each of the customer clusters, based on the commodity category of the promotional commodity and the commodity category label corresponding to the customer cluster, determining whether the customer cluster is the target customer cluster; The step of performing secondary clustering on at least one of the customers included in the customer cluster based on the purchase records and search records related to the promotional product of each customer included in the target customer cluster to determine a plurality of customer groups comprises: For each customer included in the target customer cluster, determining the consumption trend value of the customer for the major category of the promotional product based on the customer's purchase record and search record related to the promotional product by using a trend prediction model; Based on the consumption trend value of each customer included in the target customer cluster for the major category of the promotional merchandise, the customers included in the customer cluster are secondary clustered to determine the multiple customer groups.

2. The intelligent big data analysis system according to claim 1, characterized in that: The step of obtaining the purchase record and search record related to the promotional product of each customer included in the target customer cluster includes: For each customer included in the target customer cluster, obtaining the purchase record and search record of the promotional product by the customer in the first target historical time period; Determining related products of the promotional product based on the product relationship map; For each customer included in the target customer cluster, the purchase record and search record of the customer for the associated products of the promotional product in the first target historical time period are obtained.

3. The intelligent big data analysis system according to claim 2, characterized in that: Acquiring the customer's search record for the promotional product in the first target historical time period includes: Obtain all search records of the customer in a first target historical time period; For each of the search records, determining the time the customer stays on the search results page, and judging whether the search record is a valid search record based on the time the customer stays on the search results page; For each valid search record, determine the product corresponding to the valid search record based on the search results of the valid search record, and determine whether the valid search record is a search record of the customer for the promotional product in the first target historical time period based on the product corresponding to the valid search record.

4. The intelligent big data analysis system according to claim 2, characterized in that: The determining whether the customer group is a target customer group based on the purchase records and search records related to the promotional product of the cluster center of the customer group includes: Predicting the consumption probability of the cluster center of the customer group for the promotional product based on the service life and shelf life of the promotional product and the purchase record and search record related to the promotional product by the cluster center of the customer group in the first target historical time period through a consumption prediction model; Based on the consumption probability of the cluster center of the customer group predicted by the consumption prediction model for the promotional product, it is determined whether the customer group is a target customer group.

5. The intelligent big data analysis system according to claim 1, characterized in that: Also includes: For each customer included in each target customer group, obtaining the purchase record of the customer in the second target time period, and determining the delivery address of the customer based on the purchase record of the customer in the second target time period; Based on the delivery address of each customer included in all the target customer groups, performing a third clustering on all the customers included in the target customer groups to determine at least one target customer cluster; Based on the delivery address of the cluster center of the target customer cluster, the distribution quantity of the promotional merchandise in each distribution warehouse is determined.

Images