An e-commerce intelligent customer service product recommendation method based on customer behavior

Abstract

The present invention relates to the technical field of recommendation systems, and specifically to an e-commerce intelligent customer service product recommendation method based on customer behavior, comprising: S1, obtaining customer data related to all customer emails received by a client; S2, obtaining request content of the customer email, analyzing the request content, and determining customer needs; S3, based on customer needs, obtaining associated order information of the customer email, the associated order information including product information and payment methods; S4, obtaining all historical feedback information in the customer data, and generating a customer behavior portrait; the customer behavior portrait includes the customer's emotional tendency, interest points, and purchase tendency when providing relevant order feedback; S5, constructing a customer's product association portrait based on the associated order information; S6, comparing the product association portrait with the customer behavior portrait, determining products that match the current customer behavior, and generating a recommended product list; the accuracy and recommendation effect of product recommendations are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of recommendation systems, and in particular to an e-commerce intelligent customer service product recommendation method based on customer behavior. Background Art

[0002] With the rapid development of technology, the amount of data is also increasing day by day. People are increasingly feeling helpless in the face of massive data, so recommendation systems have emerged. Recommendation systems can analyze and mine user behavior, discover users' personalized needs and interest characteristics, and recommend information or products that users may be interested in to users. The application of recommendation systems is conducive to solving the problem of information overload and improving user experience.

[0003] For example, Chinese patent publication number CN116522000A discloses a method for training a recommendation model for recommending content to users, comprising: in response to receiving a user request for a page sent by a client, sending a response to the client, the response carrying a request time associated with the request and content associated with the page generated by a recommendation model; receiving embedded data sent by the client in response to user behavior for content associated with the page, the embedded data carrying the request time; in response to a current moment passing a first time period relative to the request time, performing a first training on the recommendation model based on data samples associated with all currently received embedded data carrying the request time; in response to a current moment passing a second time period greater than the first time period relative to the request time, performing a second training on the recommendation model based on data samples associated with all currently received embedded data carrying the request time.

[0004] However, when recommending content to users, existing technologies have difficulty in accurately identifying customers' specific needs and emotional tendencies, and are unable to identify the relationships between products and customers' purchasing tendencies, resulting in the inability to accurately push products that meet customers' personalized needs and purchasing behavior patterns to users. Summary of the invention

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is: an e-commerce intelligent customer service product recommendation method based on customer behavior, comprising: S1, obtaining customer data related to all customer emails received by the client.

[0006] S2, obtain the request content of the customer's email, analyze the request content, and determine the customer's needs.

[0007] S3, based on customer needs, obtains the associated order information of the customer's email, and the associated order information includes product information and payment method.

[0008] S4, obtains all historical feedback information in the customer data and generates a customer behavior profile; the customer behavior profile includes the customer's emotional tendencies, interests, and purchasing tendencies when providing relevant order feedback.

[0009] S5, build a product association profile of the customer based on the associated order information.

[0010] S6, compare the product association portrait and the customer behavior portrait, determine the products that match the current customer behavior, and generate a recommended product list.

[0011] The beneficial effects of the present invention are as follows: the present invention deeply analyzes the request content of the customer's email, combines the emotional dictionary to determine the customer's emotional tendency, and thus makes the identification of customer needs more accurate; by considering the dynamic changes of the emotional tendency and interest points of the customer during feedback, a behavioral portrait that is more in line with the customer's current state is constructed; by comprehensively considering multiple association rules and similarity indicators, a product association portrait that more accurately reflects the product association relationship and customer purchasing tendency is generated; by comparing the product association portrait and the customer behavior portrait, a recommended product list that is more in line with the customer's personalized needs and purchasing behavior patterns is generated. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0013] Figure 1 It is a flow chart of an e-commerce intelligent customer service product recommendation method based on customer behavior.

[0014] Figure 2 The present invention is a flowchart of step S2 of an e-commerce intelligent customer service product recommendation method based on customer behavior.

[0015] Figure 3 The present invention is a flowchart of step S4 of an e-commerce intelligent customer service product recommendation method based on customer behavior.

[0016] Figure 4 The present invention is a flowchart of step S5 of an e-commerce intelligent customer service product recommendation method based on customer behavior.

[0017] Figure 5 The present invention is a flowchart of step S6 of an e-commerce intelligent customer service product recommendation method based on customer behavior. DETAILED DESCRIPTION

[0018] The embodiments of the present invention are described in detail below. The embodiments described below are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. If no specific techniques or conditions are specified in the embodiments, the techniques or conditions described in the literature in the art or the product specifications are used.

[0019] See also Figure 1 ,An e-commerce intelligent customer service product recommendation method based on customer behavior, includes: S1, obtaining customer data related to all customer emails received by the client.

[0020] S2, obtain the request content of the customer's email, analyze the request content, and determine the customer's needs.

[0021] S3, based on customer needs, obtains the associated order information of the customer's email, and the associated order information includes product information and payment method.

[0022] S4, obtains all historical feedback information in the customer data and generates a customer behavior profile; the customer behavior profile includes the customer's emotional tendencies, interests, and purchasing tendencies when providing relevant order feedback.

[0023] S5. Build a product association portrait of the customer based on the associated order information. The product association portrait includes the product title, quantity, unit price, brand, and size. When building the product association portrait, it is necessary to determine the similarity of the products in the associated order information, as well as the similarity between the categories and labels corresponding to the products. For example, if the customer's order contains two products, each corresponding to a different label, then it is necessary to clarify the association between the products required by the customer to find content similar to the product.

[0024] S6, compare the product association portrait and the customer behavior portrait, determine the products that match the current customer behavior, and generate a recommended product list.

[0025] The above-mentioned customer data obtained from customer emails includes the content of the customer's request during email communication, the order pointed to by the email and the customer's feedback on the current order, that is, the relevant data of the customer's email includes the request content, customer needs, related orders and historical feedback; such as the customer's basic information: company / organization, address, name, etc.; communication history: the subject of the sent and received emails, the content of the email body, the number of replies and the date of communication; the customer's purchase history: purchased goods or services, the goods that the customer prefers to purchase; and the customer's feedback and evaluation on the product or service.

[0026] By analyzing the request content of customer emails, we can accurately identify customer needs and provide a clear direction for subsequent product recommendations; by generating customer behavior portraits, we deeply explore the emotional tendencies, interests, purchasing tendencies and other characteristics of customers when making relevant order feedback, which helps to understand customers more comprehensively; by constructing a product association portrait that includes detailed information such as product title, quantity, unit price, brand, size, etc., and considering the similarity between products and the similarity between corresponding categories and tags, we provide rich product information for product recommendations; by comparing the product association portrait with the customer behavior portrait, we can determine the products that match the current customer behavior and generate a personalized recommended product list; by identifying the real-time behavior in the customer behavior portrait and obtaining the corresponding context information of the current customer, we can make real-time adjustments to the recommended product list to improve the accuracy and timeliness of the recommendation.

[0027] like Figure 2 As shown, the implementation method of determining the customer's needs in step S2 includes the following methods: S21, obtaining the request content of the customer's email, extracting keywords in the request content, identifying the product name and order number corresponding to the keywords, and classifying the questions corresponding to the keywords to determine the customer's specific problems.

[0028] S22, compare the customer's specific questions and corresponding keywords with the preset sentiment dictionary to determine the customer's sentiment tendency.

[0029] S23, takes the customer’s emotional tendencies and the customer’s specific problems as the output of the customer’s needs.

[0030] The method of extracting keywords from the request content is to segment all the words from the request content of the customer's email, filter the segmented words, and extract the keywords of the request content given by the customer; then use the pre-annotated classification model to match the current keywords, find specific questions suitable for the keywords, and determine the emotional tendency of the current recognized keywords based on the specific questions obtained, such as determining the emotional tendency of the current customer based on the modal particles in the customer description and the logic of the customer description, and generate requirements that are consistent with the emotional tendency.

[0031] First, extract the request content from the customer's email, usually by using regular expressions or natural language processing technology to identify and extract the body part; use a keyword extraction algorithm to extract keywords from the request content.

[0032] Identify product names through a preset product name library or regular expression matching; usually order numbers have a fixed format, which can be matched and extracted using regular expressions.

[0033] A preset problem classification system, such as "shipping problem", "refund problem", "product quality problem", etc., extracts keywords and problems and classifies them into corresponding categories. For each word in the request content, check whether it appears in the sentiment dictionary. According to the score of the sentiment word set in the sentiment dictionary, the sum of the product of the sentiment word identified in the request content and the set sentiment word weight is taken as the sentiment tendency of the request content. According to the obtained sentiment tendency value, it can be identified whether the request content belongs to positive or negative sentiment. Finally, the corresponding related order information is found according to the customer's sentiment tendency and the customer's specific problem to determine the product association portrait that the customer needs to build.

[0034] like Figure 3 As shown, the method for generating the customer behavior portrait in step S4 includes the following methods: S41, obtaining the corresponding emotional tendency, interest points and purchase tendency in the historical feedback information.

[0035] S42, calculating the emotional tendency of the customer when giving feedback, and adjusting the obtained interest points according to the emotional tendency when giving feedback.

[0036] S43, takes the sentiment tendency, purchase tendency and adjusted interest points as input vectors to construct an undirected graph of customer behavior.

[0037] S44, determining the customer behavior pattern according to the output result of the customer behavior undirected graph, and combining the customer behavior pattern into a customer behavior profile.

[0038] For emotional tendency, the emotional tendency of customers at the time of feedback is determined by taking the sum of the products of emotional words identified in historical feedback information and preset weights as the value of emotional tendency. Interest points are represented by the probability of occurrence of each word under each label, and the currently required interest points are selected based on the probability of occurrence of each word. Purchase tendency is represented by the weighted value of customer purchase frequency and repeat purchase rate.

[0039] In order to make the generated customer behavior portrait more in line with the current customer, the method of adjusting the acquired interest points in step S42 is as follows: based on the acquired interest points, the score change value of the user's emotional tendency when the interest points appear is obtained; based on the score change value of the emotional tendency, the interest point corresponding to the maximum value of the score change value of the emotional tendency is marked as the target interest point; based on the identified target interest point, the currently identified emotional tendency is reorganized to determine the change value of the reorganized emotional tendency; the target interest point whose change value of the reorganized emotional tendency is greater than the preset change value is used as the adjusted interest point.

[0040] The method for reorganizing the sentiment tendency is to obtain the sentiment tendency of the target interest point corresponding to the adjacent interest points, combine the sentiment tendencies of the adjacent interest points, determine the average value of the score change values ​​of the sentiment tendencies of the adjacent interest points, and use the average value of the score change values ​​of the sentiment tendencies of the adjacent interest points as the change value of the sentiment tendency of the reorganized target interest point.

[0041] The method for obtaining the score change value of the sentiment tendency is as follows: first calculate the sum of the sentiment scores of the words associated with the interest point, and divide it by the number of these words to get the average sentiment score, and take the difference between the average sentiment scores of the previous paragraph and the next paragraph where the interest point appears as the score change value of the sentiment tendency of the current interest point.

[0042] The method for constructing the customer behavior undirected graph in step S43 includes: first converting the acquired emotional tendencies, interest points, and purchase tendencies into emotional vectors, interest vectors, and purchase vectors; and weighted averaging the emotional vectors, interest vectors, and purchase vectors to obtain a comprehensive vector after weighted averaging.

[0043] Each node in the customer behavior undirected graph represents a customer behavior. The edge between each node is determined by the similarity between the comprehensive vectors. The distance between each node is determined according to the Euclidean distance between the comprehensive vectors. The weight between each node is represented by the product of the similarity between the comprehensive vectors and the module of the comprehensive vector. All nodes are connected to generate a customer behavior undirected graph.

[0044] The implementation method of the customer behavior pattern in step S44 includes: analyzing all nodes in the customer behavior undirected graph to identify the customer behavior pattern; and determining the customer behavior portrait based on the identified behavior pattern.

[0045] Methods for identifying customer behavior patterns include calculating the degree centrality, closeness centrality, and betweenness centrality of each node in the customer behavior undirected graph, obtaining the comprehensive importance of the node based on the degree centrality, closeness centrality, and betweenness centrality, identifying the propagation path and influencing factors of customer behavior based on the node corresponding to the maximum comprehensive importance, and using the identified propagation path and influencing factors of customer behavior as the output customer behavior pattern.

[0046] Degree centrality indicates the ratio of the number of edges connected to the current node in the customer behavior undirected graph to the number of all nodes in the undirected graph except the current node; the edges connected to the current node indicate whether there is a connection relationship between the current node and other nodes in the constructed undirected graph. The value of degree centrality is the ratio of the number of nodes that have a relationship with the current node to the total number of nodes.

[0047] Closeness centrality represents the ratio of the number of all nodes except the current node in the customer behavior undirected graph to the sum of the shortest distances between the current node and all connected nodes.

[0048] Betweenness centrality represents the sum of the proportions of the number of shortest paths passing through the current node in the customer behavior undirected graph to the total number of all shortest paths.

[0049] The comprehensive importance of the node is expressed as, ;in, is the comprehensive importance of node v, is the degree centrality of node v, is the maximum value of degree centrality, is the weight of degree centrality, is the closeness centrality of node v, is the maximum value of the proximity centrality, is the weight of proximity centrality, is the betweenness centrality of node v, is the maximum value of betweenness centrality, is the weight of betweenness centrality.

[0050] The implementation method of identifying the propagation path and influencing factors of customer behavior according to the node corresponding to the maximum value of the comprehensive importance includes the following steps: taking the node corresponding to the maximum value of the comprehensive importance as the initial node, connecting the initial node and the nodes adjacent to the initial node according to the similarity between the comprehensive vectors to obtain a moving path, iterating the moving path in a loop until all nodes are connected, taking the moving path with the highest similarity of the comprehensive vectors in the moving path as the propagation path of the customer behavior, and taking the characteristic values ​​of the nodes on the propagation path as the influencing factors for identification.

[0051] Build a product association portrait of the customer based on the associated order information. When building a product association portrait, it is necessary to determine the similarity of the products in the associated order information, such as the product category, brand, price range, color, size, and the similarity between the categories and labels corresponding to the product.

[0052] like Figure 4 As shown, the processing method of the product association portrait in step S5 includes: S51, according to the acquired associated order information, determining the support and confidence between the products in the associated order information, and generating a first association rule according to the support and confidence between the products.

[0053] S52, obtaining the promotion degrees between the commodities and generating a second association rule.

[0054] S53, obtaining the cosine similarity between the products, and determining the third association rule of the products according to the cosine similarity, support, confidence and lift between the products.

[0055] S54, if the third association rule meets the preset conditions, it is determined that there is a strong association between the orders in the associated order information, the corresponding products are combined into an order list, and the indicator value corresponding to the third association rule between the order list and the products is output as a product association portrait.

[0056] The support between products is expressed as the frequency of co-occurrence of two products, that is, the ratio between the orders containing two products and the overall related order information; the confidence between products is expressed as the probability of including another product at the same time when one product is included; the lift between products is expressed as the probability of another product occurring relative to the opposite when one product has been purchased.

[0057] For example, if there are product A and product B, then the support is expressed as , the confidence level is expressed as , the lift is expressed as .

[0058] The cosine similarity of product A and product B is expressed as follows: First, the associated order information needs to be converted into a matrix form, where each row represents an order and each column represents a product. For each order, if a specific product is included, the corresponding position is marked as 1; if it is not included, it is marked as 0; for product A and product B, their vector representations are constructed respectively; the length of these vectors is equal to the total number of transactions, and each element indicates whether the product was purchased in the corresponding transaction; the cosine similarity of product A and product B vectors is obtained by calculating the dot product between them divided by the product of their respective modulus lengths.

[0059] The index value for the third association rule is expressed as, ;in, is the index value of the third association rule, is the support, is the weight of support, is the confidence level, is the confidence weight, It is the degree of improvement. is the weight of the promotion degree, is the cosine similarity, is the weight of cosine similarity, , , , They are the exponential factors of support, confidence, lift, and cosine similarity, which are used to measure the importance of support, confidence, lift, and cosine similarity in the comprehensive calculation.

[0060] When constructing a product association portrait based on the associated order information between products, it is necessary to judge in sequence based on the contents of the first association rule, the second association rule, and the third association rule to determine whether there is an association between the current products, so as to generate a highly associated product list.

[0061] The first association rule indicates that the support and confidence between commodities need to be greater than the set thresholds of support and confidence.

[0062] The second association rule indicates that the degree of promotion between commodities needs to be greater than a set threshold of the degree of promotion.

[0063] The third association rule indicates that after the first association rule and the second association rule are satisfied, the index value of the third association rule is greater than the set index value threshold.

[0064] Preferably, the setting thresholds of support and confidence in the first association rule are set to 0.6, the threshold of lift in the second association rule is set to 0.7, and the threshold of the index value of the third association rule is set to 0.6.

[0065] That is, the preset condition of the third association rule is expressed as follows: when the index value of the third association rule is greater than the threshold value of the index value of the third association rule, the corresponding commodities are assembled into an order list.

[0066] like Figure 5 As shown, the method for generating the product recommendation list in step S6 includes: S61, according to the acquired product association portrait and customer behavior portrait, determining the order of magnitude of each customer behavior in the same time series, and according to the order of magnitude of the customer behavior, determining the mapping relationship between the customer behavior and the product.

[0067] S62, based on the mapping relationship between customer behavior and products, select customer behaviors in product association portraits and customer behavior portraits whose importance exceeds a preset importance threshold, and combine them into a customer behavior chain diagram in chronological order.

[0068] S63, obtaining the index value of each link node in the customer behavior link diagram, converting the link node with the largest index value into target behavior information, and obtaining recommended product list information based on the target behavior information.

[0069] The order of magnitude of customer behavior in each time series is expressed as follows: when comparing the obtained product association portrait and customer behavior portrait for specific customer behaviors and products, the number of products corresponding to each customer behavior at a single time point is set as the order of magnitude of the customer behavior in each time series.

[0070] The mapping relationship between customer behavior and products is expressed as the relationship between each customer behavior and the product that can actually correspond to it. The relationship here is obtained by pre-setting the product type that the customer behavior can correspond to, so as to obtain the products corresponding to different customer behaviors.

[0071] Select customer behaviors in product association portraits and customer behavior portraits whose importance exceeds a preset importance threshold, where the set importance is based on one-to-one correspondence between product association portraits and customer behavior portraits according to the corresponding customer behaviors, and converting the corresponding data in product association portraits and customer behavior portraits into vectors according to the customer behaviors, and calculating the cosine similarity of the vectors as the importance of product association portraits and customer behavior portraits. The preset importance threshold is used to determine the correlation value between the customer's behavior and the product under the corresponding customer behavior. When the preset importance threshold is exceeded, it means that the similarity between the current customer behavior and the product is very high, and the product list provided at this time can better meet the customer's needs.

[0072] The customer behavior chain diagram is composed in chronological order, which is based on the time of occurrence of each customer behavior. The data in all product association portraits and customer behavior portraits are set as small events on the timeline according to the time of occurrence of the customer behavior, namely, chain nodes. The index value of the chain node is obtained by setting a score for the mapping relationship between the product and the customer on the chain node, and the product of the set score and the customer behavior weight is used as the index value of the chain node; the corresponding customer behavior and the mapped product in the chain node with the largest index value in the chain node are used as the target behavior information for conversion. At this time, the target behavior information obtained contains the list of products to be output, that is, the recommended product list information is obtained through the mapping between the target behavior information and the products.

[0073] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations on the present invention. A person skilled in the art may make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention and they are still covered by the protection scope of the present invention.

Claims

1. An e-commerce intelligent customer service product recommendation method based on customer behavior, characterized in that: include: S1, obtain customer data related to all customer emails received by the client; S2, obtaining the request content of the customer's email, analyzing the request content, and determining the customer's needs; S3, based on customer needs, obtains the order information associated with the customer's email, which includes product information and payment method; S4, obtain all historical feedback information in the customer data and generate a customer behavior profile; the customer behavior profile includes the customer's emotional tendency, interest points, and purchase tendency when providing relevant order feedback; S5, build a product association profile of the customer based on the associated order information; S6, compare the product association profile and the customer behavior profile, determine the products that match the current customer behavior, and generate a recommended product list; The method for generating the customer behavior profile in step S4 includes the following methods: S41, obtaining corresponding emotional tendency, interest point and purchase tendency in historical feedback information; S42, calculating the emotional tendency of the customer when providing feedback, and adjusting the obtained interest points according to the emotional tendency when providing feedback; S43, taking the sentiment tendency, purchase tendency and adjusted interest points as input vectors, constructs an undirected graph of customer behavior; S44, determining the customer behavior pattern according to the output result of the customer behavior undirected graph, and combining the customer behavior pattern into a customer behavior profile; The method of adjusting the obtained points of interest in step S42 is as follows: according to the obtained points of interest, the score change value of the user's emotional tendency when the points of interest appear is obtained, according to the score change value of the emotional tendency, the point of interest corresponding to the maximum value of the score change value of the emotional tendency is marked as the target point of interest, based on the identified target point of interest, the currently identified emotional tendency is reorganized to determine the change value of the reorganized emotional tendency; the target point of interest whose change value of the reorganized emotional tendency is greater than the preset change value is used as the adjusted point of interest; The method for reorganizing the sentiment tendency is to obtain the sentiment tendency of the adjacent interest points corresponding to the target interest point, combine the sentiment tendencies of the adjacent interest points, determine the average value of the score change values ​​of the sentiment tendencies of the adjacent interest points, and use the average value of the score change values ​​of the sentiment tendencies of the adjacent interest points as the change value of the sentiment tendency of the reorganized target interest point; The method for constructing the customer behavior undirected graph in step S43 includes: firstly, converting the acquired sentiment tendency, interest point, and purchase tendency into sentiment vector, interest vector, and purchase vector; weighted averaging the sentiment vector, interest vector, and purchase vector to obtain a comprehensive vector after weighted averaging; Each node in the customer behavior undirected graph represents a customer behavior. The edge between each node is determined by the similarity between the comprehensive vectors. The distance between each node is determined according to the Euclidean distance between the comprehensive vectors. The weight between each node is represented by the product of the similarity between the comprehensive vectors and the module of the comprehensive vector. All nodes are connected to generate a customer behavior undirected graph. The implementation method of the customer behavior pattern in step S44 includes: analyzing all nodes in the customer behavior undirected graph to identify the customer behavior pattern; determining the customer behavior profile according to the identified behavior pattern; The method of identifying the customer behavior pattern includes calculating the degree centrality, closeness centrality, and betweenness centrality of each node in the customer behavior undirected graph, obtaining the comprehensive importance of the node according to the degree centrality, closeness centrality, and betweenness centrality, identifying the propagation path and influencing factors of the customer behavior according to the node corresponding to the maximum value of the comprehensive importance, and using the identified propagation path and influencing factors of the customer behavior as the output customer behavior pattern; The implementation method of identifying the propagation path and influencing factors of customer behavior according to the node corresponding to the maximum value of the comprehensive importance includes the following steps: taking the node corresponding to the maximum value of the comprehensive importance as the initial node, connecting the initial node and the nodes adjacent to the initial node according to the similarity between the comprehensive vectors to obtain a moving path, iterating the moving path in a loop until all nodes are connected, taking the moving path with the highest similarity of the comprehensive vectors in the moving path as the propagation path of the customer behavior, and taking the characteristic values ​​of the nodes on the propagation path as the influencing factors for identification.

2. The e-commerce intelligent customer service product recommendation method based on customer behavior according to claim 1, characterized in that: The implementation methods of determining the customer's needs in step S2 include the following methods: S21, obtaining the request content of the customer's email, extracting keywords in the request content, identifying the product name and order number corresponding to the keywords, and classifying the questions corresponding to the keywords to determine the customer's specific questions; S22, comparing the customer's specific questions and corresponding keywords with the preset sentiment dictionary to determine the customer's sentiment tendency; S23, takes the customer’s emotional tendencies and the customer’s specific problems as the output of the customer’s needs.

3. The e-commerce intelligent customer service product recommendation method based on customer behavior according to claim 1, characterized in that: The processing method of the product association portrait in step S5 includes: S51, according to the acquired associated order information, determining the support and confidence between the products in the associated order information, and generating a first association rule according to the support and confidence between the products; S52, obtaining the promotion degree between the products and generating a second association rule; S53, obtaining the cosine similarity between the products, and determining the third association rule of the products according to the cosine similarity, support, confidence and lift between the products; S54, if the third association rule meets the preset conditions, it is determined that there is a strong association between the orders in the associated order information, the corresponding products are combined into an order list, and the indicator value corresponding to the third association rule between the order list and the products is output as a product association portrait.

4. The e-commerce intelligent customer service product recommendation method based on customer behavior according to claim 3, characterized in that: The preset condition of the third association rule is expressed as follows: when the index value of the third association rule is greater than the threshold value of the index value of the third association rule, the corresponding commodities are grouped into an order list.

5. The e-commerce intelligent customer service product recommendation method based on customer behavior according to claim 1, characterized in that: The method for generating the product recommendation list in step S6 includes: S61, determining the magnitude of each customer behavior in the same time series according to the obtained product association portrait and customer behavior portrait, and determining the mapping relationship between the customer behavior and the product according to the magnitude of the customer behavior; S62, according to the mapping relationship between customer behaviors and products, select customer behaviors whose importance exceeds a preset importance threshold in the product association portrait and the customer behavior portrait, and combine them into a customer behavior link diagram in chronological order; S63, obtaining the index value of each link node in the customer behavior link diagram, converting the link node with the largest index value into target behavior information, and obtaining recommended product list information based on the target behavior information.

Images