Maternal and infant product intelligent recommendation method and system based on big data analysis and medium

Abstract

The invention discloses a maternal and infant product intelligent recommendation method and system based on big data analysis and a medium. Relates to the technical field of intelligent recommendation. The method is improved on the basis of a traditional product intelligent recommendation technology, family multi-role division is carried out based on preprocessed basic data, decision product results of all family roles about target maternal and infant products are mined, and according to family role affiliation and recommendation modes of a target account number, the target account number is recommended to the target account number. And a recommendation strategy of the target maternal and infant product is generated in combination with a decision product result, so that the defects of an existing product intelligent recommendation system in the aspect of meeting family multi-role requirements of maternal and infant products are overcome, differentiated recommendation contents can be provided for different family role requirements based on multi-role decision under the same target account, and the user experience is improved. And expectations of all parties are met.

Description

Technical Field

[0001] The present invention relates to the field of intelligent recommendation technology, and in particular to a method, system and medium for intelligently recommending maternal and infant products based on big data analysis. Background Art

[0002] With the rapid development of intelligent recommendation technology, personalized recommendation systems in the field of maternal and infant products have gradually become an important tool for improving user experience and purchasing efficiency. Traditional intelligent recommendation systems for maternal and infant products mainly make recommendations based on the personalized needs of individual users. For example, by analyzing the browsing history, purchase history, preference tags and other data of individual users, maternal and infant products that meet their personal needs are recommended. This recommendation method meets the personalized needs of individual users to a certain extent and improves the accuracy of recommendations.

[0003] However, the purchase decision of maternal and infant products usually involves multiple family members, including mothers, fathers, grandparents, etc., and each role has different concerns and needs during the purchase process. For example, mothers may pay more attention to the safety, comfort and practicality of the product; fathers may pay more attention to the cost-effectiveness and functionality of the product; and grandparents may be more concerned about the tradition and durability of the product. Traditional recommendation systems only make recommendations for a single user and cannot effectively balance and integrate the needs of different family members, resulting in the recommendation content being difficult to meet the overall purchase decision needs of the family.

[0004] In addition, there may be conflicts or differences in priorities among family members. For example, a mother may prefer high-end brand products, while a father may prefer economical and practical products. Traditional recommendation systems cannot effectively coordinate these differences, making it difficult for recommendation results to be recognized by family members, thereby reducing the practicality of the recommendation system and user satisfaction.

[0005] Therefore, the existing intelligent recommendation system for maternal and infant products has obvious shortcomings in meeting the needs of multiple roles in the family. There is an urgent need for an intelligent recommendation method that can comprehensively consider the needs of multiple roles in the family to improve the overall effect of the recommendation system and user experience. Summary of the invention

[0006] The technical problem to be solved by the present invention is that the existing product intelligent recommendation system has obvious deficiencies in coping with the family multi-role needs of maternal and infant products; the purpose of the present invention is to provide a method, system and medium for intelligent recommendation of maternal and infant products based on big data analysis, to improve the method of traditional product intelligent recommendation technology, to divide the family into multiple roles based on the pre-processed basic data, and to mine the decision product results of each family role on the target maternal and infant products, to generate a recommendation strategy for the target maternal and infant products according to the family role attribution and recommendation mode of the target account and the decision product results, to solve the deficiencies of the existing product intelligent recommendation system in coping with the family multi-role needs of maternal and infant products, to realize multi-role decision-making under the same target account to provide differentiated recommendation content for the needs of different family roles and meet the expectations of all parties.

[0007] The present invention is achieved through the following technical solutions: This solution provides an intelligent recommendation method for maternal and infant products based on big data analysis, including: Obtaining basic data of target maternal and infant products based on big data, and preprocessing the basic data; Based on the pre-processed basic data, the family is divided into multiple roles, and the decision-making product results of each family role regarding the target maternal and infant products are mined; Obtain the family role attribution and recommendation model of the target account, and combine all decision-making product results to generate a recommendation strategy for the target maternal and infant products.

[0008] A further optimization scheme is that the basic data of the target maternal and infant products is obtained according to the big data, and the basic data is preprocessed; including the method: Obtaining basic data of target maternal and infant products, the basic data including: basic user information, user behavior data, product information and family structure data; The basic data is cleaned and integrated; the integration includes integrating the basic data from different data sources, constructing a unified user data table based on users, and associating user behavior data, family structure data and product information in the user data table.

[0009] A further optimization scheme is to divide the family into multiple roles based on the pre-processed basic data, and mine the decision-making product results of each family role on the target maternal and child products; including methods: Configure family roles including: mom user, dad user, grandparent user and other family members users; Based on the preprocessed basic data, the improved K-means clustering algorithm is used to divide family roles; According to the results of family role division, the original decision set of each family role regarding the target maternal and infant products is constructed, and the decision product set is generated from the original decision set.

[0010] A further optimization scheme is to divide family roles based on the preprocessed basic data combined with an improved K-means clustering algorithm; including the following methods: Each family role is a cluster, and the centroid of each cluster is set separately. The centroid of the cluster of mother users is the purchase frequency, browsing time and age; the centroid of the cluster of father users is the customer unit price and keyword search; the centroid of the cluster of grandparent users is the product purchase record and age; the centroid of the cluster of other family members is the brand awareness and packaging; Perform clustering iteration based on the preprocessed basic data until each cluster converges and then outputs each cluster and centroid; the clustering iteration process includes: Extracting general features and key features of each user from the preprocessed basic data, assigning a first weight to the general features and a second weight to the key features; wherein the second weight is greater than the first weight; The distance from the current user to the cluster centroid is calculated based on general features and key features.

[0011] A further optimization scheme is that the distance function of the clustering iteration is: ; Where D (x,c) represents the distance from the current user to the centroid of the cluster; x i represents the i-th general feature of the current user; x j represents the jth key feature of the current user; β i represents the weight of the i-th general feature; β j represents the weight of the jth key feature; n represents the total number of general features; m represents the total number of key features; c i represents the cluster centroid of the i-th general feature; c j represents the cluster centroid of the jth key feature.

[0012] A further optimization scheme is to construct an original decision set of each family role on the target maternal and infant product according to the family role division result, and generate a decision product set; including the method: Obtaining the family roles and decision results of all users; the decision results include purchase results and effective browsing results; the effective browsing results include that the browsing time of the current product reaches a preset time threshold, and the number of clicks on the current product during the browsing time reaches a preset number threshold; The decision results of all users with the same family role are combined to form the original decision set; After denoising the original decision set, the decision product set is obtained: the frequency of occurrence of the same decision product in the original decision set is counted, and the top N decision products with the largest occurrence frequency constitute the decision product set.

[0013] A further optimization scheme is to obtain the family role attribution and recommendation mode of the target account, and generate a recommendation strategy for the target maternal and infant products in combination with all decision-making product results; including the following methods: Obtaining the family role and recommendation mode of the target account, wherein the recommendation mode includes an associated mode and an unassociated mode, and each recommendation mode includes two options: a single role option or a multi-role option; Combining all decision product results, recommendation strategies for each option under different recommendation modes are generated respectively; in the associated mode, the recommendation strategy uses the decision product results corresponding to the target account's family role as the main reference object, and the decision product results corresponding to a single role or multiple roles as the auxiliary reference object; in the unassociated mode, the recommendation strategy uses the decision product results corresponding to a single role or multiple roles as the reference object.

[0014] A further optimization scheme is that the method for generating the recommendation strategy includes: Get all decision product results; Regarding the single-role option in the non-association mode, the decision product of the decision product result corresponding to the current single-role option is used as the recommended strategy; For multi-role options in the non-association mode, weights are assigned to the decision product results corresponding to each role option, and all decision product results after weight assignment are merged to obtain a combined result set. The top N combined results with larger weights in the combined result set are used as the recommended strategies. For single-role options in association mode, weight W1 is configured for the decision product result corresponding to the current single-role option, and weight W2 is configured for the decision product result corresponding to the target account's family role. All decision product results after weight configuration are merged to obtain a combined result set, and the top N combined results with larger weights in the combined result set are used as the recommended strategy; where W2>W1; Regarding the multi-role options in the association mode, the weight Wi is configured for the decision product results corresponding to the role option i, and the weight W3 is configured for the decision product results corresponding to the target account family role. All decision product results after the weight configuration are merged to obtain a combined result set, and the top N combined results with larger weights in the combined result set are used as the recommended strategies; where W3>Wi.

[0015] This solution provides a maternal and infant product intelligent recommendation system based on big data analysis, which is used to implement the above-mentioned maternal and infant product intelligent recommendation method based on big data analysis; the system includes: A preprocessing module, used to obtain basic data of target maternal and infant products based on big data, and preprocess the basic data; The role decision module is used to divide the family into multiple roles based on the pre-processed basic data, and to mine the decision-making product results of each family role regarding the target maternal and child products; The strategy generation module is used to obtain the family role attribution and recommendation model of the target account, and combine the results of all decision-making products to generate a recommendation strategy for the target maternal and infant products.

[0016] The present solution also provides a computer-readable medium on which a computer program is stored. The computer program is executed by a processor to implement the above-mentioned intelligent recommendation method for maternal and infant products based on big data analysis; specifically, the following steps are performed: Step 1: Obtain basic data of target maternal and infant products based on big data, and pre-process the basic data; Step 2: Divide the family into multiple roles based on the pre-processed basic data, and mine the decision-making product results of each family role regarding the target maternal and infant products; Step 3: Obtain the family role attribution and recommendation model of the target account, and combine all decision-making product results to generate a recommendation strategy for the target maternal and infant products.

[0017] Compared with the prior art, the present invention has the following advantages and beneficial effects: The present invention provides a method, system and medium for intelligent recommendation of maternal and infant products based on big data analysis. The method improves on the traditional product intelligent recommendation technology, divides the family into multiple roles based on the pre-processed basic data, and mines the decision product results of each family role on the target maternal and infant products. According to the family role attribution and recommendation mode of the target account, the recommendation strategy of the target maternal and infant products is generated in combination with the decision product results, which solves the shortcomings of the existing product intelligent recommendation system in dealing with the family multi-role needs of maternal and infant products. Under the same target account, it realizes multi-role decision-making to provide differentiated recommendation content according to the needs of different family roles and meet the expectations of all parties. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other relevant drawings can be obtained based on these drawings without creative work. In the drawings: Figure 1 This is a flowchart of an intelligent recommendation method for maternal and infant products based on big data analysis; Figure 2This is a schematic diagram of the structure of the intelligent recommendation system for maternal and infant products based on big data analysis. DETAILED DESCRIPTION

[0019] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments and drawings. The exemplary embodiments of the present invention and their description are only used to explain the present invention and are not intended to limit the present invention.

[0020] Existing intelligent product recommendation systems are obviously insufficient in meeting the needs of multiple roles in the family for maternal and infant products. The needs of family members may conflict or have different priorities. For example, mothers may tend to choose high-end brand products, while fathers tend to choose economical and practical products. Traditional recommendation systems cannot effectively coordinate these differences, making it difficult for family members to jointly recognize the recommendation results, thereby reducing the practicality of the recommendation system and user satisfaction. In view of this, this solution provides the following embodiments to solve the above technical problems: Embodiment 1: This embodiment provides a method for intelligently recommending maternal and infant products based on big data analysis, such as Figure 1 As shown, including: Step 1: obtaining basic data of target maternal and infant products based on big data, and preprocessing the basic data; this step specifically includes the following methods: S11, obtain the basic data of the target maternal and infant products, the basic data includes: basic user information (such as age, gender, region, income, etc.), user behavior data (such as browsing data (such as browsing time, number of views), click data, purchase data (such as purchase results, purchase frequency, etc.) comment data, etc.), product information (such as brand, price, model, function, evaluation, etc.) and family structure data (such as the number of family members, roles, etc.); maternal and infant product discussions and user feedback on social media (such as Weibo, Xiaohongshu, etc.); user behavior data of maternal and infant apps (such as Babytree, Qinbaobao, etc.); industry reports and user portrait data from third-party data providers (such as Nielsen, iResearch, etc.). S12, cleaning and integrating the basic data; the integration includes integrating the basic data from different data sources, constructing a unified user data table based on users, and associating user behavior data, family structure data and product information in the user data table. This embodiment provides a user data table as shown in Table 1 below, which only shows some of the main basic data; Table 1 User data table ; Step 2: divide the family into multiple roles based on the pre-processed basic data, and mine the decision-making product results of each family role on the target maternal and infant products; this step specifically includes the following methods: S21, configure family roles including: mother user, father user, grandparent user and other family member users; S22, based on the preprocessed basic data, combined with the improved K-means clustering algorithm, family role division is performed; this step specifically includes the following methods: S221, taking each family role as a cluster, and setting the centroid of each cluster respectively; among them, the cluster of mother users takes the purchase frequency, browsing time and age as the centroid; the cluster of father users takes the customer unit price and keyword search as the centroid; the cluster of grandparent users takes the product purchase record and age as the centroid; the cluster of other family members users takes the brand awareness and packaging as the centroid; specifically, it can be set as follows: the cluster centroid of mother users is set to the purchase frequency of maternal and child necessities>5 times / month, and the age is set to 25-40 years old; the cluster centroid of father users is set to the customer unit price higher than a certain unit price threshold; the cluster centroid of grandparent users is set to the purchase record of low-priced goods; the cluster centroid of other family members users is set to brand awareness or packaging reaching certain requirements; S222, performing clustering iteration based on the preprocessed basic data until each cluster converges and then outputting each cluster and its centroid; the clustering iteration process includes: S223, extracting general features and key features of each user from the preprocessed basic data, assigning a first weight to the general features, and assigning a second weight to the key features; wherein the second weight is greater than the first weight; S224, calculate the distance from the current user to the centroid of the cluster based on the general features and key features. The distance function of the clustering iteration is: ; Where D(x, c) represents the distance from the current user to the centroid of the cluster; x i represents the i-th general feature of the current user; x j represents the jth key feature of the current user; β i represents the weight of the i-th general feature; β j represents the weight of the jth key feature; n represents the total number of general features; m represents the total number of key features; c i represents the cluster centroid of the i-th general feature; c j Represents the cluster centroid of the jth key feature; the weights of general features and key features are dynamically adjusted according to the differences in family roles.

[0021] The above distance function is mainly used to calculate the distance from numerical features to the cluster centroid. For categorical features (such as gender), the Hamming distance can be used based on the K-prototype algorithm to better handle mixed types of data. This solution designs the above distance function to capture role differences. For example, mother users and grandparent users may have different price sensitivities. Different weights are set on the price dimension to reflect the differences between mother users and grandparent users.

[0022] When calculating the distance from the current user to the centroid of the cluster, different general features or key features can be set to the same weight, or can be set to different weights according to actual conditions.

[0023] The number of family roles in this solution has been determined to be four roles: mother, father, grandparents, and other family members. During the clustering process, the behavioral data between roles may overlap, resulting in inaccurate clustering. For example, fathers and grandparents may have similarities in purchasing behavior. At this time, the algorithm needs to be improved to distinguish these groups. In view of this, this solution introduces general features and key features, and sets higher weights for the key features of each role to emphasize the unique behavioral characteristics of the role; for example, the key features of mother users can include the frequency of purchasing baby necessities, browsing related parenting knowledge pages, and searching for related keywords. General features can be region, income, age, etc.; the key features of father users can include browsing preferred products and searching for related keywords; the key features of grandparent users include: browsing related product promotion discount pages and price sensitivity; When configuring weights for general features and key features, weights can be assigned based on the importance of key features or general features through business experience or a random forest model. Compared with general features, higher weights are assigned to key features.

[0024] S23, constructing the original decision set of each family role on the target maternal and infant product according to the family role division result, and generating the decision product set from the original decision set. This step specifically includes the following method: Obtaining the family roles and decision results of all users; the decision results include purchase results and effective browsing results; the effective browsing results include the browsing time of the current product reaching a preset time threshold, and the number of clicks on the current product during the browsing time reaching a preset number threshold; the purchase results include actual purchase orders and recommendation results in social media; The decision results of all users with the same family role are combined to form the original decision set; After denoising the original decision set, the decision product set is obtained: the frequency of occurrence of the same decision product in the original decision set is counted, and the top N decision products with the largest occurrence frequency constitute the decision product set.

[0025] Due to the different needs of family members, mother users are usually more concerned about the safety, practicality and professionalism of the product; father users may pay more attention to the cost-effectiveness, technical content and brand reputation of the product; grandparent users may pay more attention to the price, durability and ease of operation of the product; other family members may pay more attention to the packaging and brand awareness of the product; this solution is based on multi-role decision-making and can provide differentiated recommendation content according to the needs of different family roles to meet the expectations of all parties.

[0026] Step 3: Obtain the family role attribution and recommendation model of the target account, and combine all decision-making product results to generate a recommendation strategy for the target maternal and infant products. This step specifically includes the following methods: S31, obtaining the family role and recommendation mode of the target account, wherein the recommendation mode includes an associated mode and an unassociated mode, and each recommendation mode includes two options: a single-role option or a multi-role option; the associated mode indicates that the generated recommendation strategy is associated with the family role of the target account (i.e., the account user is the person himself, and the person's preferences can be considered more when recommending), and the unassociated mode indicates that the generated recommendation strategy is not associated with the family role of the target account (i.e., the account user is the person himself, and the person's preferences can be considered more when recommending); S32, combining all decision product results, respectively generates recommendation strategies for each option under different recommendation modes; in the associated mode, the recommendation strategy uses the decision product results corresponding to the target account's family role as the main reference object, and the decision product results corresponding to a single role or multiple roles as the auxiliary reference object; in the unassociated mode, the recommendation strategy uses the decision product results corresponding to a single role or multiple roles as the reference object.

[0027] The method for generating the recommendation strategy includes: Get all decision product results; Regarding the single-role option in the non-association mode, the decision product of the decision product result corresponding to the current single-role option is used as the recommended strategy; For multi-role options in the non-association mode, weights are assigned to the decision product results corresponding to each role option, and all decision product results after weight assignment are merged to obtain a combined result set. The top N combined results with larger weights in the combined result set are used as the recommended strategies. For single-role options in association mode, weight W1 is configured for the decision product result corresponding to the current single-role option, and weight W2 is configured for the decision product result corresponding to the target account's family role. All decision product results after weight configuration are merged to obtain a combined result set, and the top N combined results with larger weights in the combined result set are used as the recommended strategy; where W2>W1; Regarding the multi-role options in the association mode, the weight Wi is configured for the decision product results corresponding to the role option i, and the weight W3 is configured for the decision product results corresponding to the target account family role. All decision product results after the weight configuration are merged to obtain a combined result set, and the top N combined results with larger weights in the combined result set are used as the recommended strategies; where W3>Wi.

[0028] When all decision product results after configuring weights are merged, some decision product results may correspond to different role options, and the same decision product result may have multiple weights. When merging, all weights of the same decision product result are added together. For example, product a of brand A appears as a decision product result in both the mother user and the grandparent user's decision product results, where the weight of the mother user's decision product result is 2.0 and the weight of the grandparent user's decision product result is 1.5; then in the combined result, the weight of product a of brand A is 3.5.

[0029] The association mode in this solution means that the currently generated recommendation strategy needs to correspond to the target account family role. For example, when the mother user uses her own account to purchase maternal and child products, the decision-making habits of the mother user need to be the main consideration when generating the recommendation strategy. Then the decision-making habits of other family roles may or may not be considered. When the decision-making habits of multiple other family roles are considered, it is multi-role. At this time, when generating the recommendation strategy, it is assumed that the decision-making power of the target account family role is significant, and the decision-making weights of other roles are small, so as to facilitate the generation of a decision-making strategy that satisfies the target account user himself. In addition, the decision-making weight of the target account and the decision-making weights of other roles can also be customized as needed. When only the decision-making habits of the target account family role are considered, it is a single role. The non-association mode means that the current The generated recommendation strategy does not need to correspond to the family role of the target account. For example, when users of multiple roles use the mother user account to purchase maternal and child products, the decision-making habits of the mother user do not need to be the main focus when generating the recommendation strategy. You can choose the decision-making habits of any one or more family roles. When considering the decision-making habits of multiple other family roles, it is multiple roles. At this time, when generating the recommendation strategy, the decision weight of the target account family role is equal to the decision weights of other roles by default, so as to generate a decision strategy that satisfies users of each role. At the same time, the decision weight of the target account and the decision weights of other roles can also be customized as needed; when only the decision habits of one family role are considered, it is a single role, which can be the mother user or any other family role user.

[0030] Traditional recommendation systems usually take a single user as a unit, ignoring the differences in needs of different roles in the family (such as mother, father, and grandparents); this solution accurately identifies the role and preferences of each family member through multi-role division of the family; generates personalized recommendation strategies for different roles to meet the diverse needs of multiple roles in the family. Traditional recommendation systems rely on a single data source (such as purchase records), with limited data dimensions, and it is difficult to fully reflect user needs; there is a lack of in-depth mining of user behavior (such as browsing, searching, and commenting); this solution integrates multi-source data (purchase records, browsing behavior, search keywords, comment sentiment, etc.) to build a comprehensive user portrait. Through big data analysis, we can explore the deep needs and preferences of users, improve the accuracy of recommendations, and dynamically generate recommendation strategies based on the role attribution and recommendation mode of the target account to meet the needs of family collaborative decision-making.

[0031] Example 2

[0032] This embodiment provides an intelligent recommendation system for maternal and infant products based on big data analysis. Figure 2 As shown, the method for intelligently recommending maternal and infant products based on big data analysis described in Example 1 is implemented; the system includes: A preprocessing module, used to obtain basic data of target maternal and infant products based on big data, and preprocess the basic data; The role decision module is used to divide the family into multiple roles based on the pre-processed basic data, and to mine the decision-making product results of each family role regarding the target maternal and child products; The strategy generation module is used to obtain the family role attribution and recommendation model of the target account, and combine the results of all decision-making products to generate a recommendation strategy for the target maternal and infant products.

[0033] Example 3

[0034] This embodiment provides a computer-readable medium on which a computer program is stored. The computer program is executed by a processor to implement the method for intelligently recommending maternal and infant products based on big data analysis as described in Embodiment 1; specifically, the following steps are performed: Step 1: Obtain basic data of target maternal and infant products based on big data, and pre-process the basic data; Step 2: Divide the family into multiple roles based on the pre-processed basic data, and mine the decision-making product results of each family role regarding the target maternal and infant products; Step 3: Obtain the family role attribution and recommendation model of the target account, and combine all decision-making product results to generate a recommendation strategy for the target maternal and infant products.

[0035] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An intelligent recommendation method for maternal and infant products based on big data analysis, characterized in that: include: Obtaining basic data of target maternal and infant products based on big data, and preprocessing the basic data; Based on the pre-processed basic data, the family is divided into multiple roles, and the decision-making product results of each family role regarding the target maternal and infant products are mined; Obtain the family role attribution and recommendation model of the target account, and combine all decision-making product results to generate a recommendation strategy for the target maternal and infant products.

2. The method for intelligently recommending maternal and infant products based on big data analysis according to claim 1, characterized in that: The basic data of the target maternal and infant products are obtained according to the big data, and the basic data are preprocessed; Included methods: Obtaining basic data of target maternal and infant products, the basic data including: basic user information, user behavior data, product information and family structure data; The basic data is cleaned and integrated; the integration includes integrating the basic data from different data sources, constructing a unified user data table based on users, and associating user behavior data, family structure data and product information in the user data table.

3. The method for intelligently recommending maternal and infant products based on big data analysis according to claim 2, characterized in that: The method divides the family into multiple roles based on the pre-processed basic data, and mines the decision-making product results of each family role regarding the target maternal and infant products; Included methods: Configure family roles including: mom user, dad user, grandparent user and other family members users; Based on the preprocessed basic data, the improved K-means clustering algorithm is used to divide family roles; According to the results of family role division, the original decision set of each family role regarding the target maternal and infant products is constructed, and the decision product set is generated from the original decision set.

4. The method for intelligently recommending maternal and infant products based on big data analysis according to claim 3, characterized in that: The method of dividing family roles based on the preprocessed basic data and combining the improved K-means clustering algorithm includes: Each family role is a cluster, and the centroid of each cluster is set separately. The centroid of the cluster of mother users is the purchase frequency, browsing time and age; the centroid of the cluster of father users is the customer unit price and keyword search; the centroid of the cluster of grandparent users is the product purchase record and age; the centroid of the cluster of other family members is the brand awareness and packaging; Perform clustering iteration based on the preprocessed basic data until each cluster converges and then outputs each cluster and centroid; the clustering iteration process includes: Extracting general features and key features of each user from the preprocessed basic data, assigning a first weight to the general features and a second weight to the key features; wherein the second weight is greater than the first weight; The distance from each user to the cluster centroid is calculated based on the general feature, the first weight, the key feature and the second weight.

5. The method for intelligently recommending maternal and infant products based on big data analysis according to claim 4, characterized in that: The calculation method of the distance D(x,c) from the current user to the cluster centroid includes: ； Among them, x i represents the i-th general feature of the current user; x j represents the jth key feature of the current user; β i represents the weight of the i-th general feature; β j represents the weight of the jth key feature; n represents the total number of general features; m represents the total number of key features; c i represents the cluster centroid of the i-th general feature; c j represents the cluster centroid of the jth key feature.

6. The method for intelligently recommending maternal and infant products based on big data analysis according to claim 5, characterized in that: The method constructs an original decision set of each family role on the target maternal and infant product according to the family role division result, and generates a decision product set from the original decision set; Included methods: Obtaining the family roles and decision results of all users; the decision results include purchase results and effective browsing results; the effective browsing results include that the browsing time of the current product reaches a preset time threshold, and the number of clicks on the current product during the browsing time reaches a preset number threshold; The decision results of all users with the same family role are combined to form the original decision set; After denoising the original decision set, the decision product set is obtained: the frequency of occurrence of the same decision product in the original decision set is counted, and the top N decision products with the largest occurrence frequency constitute the decision product set.

7. The method for intelligently recommending maternal and infant products based on big data analysis according to claim 1, characterized in that: The target account's family role attribution and recommendation model are obtained, and a recommendation strategy for the target maternal and infant products is generated in combination with all decision-making product results; Included methods: Obtaining the family role and recommendation mode of the target account, wherein the recommendation mode includes an associated mode and an unassociated mode, and each recommendation mode includes two options: a single role option or a multi-role option; Combine all decision-making product results to generate recommendation strategies for each option under different recommendation modes; In the association mode, the recommendation strategy takes the decision-making product results corresponding to the target account's family role as the main reference object, and the decision-making product results corresponding to a single role or multiple roles as the auxiliary reference object; In the non-association mode, the recommendation strategy uses the decision product results corresponding to a single role or multiple roles as a reference object.

8. The method for intelligently recommending maternal and infant products based on big data analysis according to claim 7, characterized in that: The method for generating the recommendation strategy includes: Get all decision product results; For single-role options in non-association mode, the decision product in the decision product result corresponding to the current single-role option is used as the recommended strategy; For multi-role options in the non-association mode, weights are assigned to the decision product results corresponding to each role option, and all decision product results after weight assignment are merged to obtain a combined result set. The top N combined results with larger weights in the combined result set are used as the recommended strategies. For single-role options in association mode, weight W1 is configured for the decision product result corresponding to the current single-role option, and weight W2 is configured for the decision product result corresponding to the target account's family role. All decision product results after weight configuration are merged to obtain a combined result set, and the top N combined results with larger weights in the combined result set are used as the recommended strategy; where W2>W1; Regarding the multi-role options in the association mode, the weight Wi is configured for the decision product results corresponding to the role option i, and the weight W3 is configured for the decision product results corresponding to the target account family role. All decision product results after the weight configuration are merged to obtain a combined result set, and the top N combined results with larger weights in the combined result set are used as the recommended strategies; where W3>Wi.

9. An intelligent recommendation system for maternal and infant products based on big data analysis, characterized in that: Used to implement the intelligent recommendation method for maternal and infant products based on big data analysis as described in any one of claims 1 to 8; the system comprises: A preprocessing module, used to obtain basic data of target maternal and infant products based on big data, and preprocess the basic data; The role decision module is used to divide the family into multiple roles based on the pre-processed basic data, and to mine the decision-making product results of each family role regarding the target maternal and child products; The strategy generation module is used to obtain the family role attribution and recommendation model of the target account, and combine the results of all decision-making products to generate a recommendation strategy for the target maternal and infant products.

10. A computer readable medium having a computer program stored thereon, characterized in that: The computer program is executed by a processor to implement the intelligent recommendation method for maternal and infant products based on big data analysis as described in any one of claims 1 to 8.

Images