E-commerce operation management system based on cloud platform

Abstract

The invention relates to the field of e-commerce operation management systems, in particular to an e-commerce operation management system based on a cloud platform, and is used for solving the problems that in a traditional e-commerce operation management system, the recommendation coverage range is wide, the accuracy is poor, the flow is wasted, frequent useless recommendation may cause user dislike, and the user experience is poor. And useless recommendation cannot improve the sales volume of the e-commerce platform, matching recommendation cannot be performed on commodities purchased by the user, the user is not facilitated to maintain the purchased commodities, and the user experience is easy to cause. The e-commerce operation management system comprises the following modules: a data collection module, a user portrait module, a conventional recommendation module and an intelligent recommendation module. According to the e-commerce operation management system, a conventional recommendation and intelligent recommendation complementary strategy is adopted, the dependence of the user on the platform can be better increased, and the commodity sales volume of the platform is improved.

Description

Technical Field

[0001] The present invention relates to the field of e-commerce operation management systems, and in particular to an e-commerce operation management system based on a cloud platform. Background Art

[0002] E-commerce is one of the mainstream ways of shopping in today's society, so many e-commerce platforms have been born. Due to the competition between e-commerce platforms, e-commerce platforms need to develop a complete operation and management system to maintain users and increase platform sales. Among them, user personalized recommendations are a function that can use user tags, such as age, occupation, interests and hobbies for personalized recommendations. Personalized recommendations can also be made based on users' various search and browsing behavior data.

[0003] The patent with the publication number CN115080865A discloses an e-commerce data operation and management system based on multidimensional data analysis. The system includes: a data acquisition module: used to obtain the browsing record information of the user; a feature extraction module: used to obtain the text area and the picture area in the browsing image, and obtain the texture style feature index based on the channel value of the text feature pixel point set and the picture feature pixel point set in each feature channel image corresponding to the feature image after each downsampling of the feature extraction encoder; obtain the color distribution feature based on the value of the picture area pixel point in each channel; a management recommendation module: used to select candidate products from the products in the database to build a candidate product library; based on the image features of the browsing image and the image features of each product in the candidate product library, select the recommended products from the candidate product library. The present invention can improve the recommendation accuracy of products.

[0004] However, in traditional e-commerce operation and management systems, personalized recommendations can only be made for mainstream products, user tags, and routine recommendations based on user behavior data. The recommendations cover a wide range but have poor accuracy, which not only wastes traffic, but frequent useless recommendations may cause user disgust. Moreover, useless recommendations will not increase the sales volume of the e-commerce platform. It is impossible to make supporting recommendations for products after the user has purchased them, which is not conducive to users maintaining the purchased products, and can easily cause poor user experience.

[0005] To this end, this case proposes an e-commerce operation management system based on a cloud platform to solve the above technical problems. Summary of the invention

[0006] In order to overcome the above-mentioned technical problems, the purpose of the present invention is to provide an e-commerce operation and management system based on a cloud platform: by adding an intelligent recommendation strategy based on conventional recommendations, timely predicting the time when a user purchases a product for the second time, and calculating the time when the user needs supporting products based on the products already purchased by the user, timely recommendations are made, and a recommendation strategy within the shelf life and a recommendation strategy for the replacement of main products are formulated for consumables xhs. This solves the problem that in traditional e-commerce operation and management systems, personalized recommendations can only be made for mainstream products, user labels, and conventional recommendations based on user behavior data. The recommendations cover a wide range but have poor accuracy, which not only wastes traffic, but frequent useless recommendations may cause user disgust, and useless recommendations will not increase the sales of the e-commerce platform. It is impossible to make supporting recommendations for products after the user has purchased them, which is not conducive to the user maintaining the purchased products, and can easily cause poor user experience.

[0007] The purpose of the present invention can be achieved through the following technical solutions: An e-commerce operation management system based on a cloud platform, comprising: The data collection module is used to collect user registration information and user behavior data, where the registration information includes user age, gender, occupation, and address, and the user behavior data includes browsing behavior data, search behavior data, favorite / follow behavior data, and purchase behavior data, and upload all data to the cloud platform; The user portrait module creates multi-dimensional data labels based on user data collected by the cloud platform, establishes basic labels based on user registration information, creates consumer labels of the same level for users of the same age and gender, creates preference approximation labels for users with similar behavior data, creates consumption capacity labels for users who purchase the same products in similar price ranges, and uses the K-means clustering algorithm to perform cluster analysis on user labels, classifying users with similar characteristics and behavior patterns into one category to form different user group portraits; The conventional recommendation module matches product recommendations based on the portrait tags, and marks the products corresponding to the browsing behavior data, search behavior data, and collection / follow behavior data as unpurchased products wgs, and the products corresponding to the purchase behavior data as purchased products ygs, and then makes conventional recommendations based on the unpurchased products wgs through the conventional recommendation strategy; The intelligent recommendation module establishes a secondary recommendation strategy for the purchased goods ygs to predict the secondary recommendation time, and marks the goods without ancillary products in the purchased goods ygs as consumables xhs, and the goods with ancillary products as matching products pts. Matching products pts include replacement parts, repair and maintenance tools, and maintenance and decoration products. For matching product pts, a matching recommendation strategy is established and a prediction recommendation is made As a further solution of the present invention: the conventional recommendation module formulates a conventional recommendation strategy based on the user portrait, which is as follows: For browsing behavior data, products with a browsing time of less than 5 seconds and a browsing frequency of less than 2 times are marked as accidentally clicked products ws, products with a browsing frequency of 2-5 times and an average browsing time of 10-40 seconds are marked as initially preferred products cs, and products with a browsing frequency of more than 5 times and a browsing time of more than 40 seconds are marked as key preferred products zs; For search behavior data, the products clicked after the search are marked as the desired products ys, the products related to the search keywords are marked as recommended products ks, and the products added to the shopping cart after the search are included in the collection / follow behavior data; According to the collection / attention behavior data, these products are marked as key recommended products. The conventional recommendation module formulates a conventional recommendation plan based on the conventional recommendation strategy for recommendation.

[0008] As a further solution of the present invention: the conventional recommendation plan is as follows: First, we collect the behavior data of accidentally clicked products ws, initially preferred products cs, key preferred products zs, intended products ys, recommended products ks, and key recommended products zds. Since these products are all unpurchased products wgs, they are recommended to users every day. The recommendation list is based on the rule list of key recommended products zds> key preferred products zs> intended products ys> initially preferred products cs> recommended products ks> accidentally clicked products ws. In the recommendation process, the total weights of the six types of goods are calculated as 1, among which the key recommended goods zds are 0.35, the key inclined goods zs are 0.20, the intended purchase goods ys are 0.15, the preliminary inclined goods cs are 0.15, the recommendable goods ks are 0.10, and the accidentally clicked goods ws are 0.05. The recommended goods are allocated on the recommendation page according to the weight values, and all unpurchased goods wgs are recommended every day with one week as a unit. Based on the daily recommendation, the user click data is collected again, and the clicked goods are re-marked as inclined goods yqs. The products that are not clicked every day in a week are re-marked as abandoned goods dfs. In the second week, the recommendation priority of inclined goods yqs>to-be-abandoned goods dfs is used for recommendation. Among them, the inclined goods yqs are still allocated with the recommendation share according to their respective weights and recommended daily, and the to-be-abandoned goods dfs are recommended every other day according to their respective weights. This cycle repeats. The goods that users have not clicked for 1.5-3 months are marked as not recommended goods bts and are no longer recommended.

[0009] As a further solution of the present invention: the recommendation strategy of the intelligent recommendation module for the purchased product ygs is as follows, wherein the purchased product ygs includes the consumable product xhs body and the matching product pts body: Collect the product information of the user's purchased products ygs, screen the user's historical purchase data, and if the same or the same product appears, it is determined that the user has purchased the product twice or more. Collect the time of each purchase of the user, analyze the user's historical purchase time interval sequence, and mark it as T1, T2, ..., Tn, where T1 is the first purchase time and Tn is the most recent purchase time, and then calculate the average purchase time interval , the calculation formula is , then the next purchase time The calculation formula is, ,in The time of the most recent purchase, based on the next purchase time The base date is moved forward by 7 days, and the resulting date is the recommended time for the purchased goods for the second purchase. If no similar or identical goods appear, it is determined that the user is purchasing the goods for the first time. For users who purchase the goods for the first time, the cycle is converted into days based on the average replacement cycle of the goods, and the days are added to the date of the user's first purchased goods to obtain the predicted second purchase date of the user. Based on the predicted second purchase date of the user, the date is moved forward by 30 days, and the resulting date is the recommended time for the purchased goods for the first purchase. The intelligent recommendation module regularly makes second purchase recommendations for the purchased goods, and after the recommendation, the conventional recommendation module continues to collect user behavior data, and recommends the collected user behavior data again according to the conventional recommendation strategy.

[0010] As a further solution of the present invention: the intelligent recommendation module establishes a matching recommendation strategy for matching product pts, and the specific strategy is as follows: Obtain all product information on the platform, establish a supporting product recommendation database for all product information, connect the recommendation database with each product manufacturer, obtain detailed information on various products from the product manufacturers, including the main product and the corresponding original replacement parts, repair and maintenance tools, and maintenance and decoration products, and establish the product table as follows: Create a main product table, including the main product ID, name, model, brand, and category fields; Create a replacement parts table and associate it with the main product table, including the parts ID, parts name, model, brand, applicable main product ID, original manufacturer identification, and replacement cycle fields; Create a maintenance tool table and associate it with the main product table, including fields such as tool ID, tool name, model, brand, applicable main product ID, and recommended usage period; Create a maintenance and decoration product table and associate it with the main product table, including product ID, product name, model, brand, applicable main product ID, and recommended matching cycle fields; After obtaining the information of the matching items pts in the user's purchased product ygs, the information includes product ID, name, model, brand, and category. The matching items pts are matched with the main product table in the recommendation database to obtain the replacement parts, repair and maintenance tools, and maintenance and decoration products related to the matching items pts. The recommended use period of the related replacement parts, repair and maintenance tools, and maintenance and decoration products is converted into days, and the days are added to the purchase date of the main product. The demand date of the replacement parts, repair and maintenance tools, and maintenance and decoration products is calculated according to the current calendar. On the basis of this date, 14 days are moved forward. The obtained date is the recommended date of the replacement parts, repair and maintenance tools, and maintenance and decoration products related to the matching items pts in the purchased product ygs. The intelligent recommendation module makes recommendations after expiration. When recommending, if there are more than two matching products for the matching items pts, the priority order of replacement parts>repair and maintenance tools>maintenance and decoration products is used for recommendation. After the recommendation, the conventional recommendation module continues to collect user behavior data, and the collected user behavior data is recommended again according to the conventional recommendation strategy.

[0011] As a further solution of the present invention: the intelligent recommendation module formulates a comprehensive recommendation strategy for original parts and non-original parts based on the principle of recommendation in the priority order of replacement parts > repair and maintenance tools > maintenance and decoration products, as follows: Obtain the main product information of the matching products pts, and collect all users who have purchased the product within 12 months before the current time on the platform, track the proportion of original and non-original parts purchased by these users, as well as the brand, model, price k, and evaluation of the non-original parts purchased, and divide the priority of original and non-original parts according to the obtained original and non-original parts ratio. If the original parts ratio is greater than the non-original parts ratio, increase the recommended share of original parts, otherwise increase the share of non-original parts. Based on this principle, eliminate matching products with a non-original parts negative review rate exceeding 40%.

[0012] As a further solution of the present invention: based on the remaining 60% of non-original parts, if the supporting products of the matching product pts have original parts, the price of the original parts is marked as β, and then the price of the original parts price β is calculated by 30% up and 20% down respectively, and the calculation method is as follows: 30% price increase = , price reduction of 20% = , get the price range, and then arrange the remaining 60% of non-original parts prices k in descending order, select non-original parts with price k between 30% increase and 20% decrease of original parts price, and recommend the non-original parts in this range according to the priority of price in descending order, that is, the higher the price, the higher the recommended share, and vice versa. After the recommendation, the conventional recommendation module continues to collect user behavior data, and the collected user behavior data is recommended again according to the conventional recommendation strategy.

[0013] As a further solution of the present invention: based on the remaining 60% of non-original parts, if the supporting products of the matching product pts do not have original parts, the price of the matching product pts is marked as P 主商品价格 Then get the minimum price range of the same category of the main product and mark it as P 最低价格 , the maximum value is marked as P 最高价格 , calculate the relative position ratio of the price of the main product purchased by the user within the price range of products of the same category, the calculation formula is: , P 倾向值 Between 0 and 1, the closer it is to 0, the more the user prefers low-priced products, and the closer it is to 1, the more the user prefers high-priced products. 倾向值 To calculate the price preference range of users for supporting products, the same principle of 30% increase and 20% decrease is used. The calculation formula is as follows: ; ; Get P from the remaining 60% of the non-original parts price K 配套品下限 and P 配套品上限 Price range, non-original parts within this range are recommended in descending order of price priority. After the recommendation, the conventional recommendation module continues to collect user behavior data and recommends the collected user behavior data again according to the conventional recommendation strategy.

[0014] As a further solution of the present invention: the intelligent recommendation module formulates a recommendation strategy within the shelf life for consumables xhs without ancillary products in the purchased goods ygs, which is as follows: Collect the information of consumables xhs, obtain its shelf life, and obtain the expiration time of consumables xhs through the calculation formula of factory date + shelf life = expiration time. Since the use cycle of consumables xhs is short, no user feedback can be obtained after the prediction recommendation, and the conventional recommendation module regards it as a non-recommended product bts. The intelligent recommendation module moves forward 30 days based on the expiration time of consumables xhs as the reminder recommendation time, and determines in real time through user purchase behavior data whether the user has purchased the consumables xhs twice within the shelf life. When the judgment result shows that there is a second purchase, the reminder recommendation time is cancelled, and the expiration time of the most recent purchase of consumables xhs is recalculated to set the reminder recommendation time. When the judgment result shows that there is no second purchase, after the reminder recommendation time arrives, the intelligent recommendation module triggers the recommendation mechanism for recommendation, and sets the reminder words of the expiration date approaching on the recommendation page. After the recommendation, the conventional recommendation module continues to collect user behavior data, and the collected user behavior data is recommended again according to the conventional recommendation strategy.

[0015] As a further solution of the present invention: the intelligent recommendation module is based on the main product table in the recommendation database. When the manufacturer updates the products in the main product table, the intelligent recommendation module first determines the applicable grade, occupation and function of the product based on the main product information, and then determines the user tags for the updated product to be recommended. The priority adopts the strategy of occupation applicability > interest and hobby applicability > age applicability for trial recommendation. The trial recommendation method is daily recommendation for occupation applicability, every other day recommendation for interest and hobby applicability, and weekly recommendation for age applicability. According to this recommendation method, the regular recommendation module continuously collects user behavior data, and the collected user behavior data is recommended again according to the regular recommendation strategy.

[0016] Beneficial effects of the present invention: The e-commerce operation management system based on the cloud platform of the present invention adds multiple intelligent recommendation strategies on the basis of conventional recommendation; 1. For purchased products, YGS can timely predict the time of users' second purchase, and calculate the time of second recommendation of purchased products to users through the second purchase time. On the one hand, it is convenient for users to purchase in advance when the product usage cycle is about to end, avoiding the gap in product usage, improving user experience, and also indirectly increasing platform sales and profits; 2. For purchased products, YGS has formulated different recommendation plans based on the first purchase and second purchase or above of the user, so as to avoid neglecting the second purchase time of the first-time purchaser, increase the probability of the user's second purchase of the product on the platform, and at the same time still avoid the user's gap in using the product; 3. A matching recommendation strategy has been established for the matching products pts in the user's purchased products ygs. It can predict and recommend products with matching products in the user's purchased products ygs. The matching products include accessories, repair and maintenance tools, and maintenance and decoration products. On the one hand, it can remind users to replace accessories and maintain the purchased products ygs in time. On the other hand, it can recommend matching repair and maintenance tools to users when the products may be damaged. When users browse the recommended matching products, they can also increase the impression of matching products, repair tools, and maintenance and decoration products needed for the purchased products ygs. When users need repair and maintenance, maintenance and decoration, it can facilitate their search for related matching products, further improve user experience, and increase user dependence on the platform; 4. A comprehensive recommendation strategy for original and non-original parts has been formulated for the related supporting products of the matching product pts. On the one hand, it ensures that users who pursue original parts can receive the recommended original parts information. On the other hand, users who have unique insights into product supporting products can also receive non-original parts that match the main product, which increases the user's purchase selection range and avoids the problem of user loss or purchase order loss due to users not seeing the relevant non-original supporting products; 5. For the recommendation of non-original parts, we combine the price range of the main product purchased by the user in this category to calculate the user's price propensity value, and select products with a price increase of 30% and a price decrease of 20% from non-original parts according to the propensity value for priority recommendation in descending order, so that the user's choices are more diversified and more in line with the user's label, increasing the user's dependence on the platform; 6. In the recommendation of supporting products, supporting products are connected with manufacturers through the database. On the one hand, it can reduce the workload of the platform in processing product information. On the other hand, manufacturers can also update their own product information in time for better sales, ensuring the completeness and timeliness of product information; 7. A recommendation strategy within the shelf life has been formulated for consumables xhs in the purchased goods ygs. When users purchase consumables xhs and do not make a second purchase within the shelf life, there is a problem that the user forgets to use or uses it slowly, causing the product to expire or the user is unwilling to use it. The function of recommending consumables xhs within the shelf life can remind users who forget to use or use it slowly before the expiration date of the product, and use it in time or avoid using expired products. On the other hand, the recommendation within the shelf life determines whether the user needs the product based on the user behavior data, avoiding the problem of mis-push later affecting the user's shopping experience; 8. Combined with the main product table of the recommendation database, when products are updated, combined with user standards, a strategy of occupational applicability > hobby applicability > age applicability is used for trial recommendation. By recommending updated main products to specific users, on the one hand, it is convenient for users who tend to pursue updated products to make timely recommendations, and on the other hand, it can also increase product sales and improve platform profits; 9. All regular recommendation modules for products recommended by the intelligent recommendation modules are updated in real time by continuously collecting user behavior data, thereby continuously optimizing the recommendation results, improving the fit of recommended products to user purchasing needs, and increasing the probability of users purchasing recommended products. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The present invention will be further described below in conjunction with the accompanying drawings.

[0018] Figure 1 It is a principle block diagram of an e-commerce operation management system based on a cloud platform in the present invention. DETAILED DESCRIPTION

[0019] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] Embodiment 1: See also Figure 1 As shown, this embodiment is an e-commerce operation management system based on a cloud platform, including the following modules: a data collection module, a user portrait module, a conventional recommendation module, and an intelligent recommendation module; The data collection module is used to collect user registration information and user behavior data, wherein the registration information includes user age, gender, occupation, and address, and the user behavior data includes browsing behavior data, search behavior data, collection / attention behavior data, and purchase behavior data, and upload all data to the cloud platform; Among them, the user portrait module creates multi-dimensional data labels based on user data collected by the cloud platform, establishes basic labels based on user registration information, creates consumer labels of the same level for users of the same age and gender, creates preference approximation labels for users with similar behavior data, creates consumption capacity labels for users who purchase the same products in similar price ranges, and uses the K-means clustering algorithm to perform cluster analysis on user labels, classifying users with similar characteristics and behavior patterns into one category to form different user group portraits; Among them, the conventional recommendation module matches product recommendations based on the portrait tags, and marks the products corresponding to the browsing behavior data, search behavior data, and collection / follow behavior data as unpurchased products wgs, and the products corresponding to the purchase behavior data as purchased products ygs, and then makes conventional recommendations based on the unpurchased products wgs through the conventional recommendation strategy; Among them, the intelligent recommendation module establishes a secondary recommendation strategy for the purchased goods ygs to predict the secondary recommendation time, and marks the goods without ancillary products in the purchased goods ygs as consumables xhs, and the goods with ancillary products as matching products pts, where matching products pts include replacement parts, repair and maintenance tools, and maintenance and decoration products of the goods. A matching recommendation strategy is established for the matching products pts and predictive recommendations are made.

[0021] It should be noted that the data collection module, user portrait module, general recommendation module, and intelligent recommendation module all perform data storage, calculation, and processing based on the cloud server platform.

[0022] Embodiment 2: See also Figure 1 As shown, based on Example 1, it is further obtained that the conventional recommendation module formulates a conventional recommendation strategy based on the user portrait, which is specifically as follows: For browsing behavior data, products with a browsing time of less than 5 seconds and a browsing frequency of less than 2 times are marked as accidentally clicked products ws, products with a browsing frequency of 2-5 times and an average browsing time of 10-40 seconds are marked as initially preferred products cs, and products with a browsing frequency of more than 5 times and a browsing time of more than 40 seconds are marked as key preferred products zs; For search behavior data, the products clicked after the search are marked as the desired products ys, the products related to the search keywords are marked as recommended products ks, and the products added to the shopping cart after the search are included in the collection / follow behavior data; According to the collection / attention behavior data, these products are marked as key recommended products. The conventional recommendation module formulates a conventional recommendation plan based on the conventional recommendation strategy for recommendation.

[0023] The recommended routine schedule is as follows: First, we collect the behavior data of accidentally clicked products ws, initially preferred products cs, key preferred products zs, intended products ys, recommended products ks, and key recommended products zds. Since these products are all unpurchased products wgs, they are recommended to users every day. The recommendation list is based on the rule list of key recommended products zds> key preferred products zs> intended products ys> initially preferred products cs> recommended products ks> accidentally clicked products ws. In the recommendation process, the total weights of the six types of goods are calculated as 1, among which the key recommended goods zds are 0.35, the key inclined goods zs are 0.20, the intended purchase goods ys are 0.15, the preliminary inclined goods cs are 0.15, the recommendable goods ks are 0.10, and the accidentally clicked goods ws are 0.05. The recommended goods are allocated on the recommendation page according to the weight values, and all unpurchased goods wgs are recommended every day with one week as a unit. Based on the daily recommendation, the user click data is collected again, and the clicked goods are re-marked as inclined goods yqs. The products that are not clicked every day in a week are re-marked as abandoned goods dfs. In the second week, the recommendation priority of inclined goods yqs>to-be-abandoned goods dfs is used for recommendation. Among them, the inclined goods yqs are still allocated with the recommendation share according to their respective weights and recommended daily, and the to-be-abandoned goods dfs are recommended every other day according to their respective weights. This cycle repeats. The goods that users have not clicked for 1.5-3 months are marked as not recommended goods bts and are no longer recommended.

[0024] It should be noted that regular recommendations are based on detailed labeling and classification strategies of user behavior, which can more accurately grasp users' purchasing tendencies and improve the accuracy and effectiveness of regular recommendations. By re-collecting and labeling, dynamic adjustment of recommendation strategies can better adapt to changes in user behavior, improve the accuracy and timeliness of recommendations, and thus enhance users' shopping experience.

[0025] Example 3 See also Figure 1 As shown, based on Examples 1 and 2, it is further obtained that the recommendation strategy of the intelligent recommendation module for the purchased product ygs is as follows, wherein the purchased product ygs includes the consumable xhs body and the matching product pts body: Collect the product information of the user's purchased products ygs, screen the user's historical purchase data, and if the same or the same product appears, it is determined that the user has purchased the product twice or more. Collect the time of each purchase of the user, analyze the user's historical purchase time interval sequence, and mark it as T1, T2, ..., Tn, where T1 is the first purchase time and Tn is the most recent purchase time, and then calculate the average purchase time interval , the calculation formula is , then the next purchase time The calculation formula is, ,in The time of the most recent purchase, based on the next purchase time The base date is moved forward by 7 days, and the resulting date is the recommended time for the purchased goods for the second purchase. If no similar or identical goods appear, it is determined that the user is purchasing the goods for the first time. For users who purchase the goods for the first time, the cycle is converted into days based on the average replacement cycle of the goods, and the days are added to the date of the user's first purchased goods to obtain the predicted second purchase date of the user. Based on the predicted second purchase date of the user, the date is moved forward by 30 days, and the resulting date is the recommended time for the purchased goods for the first purchase. The intelligent recommendation module regularly makes second purchase recommendations for the purchased goods, and after the recommendation, the conventional recommendation module continues to collect user behavior data, and recommends the collected user behavior data again according to the conventional recommendation strategy.

[0026] The intelligent recommendation module establishes a matching recommendation strategy for matching product pts. The specific strategy is as follows: Obtain all product information on the platform, establish a supporting product recommendation database for all product information, connect the recommendation database with each product manufacturer, obtain detailed information on various products from the product manufacturers, including the main product and the corresponding original replacement parts, repair and maintenance tools, and maintenance and decoration products, and establish the product table as follows: Create a main product table, including the main product ID, name, model, brand, and category fields; Create a replacement parts table and associate it with the main product table, including the parts ID, parts name, model, brand, applicable main product ID, original manufacturer identification, and replacement cycle fields; Create a maintenance tool table and associate it with the main product table, including fields such as tool ID, tool name, model, brand, applicable main product ID, and recommended usage period; Create a maintenance and decoration product table and associate it with the main product table, including product ID, product name, model, brand, applicable main product ID, and recommended matching cycle fields; After obtaining the information of the matching items pts in the user's purchased product ygs, the information includes product ID, name, model, brand, and category. The matching items pts are matched with the main product table in the recommendation database to obtain the replacement parts, repair and maintenance tools, and maintenance and decoration products related to the matching items pts. The recommended use period of the related replacement parts, repair and maintenance tools, and maintenance and decoration products is converted into days, and the days are added to the purchase date of the main product. The demand date of the replacement parts, repair and maintenance tools, and maintenance and decoration products is calculated according to the current calendar. On the basis of this date, 14 days are moved forward. The obtained date is the recommended date of the replacement parts, repair and maintenance tools, and maintenance and decoration products related to the matching items pts in the purchased product ygs. The intelligent recommendation module makes recommendations after expiration. When recommending, if there are more than two matching products for the matching items pts, the priority order of replacement parts>repair and maintenance tools>maintenance and decoration products is used for recommendation. After the recommendation, the conventional recommendation module continues to collect user behavior data, and the collected user behavior data is recommended again according to the conventional recommendation strategy.

[0027] The intelligent recommendation module formulates a comprehensive recommendation strategy for original and non-original parts based on the principle of recommendation in the order of priority of replacement parts > repair and maintenance tools > maintenance and decoration products, as follows: Obtain the main product information of the matching products pts, and collect all users who have purchased the product within 12 months before the current time on the platform, track the proportion of original and non-original parts purchased by these users, as well as the brand, model, price k, and evaluation of the non-original parts purchased, and divide the priority of original and non-original parts according to the obtained original and non-original parts ratio. If the original parts ratio is greater than the non-original parts ratio, increase the recommended share of original parts, otherwise increase the share of non-original parts. Based on this principle, eliminate matching products with a non-original parts negative review rate exceeding 40%.

[0028] Based on the remaining 60% of non-original parts, if the supporting products of the matching product pts have original parts, the price of the original parts is marked as β, and then the price of the original parts price β is calculated by 30% up and 20% down respectively. The calculation method is as follows: 30% price increase = , price reduction of 20% = , get the price range, and then arrange the remaining 60% of non-original parts prices k in descending order, select non-original parts with price k between 30% increase and 20% decrease of original parts price, and recommend the non-original parts in this range according to the priority of price in descending order, that is, the higher the price, the higher the recommended share, and vice versa. After the recommendation, the conventional recommendation module continues to collect user behavior data, and the collected user behavior data is recommended again according to the conventional recommendation strategy.

[0029] It should be noted that, assuming the price of original parts is 100, according to the above formula, the price that increases by 30% is 130, and the price that decreases by 20% is 80. Therefore, the recommended price k of non-original parts ranges from 130 to 80.

[0030] Based on the remaining 60% of non-original parts, if the supporting products of the matching product pts do not have original parts, the price of the matching product pts is marked as P 主商品价格 Then get the minimum price range of the same category of the main product and mark it as P 最低价格 , the maximum value is marked as P 最高价格 , calculate the relative position ratio of the price of the main product purchased by the user within the price range of products of the same category, the calculation formula is: , P 倾向值 Between 0 and 1, the closer it is to 0, the more the user prefers low-priced products, and the closer it is to 1, the more the user prefers high-priced products. 倾向值 To calculate the price preference range of users for supporting products, the same principle of 30% increase and 20% decrease is used. The calculation formula is as follows: ; ; Get P from the remaining 60% of the non-original parts price K 配套品下限 and P 配套品上限 Price range, non-original parts within this range are recommended in descending order of price priority. After the recommendation, the conventional recommendation module continues to collect user behavior data and recommends the collected user behavior data again according to the conventional recommendation strategy.

[0031] It should be noted that, assuming that a user buys a refrigerator P 主商品价格 The lowest price of refrigerators on the platform is P 最低价格 is 1000, the highest price P 最高价格 is 20000, then , plus the lowest price P of a certain accessory of the refrigerator 配套品最低价格 is 100, the highest price of the supporting product is P 配套品最高价格 is 600, then the price range of non-original accessories preferred by users is as follows: ; ; We recommend non-original parts in the price range of 383.5-133.5 to users.

[0032] The intelligent recommendation module is based on the main product table in the recommendation database. When the manufacturer updates the products in the main product table, the intelligent recommendation module first determines the applicable grade, occupation and function of the product based on the main product information, and then determines the user tags for the updated product to be recommended. The priority adopts the strategy of occupation applicability > interest and hobby applicability > age applicability for trial recommendation. The trial recommendation method is daily recommendation for occupation applicability, every other day recommendation for interest and hobby applicability, and weekly recommendation for age applicability. According to this recommendation method, the regular recommendation module continuously collects user behavior data, and the collected user behavior data is recommended again according to the regular recommendation strategy.

[0033] It should be noted that by recommending updated main products to specific users, on the one hand, it is convenient for users who tend to pursue updated products to make timely recommendations, and on the other hand, it can also increase product sales and improve platform profits.

[0034] Example 4 See also Figure 1 As shown, based on the examples 1, 2, and 3, it is further obtained that the intelligent recommendation module formulates a recommendation strategy within the shelf life for the consumables xhs without ancillary products in the purchased goods ygs, which is as follows: Collect the information of consumables xhs, obtain its shelf life, and obtain the expiration time of consumables xhs through the calculation formula of factory date + shelf life = expiration time. Since the use cycle of consumables xhs is short, no user feedback can be obtained after the prediction recommendation, and the conventional recommendation module regards it as a non-recommended product bts. The intelligent recommendation module moves forward 30 days based on the expiration time of consumables xhs as the reminder recommendation time, and determines in real time through user purchase behavior data whether the user has purchased the consumables xhs twice within the shelf life. When the judgment result shows that there is a second purchase, the reminder recommendation time is cancelled, and the expiration time of the most recent purchase of consumables xhs is recalculated to set the reminder recommendation time. When the judgment result shows that there is no second purchase, after the reminder recommendation time arrives, the intelligent recommendation module triggers the recommendation mechanism for recommendation, and sets the reminder words of the expiration date approaching on the recommendation page. After the recommendation, the conventional recommendation module continues to collect user behavior data, and the collected user behavior data is recommended again according to the conventional recommendation strategy.

[0035] It should be noted that when users purchase consumables xhs and do not make a second purchase within the shelf life, there is a problem that the user forgets to use the product or uses it slowly, causing the product to expire or the user is unwilling to use the product. The function of recommending consumables xhs within the shelf life can remind users who forget to use the product or use it slowly about the expiration date of the product, and use it in time or avoid using expired products. On the other hand, the recommendation within the shelf life determines whether the user needs the product based on user behavior data, to avoid the problem of erroneous push later affecting the user's shopping experience.

[0036] Example 5 An e-commerce operation management system based on a cloud platform. When in use, a data collection module collects user registration information and user behavior data, and uploads all data to the cloud platform. A user portrait module creates multi-dimensional data labels based on the user data collected by the cloud platform, and uses the K-means clustering algorithm to perform cluster analysis on user labels, classifying users with similar characteristics and behavior patterns into one category to form different user group portraits. A conventional recommendation module matches product recommendations based on portrait labels, and marks products corresponding to browsing behavior data, search behavior data, and favorite / attention behavior data as unpurchased products wgs, and products corresponding to purchase behavior data as purchased products ygs. Then, conventional recommendations are made based on unpurchased products wgs through conventional recommendation strategies. An intelligent recommendation module makes secondary recommendations for purchased products ygs based on the recommendation strategy for purchased products ygs. After the recommendation, the conventional recommendation module recommends the purchased products ygs. The recommendation module continuously collects user behavior data, and recommends the collected user behavior data again according to the conventional recommendation strategy. The intelligent recommendation module recommends matching products pts according to the matching recommendation strategy, and selects suitable recommended matching products according to the formulated comprehensive recommendation strategy for original parts and non-original parts during the recommendation process. The intelligent recommendation module recommends the purchased product ygs within the shelf life according to the recommendation strategy for consumables xhs. If the user does not purchase the same product again within the shelf life, the recommendation reminder will be made in advance to let the user know that the product has been purchased. The intelligent recommendation module is based on the main product table in the recommendation database. After the docking manufacturer updates the product in the main product table, it selects the applicable tag user according to the user portrait tag for trial recommendation. The conventional recommendation module continuously collects user behavior data, and recommends the collected user behavior data again according to the conventional recommendation strategy.

[0037] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0038] The above contents are merely examples and explanations of the present invention. Those skilled in the art may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the invention or exceed the scope defined by the claims, they shall all fall within the protection scope of the present invention.

Claims

1. An e-commerce operation management system based on a cloud platform, characterized in that: include: The data collection module is used to collect user registration information and user behavior data, where the registration information includes user age, gender, occupation, and address, and the user behavior data includes browsing behavior data, search behavior data, favorite / follow behavior data, and purchase behavior data, and upload all data to the cloud platform; The user portrait module creates multi-dimensional data labels based on user data collected by the cloud platform, establishes basic labels based on user registration information, creates consumer labels of the same level for users of the same age and gender, creates preference approximation labels for users with similar behavior data, creates consumption capacity labels for users who purchase the same products in similar price ranges, and uses the K-means clustering algorithm to perform cluster analysis on user labels, classifying users with similar characteristics and behavior patterns into one category to form different user group portraits; The conventional recommendation module matches product recommendations based on the portrait tags, and marks the products corresponding to the browsing behavior data, search behavior data, and collection / follow behavior data as unpurchased products wgs, and the products corresponding to the purchase behavior data as purchased products ygs, and then makes conventional recommendations based on the unpurchased products wgs through the conventional recommendation strategy; The intelligent recommendation module establishes a secondary recommendation strategy for the purchased goods ygs and predicts the secondary recommendation time. It also marks the goods without ancillary products in the purchased goods ygs as consumables xhs, and marks the goods with ancillary products as matching products pts. The matching products pts include replacement parts, repair and maintenance tools, and maintenance and decoration products for the goods. A matching recommendation strategy is established for the matching products pts and predictive recommendations are made.

2. The e-commerce operation management system based on a cloud platform according to claim 1, characterized in that: The conventional recommendation module formulates a conventional recommendation strategy based on the user portrait, as follows: For browsing behavior data, products with a browsing time of less than 5 seconds and a browsing frequency of less than 2 times are marked as accidentally clicked products ws, products with a browsing frequency of 2-5 times and an average browsing time of 10-40 seconds are marked as initially preferred products cs, and products with a browsing frequency of more than 5 times and a browsing time of more than 40 seconds are marked as key preferred products zs; For search behavior data, the products clicked after the search are marked as the desired products ys, the products related to the search keywords are marked as recommended products ks, and the products added to the shopping cart after the search are included in the collection / follow behavior data; According to the collection / attention behavior data, these products are marked as key recommended products. The conventional recommendation module formulates a conventional recommendation plan based on the conventional recommendation strategy for recommendation.

3. The e-commerce operation management system based on a cloud platform according to claim 2, characterized in that: The recommended routine schedule is as follows: First, we collect the behavior data of accidentally clicked products ws, initially preferred products cs, key preferred products zs, intended products ys, recommended products ks, and key recommended products zds. Since these products are all unpurchased products wgs, they are recommended to users every day. The recommendation list is based on the rule list of key recommended products zds> key preferred products zs> intended products ys> initially preferred products cs> recommended products ks> accidentally clicked products ws. In the recommendation process, the total weights of the six types of goods are calculated as 1, among which the key recommended goods zds are 0.35, the key inclined goods zs are 0.20, the intended purchase goods ys are 0.15, the preliminary inclined goods cs are 0.15, the recommendable goods ks are 0.10, and the accidentally clicked goods ws are 0.

05. The recommended goods are allocated on the recommendation page according to the weight values, and all unpurchased goods wgs are recommended every day with one week as a unit. Based on the daily recommendation, the user click data is collected again, and the clicked goods are re-marked as inclined goods yqs. The products that are not clicked every day in a week are re-marked as abandoned goods dfs. In the second week, the recommendation priority of inclined goods yqs>to-be-abandoned goods dfs is used for recommendation. Among them, the inclined goods yqs are still allocated with the recommendation share according to their respective weights and recommended daily, and the to-be-abandoned goods dfs are recommended every other day according to their respective weights. This cycle repeats. The goods that users have not clicked for 1.5-3 months are marked as not recommended goods bts and are no longer recommended.

4. The e-commerce operation management system based on a cloud platform according to claim 1, characterized in that: The recommendation strategy of the intelligent recommendation module for the purchased product ygs is as follows, where the purchased product ygs includes the consumable xhs body and the matching product pts body: Collect the product information of the user's purchased products ygs, screen the user's historical purchase data, and if the same or the same product appears, it is determined that the user has purchased the product twice or more. Collect the time of each purchase of the user, analyze the user's historical purchase time interval sequence, and mark it as T1, T2, ..., Tn, where T1 is the first purchase time and Tn is the most recent purchase time, and then calculate the average purchase time interval , the calculation formula is , then the next purchase time The calculation formula is, ,in The time of the most recent purchase, based on the next purchase time The base date is moved forward by 7 days, and the resulting date is the recommended time for the purchased goods for the second purchase. If no similar or identical goods appear, it is determined that the user is purchasing the goods for the first time. For users who purchase the goods for the first time, the cycle is converted into days based on the average replacement cycle of the goods, and the days are added to the date of the user's first purchased goods to obtain the predicted second purchase date of the user. Based on the predicted second purchase date of the user, the date is moved forward by 30 days, and the resulting date is the recommended time for the purchased goods for the first purchase. The intelligent recommendation module regularly makes second purchase recommendations for the purchased goods, and after the recommendation, the conventional recommendation module continues to collect user behavior data, and recommends the collected user behavior data again according to the conventional recommendation strategy.

5. The e-commerce operation management system based on a cloud platform according to claim 1, characterized in that: The intelligent recommendation module establishes a matching recommendation strategy for matching product pts. The specific strategy is as follows: Obtain all product information on the platform, establish a supporting product recommendation database for all product information, connect the recommendation database with each product manufacturer, obtain detailed information on each product from the product manufacturer, including the main product and the corresponding original replacement parts, repair and maintenance tools, maintenance and decoration products, and establish the product table as follows: Create a main product table, including the main product ID, name, model, brand, and category fields; Create a replacement parts table and associate it with the main product table, including the parts ID, parts name, model, brand, applicable main product ID, original manufacturer identification, and replacement cycle fields; Create a maintenance tool table and associate it with the main product table, including fields such as tool ID, tool name, model, brand, applicable main product ID, and recommended usage period; Create a maintenance and decoration product table and associate it with the main product table, including product ID, product name, model, brand, applicable main product ID, and recommended matching cycle fields; After obtaining the information of the matching items pts in the user's purchased product ygs, the information includes product ID, name, model, brand, and category. The matching items pts are matched with the main product table in the recommendation database to obtain the replacement parts, repair and maintenance tools, and maintenance and decoration products related to the matching items pts. The recommended use period of the related replacement parts, repair and maintenance tools, and maintenance and decoration products is converted into days, and the days are added to the purchase date of the main product. The demand date of the replacement parts, repair and maintenance tools, and maintenance and decoration products is calculated according to the current calendar. On the basis of this date, 14 days are moved forward. The obtained date is the recommended date of the replacement parts, repair and maintenance tools, and maintenance and decoration products related to the matching items pts in the purchased product ygs. The intelligent recommendation module makes recommendations after expiration. When recommending, if there are more than two matching products for the matching items pts, the priority order of replacement parts>repair and maintenance tools>maintenance and decoration products is used for recommendation. After the recommendation, the conventional recommendation module continues to collect user behavior data, and the collected user behavior data is recommended again according to the conventional recommendation strategy.

6. The e-commerce operation management system based on a cloud platform according to claim 5, characterized in that: The intelligent recommendation module formulates a comprehensive recommendation strategy for original and non-original parts based on the principle of recommendation in the order of priority of replacement parts > repair and maintenance tools > maintenance and decoration products, as follows: Obtain the main product information of the matching products pts, and collect all users who have purchased the product within 12 months before the current time on the platform, track the proportion of original and non-original parts purchased by these users, as well as the brand, model, price k, and evaluation of the non-original parts purchased, and divide the priority of original and non-original parts according to the obtained original and non-original parts ratio. If the original parts ratio is greater than the non-original parts ratio, increase the recommended share of original parts, otherwise increase the share of non-original parts. Based on this principle, eliminate matching products with a non-original parts negative review rate exceeding 40%.

7. The e-commerce operation management system based on a cloud platform according to claim 6, characterized in that: Based on the remaining 60% of non-original parts, if the supporting products of the matching product pts have original parts, the price of the original parts is marked as β, and then the price of the original parts price β is calculated by 30% up and 20% down respectively. The calculation method is as follows: 30% price increase = , price reduction of 20% = , get the price range, and then arrange the remaining 60% of non-original parts prices k in descending order, select non-original parts with price k between 30% increase and 20% decrease of original parts price, and recommend the non-original parts in this range according to the priority of price in descending order, that is, the higher the price, the higher the recommended share, and vice versa. After the recommendation, the conventional recommendation module continues to collect user behavior data, and the collected user behavior data is recommended again according to the conventional recommendation strategy.

8. The e-commerce operation management system based on a cloud platform according to claim 7, characterized in that: Based on the remaining 60% of non-original parts, if the supporting products of the matching product pts do not have original parts, the price of the matching product pts is marked as P 主商品价格 Then get the minimum price range of the same category of the main product and mark it as P 最低价格 , the maximum value is marked as P 最高价格 , calculate the relative position ratio of the price of the main product purchased by the user within the price range of products of the same category, the calculation formula is: , P 倾向值 Between 0 and 1, the closer it is to 0, the more the user prefers low-priced products, and the closer it is to 1, the more the user prefers high-priced products. 倾向值 To calculate the price preference range of users for supporting products, the same principle of 30% increase and 20% decrease is used. The calculation formula is as follows: ； ； 60% of the non-original parts price K gets P 配套品下限 and P 配套品上限 Price range, non-original parts within this range are recommended in descending order of price priority. After the recommendation, the conventional recommendation module continues to collect user behavior data and recommends the collected user behavior data again according to the conventional recommendation strategy.

9. The e-commerce operation management system based on a cloud platform according to claim 1, characterized in that: The intelligent recommendation module formulates a recommendation strategy within the shelf life for consumables xhs without ancillary products in the purchased goods ygs, as follows: Collect the information of consumables xhs, obtain its shelf life, and obtain the expiration time of consumables xhs through the calculation formula of factory date + shelf life = expiration time. Since the use cycle of consumables xhs is short, no user feedback can be obtained after the prediction recommendation, and the conventional recommendation module regards it as a non-recommended product bts. The intelligent recommendation module moves forward 30 days based on the expiration time of consumables xhs as the reminder recommendation time, and determines in real time through user purchase behavior data whether the user has purchased the consumables xhs twice within the shelf life. When the judgment result shows that there is a second purchase, the reminder recommendation time is cancelled, and the expiration time of the most recent purchase of consumables xhs is recalculated to set the reminder recommendation time. When the judgment result shows that there is no second purchase, after the reminder recommendation time arrives, the intelligent recommendation module triggers the recommendation mechanism for recommendation, and sets the reminder words of the expiration date approaching on the recommendation page. After the recommendation, the conventional recommendation module continues to collect user behavior data, and the collected user behavior data is recommended again according to the conventional recommendation strategy.

10. The e-commerce operation management system based on a cloud platform according to claim 5, characterized in that: The intelligent recommendation module is based on the main product table in the recommendation database. When the manufacturer updates the products in the main product table, the intelligent recommendation module first determines the applicable grade, occupation and function of the product based on the main product information, and then determines the user tags for the updated product to be recommended. The priority adopts the strategy of occupation applicability > interest and hobby applicability > age applicability for trial recommendation. The trial recommendation method is daily recommendation for occupation applicability, every other day recommendation for interest and hobby applicability, and weekly recommendation for age applicability. According to this recommendation method, the regular recommendation module continuously collects user behavior data, and the collected user behavior data is recommended again according to the regular recommendation strategy.

Images