An intelligent information recommendation system for e-commerce sales

Abstract

The application discloses an e-commerce sales intelligent information recommendation system, and belongs to the technical field of e-commerce sales information recommendation, which comprises a user module, a retrieval module and a recommendation module; the user module is used for inputting demand information by a user and sending the recognized demand information to the retrieval module; the retrieval module is used for retrieving sales information according to the demand information, setting a target channel, and inputting the obtained sales information into each target channel for retrieval to obtain a large amount of sales information; the obtained sales information is evaluated to obtain a corresponding recommendation value, each sales information is arranged in a descending order of the recommendation value to obtain a first sequence; and corresponding quality characteristic data is supplemented; the recommendation module is used for recommending the sales information to the user, obtaining the corresponding first sequence, and recommending the sales information to the user according to the first sequence.

Description

[0001] This application is a divisional application of application filed on November 4, 2024, with application number 2024115566735 and invention title "An Intelligent Information Recommendation System for Online Sales". Technical Field

[0002] This invention belongs to the field of sales information recommendation technology, specifically an intelligent e-commerce sales information recommendation system. Background Technology

[0003] To meet consumers' growing shopping and personalization needs, and with the rapid development of the internet and e-commerce, people can easily shop online. However, the vast amount of goods and information also brings users problems of choice difficulties and information overload.

[0004] Furthermore, existing sales information recommendation systems are not ideal in practical applications, especially as more and more consumers are tired of push notifications and simply delete or block push notifications from various shopping platforms. This results in poor performance of existing sales information recommendation systems and also constitutes harassment for consumers. The lack of reliable ways to verify sales information leaves most consumers unable to understand detailed information about products, raising concerns about product quality and potential scams. Therefore, to address these issues and achieve consumer-oriented sales information recommendations, this invention provides an intelligent e-commerce sales information recommendation system. Summary of the Invention

[0005] To address the problems of the aforementioned solutions, this invention provides an intelligent information recommendation system for e-commerce sales.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] An intelligent information recommendation system for e-commerce sales includes a user module, a search module, and a recommendation module;

[0008] The user module is used for users to input their needs and to send the identified needs to the retrieval module.

[0009] The retrieval module is used to retrieve sales information based on demand information, set target channels (which refer to various selected e-commerce platforms), input the obtained demand information into each target channel for retrieval, obtain a large amount of sales information, evaluate the obtained sales information to obtain corresponding recommendation values, arrange the sales information in descending order of recommendation values ​​to obtain the first sequence, and supplement the corresponding quality characteristic data.

[0010] Furthermore, methods for setting target channels include:

[0011] Identify the existing sales channels and calculate the average recommendation value for each product category in each sales channel;

[0012] Obtain the market share for each sales channel, and match the corresponding correction coefficient based on the market share.

[0013] The obtained correction coefficient and average recommendation value are labeled as ε and PTZ, respectively. The corresponding screening value SP is calculated according to the screening formula SP=ε×PTZ. Product categories with screening values ​​lower than the threshold X2 are removed from the sales channel.

[0014] Target channels are set based on the processing results of each product category, and corresponding target category tags are added.

[0015] Furthermore, the average recommendation value is calculated using a sampling method.

[0016] Furthermore, the recommendation value is calculated based on the product quality and price corresponding to the sales information.

[0017] Furthermore, the methods for evaluating recommendation values ​​include:

[0018] Identify the quality data and price data corresponding to each sales information, evaluate the obtained quality data, obtain the corresponding quality value, and mark the obtained quality value as PZ; set the corresponding discounted value and mark value based on the price data, and mark the obtained discounted value and mark value as ZX and BZ respectively;

[0019] The recommended value is calculated using the formula TZ=PZ / (BZ-ZX).

[0020] Furthermore, the recommendation values ​​for each sales information are optimized based on user usage records. The number of times users select each target channel is counted in real time, the usage ratio of each target channel is calculated, and the usage ratios are integrated into optimization features. The optimization features are analyzed to obtain the optimization coefficient corresponding to each target channel, which is marked as c.

[0021] Identify the target channels corresponding to each sales information and match the corresponding optimization coefficients. Then, the formula for calculating the recommended value is optimized as follows: TZ=c×PZ / (BZ-ZX).

[0022] Furthermore, methods for evaluating quality data include:

[0023] Identify each quality item in the quality data and label it as i, i = 1, 2, ..., n, where n is a positive integer; evaluate each quality item data to obtain the corresponding single quality value, label the obtained single quality value as PZi, and calculate the corresponding quality value PZ according to the quality value evaluation formula PZ = (∑PZi) / n.

[0024] Furthermore, methods for analyzing price data include:

[0025] Set up a discount channel, input sales information into the discount channel, obtain the corresponding discount data, and obtain the corresponding discount value based on the discount data; identify the marker value in the sales information.

[0026] Furthermore, methods for obtaining quality characteristic data include:

[0027] Obtain quality item data where the individual quality value is lower than the threshold X1, extract the cause data from the quality item data (the cause data is the data that makes the individual quality value lower than the threshold X1), obtain the impact data corresponding to the cause data, and combine and integrate the cause data and impact data into quality feature data.

[0028] The recommendation module is used to recommend sales information to users, obtain the corresponding first sequence, and recommend sales information to users according to the first sequence.

[0029] Furthermore, before recommending sales information to users based on the first sequence, the recommendation module identifies large items in the first sequence, obtains the user's location, obtains the location information of the large items for sale based on the user's location, and supplements the first sequence with the obtained location information of the large items.

[0030] Compared with the prior art, the beneficial effects of the present invention are:

[0031] By coordinating the user module, search module, and recommendation module, intelligent recommendations of sales information are made for users, breaking down platform barriers and obtaining suitable sales information based on user needs. Furthermore, relevant quality characteristic data is supplemented into the recommended sales information, allowing users to intuitively understand the quality and other information of each product. This addresses the concerns of most consumers who lack detailed information about the products being sold, such as concerns about product quality and being scammed. Attached Figure Description

[0032] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0033] Figure 1 This is a block diagram illustrating the principle of the present invention;

[0034] Figure 2 This is a schematic diagram illustrating the working principle of the retrieval module of this invention. Detailed Implementation

[0035] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] like Figures 1 to 2 As shown, an intelligent information recommendation system for e-commerce sales includes a user module, a search module, and a recommendation module.

[0037] The user module is used for users to input their needs, including product names, price ranges, etc.; the identified needs are sent to the retrieval module; and the user module has user information management functions, allowing users to manage their personal information.

[0038] The retrieval module is used to retrieve sales information based on demand information, and the specific methods include:

[0039] Set target channels, which refer to various selected e-commerce platforms; input the obtained demand information into each target channel for retrieval to obtain a large amount of sales information, including corresponding source information; evaluate the obtained sales information to obtain corresponding recommendation values, arrange the sales information in descending order of recommendation value to obtain the first sequence; and supplement the corresponding quality characteristic data.

[0040] Methods for setting target channels include:

[0041] Obtain all available sales channels in the market, as long as they are searchable, legitimate, and secure. Evaluate the recommendation value (RWV) for each product; the specific calculation method for RWV is detailed in the evaluation section below. Calculate the average RWV for the product category. If there are too many products, sampling can be used to calculate the average RWV. Identify the market share of the sales channel and match a corresponding correction coefficient based on the market share. This correction coefficient is used to adjust the sales channel with the largest market share. The correction coefficient should be greater than or equal to 1. This is done manually, setting a correction coefficient for each market share. For shares that are too low or too high, no individual setting is needed. Correction coefficients below a certain share are all set to 1; those above a certain share directly indicate compliance and can be considered infinitely large. Only the share within the interval is marked with the corresponding coefficient. The correction coefficient is set one by one with a share unit of 1. The subsequent calculation is performed using the interpolation method. For example, if the share is 3.8%, the calculation is performed using the interpolation method based on the correction coefficients corresponding to 3% and 4%. The number of correction coefficients is reduced to prevent sales channels with a certain share from being directly eliminated. The obtained correction coefficients and average recommended values ​​are marked as ε and PTZ, respectively. The corresponding screening value is calculated according to the screening formula SP=ε×PTZ. Product categories with screening values ​​lower than the threshold X2 are eliminated from the sales channel until all product categories of the sales channel are evaluated. Sales channels whose product categories have not been completely eliminated are marked as target channels and labeled with the corresponding target category label. The category label refers to the product categories that have not been eliminated.

[0042] In other embodiments, the target channel can also be specified manually.

[0043] Methods for evaluating the obtained sales information include:

[0044] The system identifies the quality and price data corresponding to each sales message. Quality data refers to information related to product quality, such as ingredient lists, materials, manufacturers, and reviews. Price data includes information related to price, discount coupons, and cashback. For example, it evaluates the channels for cashback and discount coupons for various products, which are preset by the platform. For instance, various chat groups offering cashback and discounts are uniformly marked as cashback channels. After entering the corresponding sales link into the cashback channel, the system obtains the relevant price data, including discount coupon and cashback information. In other words, the cashback channels are set by the platform itself.

[0045] Based on the price data, the corresponding marked value and discounted value are evaluated. The marked value refers to the marked price in the sales information, and the discounted value refers to the maximum combined price that can be obtained through various discounting channels, such as the maximum discount, cashback, etc. The total price of discount plus cashback is the discounted value. Depending on whether various discounting channels can be used simultaneously, the corresponding discounted value can be obtained using existing identification and calculation methods. The obtained discounted value and marked value are marked as ZX and BZ, respectively.

[0046] The obtained quality data is evaluated to obtain the corresponding quality value, which is marked as PZ; the corresponding recommended value is calculated according to the recommended value formula TZ=PZ / (BZ-ZX).

[0047] Methods for evaluating the obtained quality data include:

[0048] The quality data for each corresponding product category has fixed quality items. For example, food products generally include data such as ingredient list, manufacturer, and reviews. The specific quality items that should be included in the quality data for each product category are set by an expert group based on the relevant aspects of product quality in the corresponding product category. For example, manufacturer information includes the manufacturer's registration information, place of origin, and other production quality-related data that can be obtained based on current big data. When setting the quality items for each product category, an initial value is preset. That is, when the data for a quality item cannot be obtained, the initial value will be used as the data for that quality item and as the individual quality value for that quality item in subsequent searches. This is because in the actual retrieval process, some information may not be available due to the product. The initial values ​​for different categories may not be the same because the weight of the same quality item in the quality assessment of the product varies in different categories. Therefore, when setting each quality item, the corresponding initial value is set simultaneously. The initial value is a percentage, that is, the value range is [0, 100].

[0049] Identify each quality item in the quality data and label it as i, i = 1, 2, ..., n, where n is a positive integer; evaluate each quality item data to obtain the corresponding single quality value, label the obtained single quality value as PZi, and calculate the corresponding quality value PZ according to the quality value evaluation formula PZ = (∑PZi) / n, i = 1, 2, ..., n.

[0050] The evaluation of each quality item involves establishing corresponding evaluation criteria based on relevant standard data. For example, evaluation criteria can be set based on the latest version of food standards and data on the harmful effects of related additives on the human body. Taking the ingredient list as an example, first assess whether it complies with food standards. If it does not, a corresponding single quality value is assigned. If it does, a subtraction is performed from a maximum score of 100 to identify whether harmful elements are present. The corresponding single quality value is obtained by subtracting from the value based on the content, type, and quantity of these elements. Evaluation criteria can be set using current research data, as each quality item in various industries has its own set of quality assessment standards. The optimal quality has a corresponding consensus, therefore, it can be organized into corresponding evaluation criteria, and the corresponding individual quality values ​​can be evaluated in combination with the corresponding evaluation criteria; existing methods can be applied to evaluate each quality item; for example, a corresponding quality evaluation model can be built based on CNN or DNN networks, and a corresponding training set can be built manually for training. The training set includes various simulated quality data of the product and the corresponding individual quality values ​​of each quality item; the quality evaluation model is then used for evaluation to obtain the individual quality values ​​corresponding to each quality item; since neural networks are existing technology in this field, the specific building and training process will not be described in detail in this invention.

[0051] Supplementing the corresponding quality feature data involves extracting the reasons why each individual quality value is lower than the threshold X1 from the quality item data, obtaining its corresponding impact data, and integrating it into quality feature data. This data helps users directly understand the defects of the product, such as a harmful additive in food, the additive content, and its corresponding impact.

[0052] The recommendation module is used to recommend sales information to users, obtain the corresponding first sequence, and recommend sales information to users according to the first sequence.

[0053] The recommended information includes relevant discounting information, such as discounting channels, and may also include corresponding tutorials to help users save costs.

[0054] By coordinating the user module, search module, and recommendation module, intelligent recommendations of sales information are made for users, breaking down platform barriers and obtaining suitable sales information based on user needs. Furthermore, relevant quality characteristic data is supplemented into the recommended sales information, allowing users to intuitively understand the quality and other information of each product. This addresses the concerns of most consumers who lack detailed information about the products being sold, such as concerns about product quality and being scammed.

[0055] In one embodiment, it is often difficult to understand the actual offline condition of an item through images during online shopping. For large items, the inconvenience of returns and exchanges may prevent some people from purchasing online. Therefore, to solve this problem, the following method is proposed:

[0056] Identify large items in the first sequence and evaluate them based on both volume and price. Items exceeding a certain price or volume are directly identified as large items, while others can be pre-calculated with corresponding weights. Alternatively, other existing technologies can be used to set corresponding evaluation standards for large items.

[0057] Obtain the user's location, and identify the location information of large items for sale around the user, such as the location of each store selling the large item; supplement the obtained large item location information into the first sequence.

[0058] In one embodiment, as users use the service more frequently, their preference for one or more target channels can be understood. Therefore, the first sequence can be optimized based on user usage records. The specific method is as follows:

[0059] The system continuously monitors the number of times users select each target channel, calculates the usage percentage of each target channel, and integrates these usage percentages into optimization features. For example, if the usage percentages are 0.5, 0.2, 0.1, 0.1, 0.05, 0.05, 0, 0, 0, then the optimization feature is (0.5, 0.2, 0.1, 0.1, 0.05, 0.05, 0, 0, 0). A corresponding optimization analysis model is built based on a CNN or DNN network. A training set is manually created for training, including optimization features for various simulation settings and corresponding optimization coefficients for each target channel. The successfully trained optimization analysis model is then analyzed to obtain the optimization coefficients for each target channel, denoted as c.

[0060] Identify the target channels corresponding to each sales information and match the corresponding optimization coefficients. Then, the formula for calculating the recommended value is optimized as follows: TZ=c×PZ / (BZ-ZX).

[0061] The above formulas are all numerical calculations after removing dimensions. The formulas are obtained by software simulation based on a large amount of data and are closest to the real situation. The preset parameters and preset thresholds in the formulas are set by those skilled in the art according to the actual situation or obtained by simulation based on a large amount of data.

[0062] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. An intelligent information recommendation system for e-commerce sales, characterized in that, The user module, the retrieval module and the recommendation module are included. The user module is used for inputting demand information and sending the identified demand information to the retrieval module. The retrieval module is used for performing sales information retrieval according to the demand information, setting a target channel, the target channel refers to various selected e-commerce platforms, inputting the obtained demand information into each target channel for retrieval, obtaining a large amount of sales information, evaluating the obtained sales information, obtaining a corresponding recommendation value, arranging each sales information in a descending order of the recommendation value to obtain a first sequence, and supplementing corresponding quality characteristic data. The recommendation module is used for recommending the sales information to the user, obtaining the corresponding first sequence, and recommending the sales information to the user according to the first sequence. The method for setting the target channel comprises: identifying the sales channels, calculating the average recommendation value of each commodity classification of the sales channels; obtaining the market share corresponding to each sales channel, and matching a corresponding correction coefficient based on the market share; marking the obtained correction coefficient and average recommendation value as ε and PTZ respectively, calculating a corresponding screening value SP according to a screening formula SP = ε × PTZ, and removing the commodity classification with a screening value lower than a threshold X2 from the sales channel; setting the target channel based on the processing result of each commodity classification and marking the corresponding target classification label; The recommendation module identifies large-size goods in the first sequence before recommending the sales information to the user according to the first sequence, obtains the user location, obtains large-size location information of the large-size goods for sale according to the user location, and supplements the obtained large-size location information to the first sequence. The method for evaluating the recommendation value comprises: identifying the quality data and price data corresponding to each sales information, evaluating the obtained quality data to obtain a corresponding quality value, marking the obtained quality value as PZ, setting a corresponding discount value and marking value based on the price data, and marking the obtained discount value and marking value as ZX and BZ respectively; calculating a corresponding recommendation value according to a recommendation value formula TZ = PZ / (BZ - ZX); optimizing the recommendation value of each sales information according to the user usage record, statistically analyzing the number of times that the user selects each target channel, calculating the usage proportion of each target channel, integrating each usage proportion into an optimization feature, analyzing the optimization feature, obtaining an optimization coefficient corresponding to each target channel, and marking the optimization coefficient as c; identifying the target channel corresponding to each sales information, matching the corresponding optimization coefficient, and optimizing the calculation formula of the recommendation value to TZ = c × PZ / (BZ - ZX). 2.The intelligent information recommendation system for e-commerce sales according to claim 1, wherein, The average recommendation value is calculated by sampling. 3.The intelligent information recommendation system for e-commerce sales according to claim 2, wherein, The recommendation value is calculated according to the quality and price of the goods corresponding to the sales information. 4.The intelligent information recommendation system for e-commerce sales according to claim 1, wherein, The method for evaluating the quality data comprises: identifying each quality item in the quality data, marking the quality item as i, i = 1, 2, …, n, n is a positive integer, evaluating each quality item data to obtain a corresponding single quality value, marking the obtained single quality value as PZi, and calculating a corresponding quality value PZ according to a quality value evaluation formula PZ = (∑PZi) / n. 5.The intelligent information recommendation system for e-commerce sales according to claim 1, wherein, The method for analyzing the price data comprises: A discount channel is set, sales information is input into the discount channel, corresponding discount data is obtained, and corresponding discount values are obtained according to the discount data; and a mark value in the sales information is identified. 6.The intelligent information recommendation system for e-commerce sales according to claim 5, wherein, The quality characteristic data acquisition method comprises: Quality item data with a single quality value lower than a threshold X1 is acquired, reason data in the quality item data is extracted, the reason data being data that causes the single quality value to be lower than the threshold X1, influence data corresponding to the reason data is acquired, and each reason data and influence data are combined and integrated as quality characteristic data.

Images