A smart e-commerce business data management system

Abstract

The present invention provides a smart e-commerce business data middle-end management system, including a data collection module, a data analysis module and a data display module; wherein the data collection module is used to collect sales data of various commodities sold by merchant users in different sales platforms, wherein the sales platforms include online platforms and offline platforms; the sales data include browsing data and transaction data; the data analysis module is used to analyze the heat information of the current commodity according to the acquired sales data of each commodity, and obtain the commodity heat information analysis result; the data display module is used to integrate and generate a real-time heat report according to the commodity heat information analysis result of various commodities sold by merchant users, and push the initial heat report to the merchant user terminal. The present invention helps to improve the comprehensiveness and real-time nature of commodity sales data collection, and improve the overall management effect of merchant users on commodity sales.

Description

Technical Field

[0001] The present invention relates to the technical field of data middle platforms, and particularly to a management system for a smart e-commerce business data middle platform. Background Art

[0002] Currently, with the development of e-commerce business, more and more enterprises and merchants choose to sell goods through an online combined with offline sales method. When conducting online sales, goods are usually listed on different online sales platforms, and the sales of goods are achieved through different channels and platforms.

[0003] In the prior art, when merchants list and sell goods on different sales channels (online or offline) or platforms (such as different online sales platforms), they usually adopt independent operation strategies. Different sales channels and platforms adopt different sales methods, and the sales data obtained from different channels or platforms cannot be interconnected. The management difficulty of merchants for product sales data increases, resulting in a slow sales feedback speed for merchants (for example, a product is popular offline but cannot be promptly promoted online), which is not conducive to the effective management of product sales by merchants. Therefore, it is extremely necessary to design a management system for a smart e-commerce business data middle platform to improve the intelligence and real-time level of merchants' management of product sales business data. Summary of the Invention

[0004] Aiming at the technical problem that in the traditional mode, the sales data obtained from different channels or platforms cannot be interconnected and the management difficulty of merchants for product sales data increases, the present invention aims to provide a management system for a smart e-commerce business data middle platform.

[0005] The object of the present invention is achieved by the following technical solutions:

[0006] The present invention provides a management system for a smart e-commerce business data middle platform, including a data collection module, a data analysis module, and a data display module; wherein,

[0007] The data collection module is used to collect the sales data of each product sold by merchant users on different sales platforms, where the sales platforms include online platforms and offline platforms; the sales data includes browsing data, transaction data, etc.;

[0008] The data analysis module is used to analyze the popularity information of the current product based on the obtained sales data of each product to obtain the analysis result of the product popularity information;

[0009] The data display module is used to integrate and generate a real-time popularity report based on the analysis result of the product popularity information of various products sold by merchant users, and push the initial popularity report to the merchant user terminal.

[0010] Preferably, the system further includes a data storage module;

[0011] The data storage module is used to analyze the obtained sales data and product popularity information analysis results, associate and store them with the corresponding products, build a product data file, classify and store and manage the product data file, and build a product database.

[0012] Preferably, the system further includes an intelligent shelving module;

[0013] The intelligent shelving module is used to perform intelligent product shelving or push according to the obtained implementation heat report and product popularity information.

[0014] Preferably, the data acquisition module includes an online data acquisition unit;

[0015] The online data acquisition unit is used to connect to the data interfaces of each online platform respectively, obtain the sales data of the specified products provided by the online platform, and associate the obtained sales data with the corresponding products; where the online platforms include merchant online sales platforms, third-party e-commerce platforms, etc.

[0016] Preferably, the data acquisition module includes an offline data acquisition unit;

[0017] The offline data acquisition unit is used to connect to the on-site terminal set on the offline platform, obtain the browsing data and transaction data of the specified products collected by the on-site terminal, and associate the obtained browsing data and transaction data with the corresponding products; where the offline platforms include merchant offline stores, offline shopping malls, etc., and the browsing data includes the pedestrian flow data and customer stay time data in front of the product display stand.

[0018] Preferably, the on-site terminal includes an image acquisition device, and the pedestrian flow image data in front of the product display stand is obtained through the image acquisition device;

[0019] The offline acquisition unit further includes an image analysis unit and an association unit;

[0020] The image analysis unit is used to identify and track the customers in the image based on the Yolov5 image analysis model according to the obtained pedestrian flow image data, and obtain the pedestrian flow data and customer stay time data in front of the corresponding product display stand;

[0021] The association unit is used to associate the obtained pedestrian flow data and customer stay time data with the corresponding products.

[0022] Preferably, the data acquisition module further includes a data security management unit;

[0023] The data security management unit is used to perform security verification on the obtained sales data of the products, and obtain the original product sales data after the security verification passes.

[0024] Preferably, the data analysis module includes a statistics unit and a popularity analysis unit;

[0025] The statistics unit is used to classify and statistically analyze the sales data of a specified product on each sales platform to obtain the statistical information of the product's sales data;

[0026] The popularity analysis unit is used to intelligently analyze the popularity information of a product based on the product's sales data to obtain the analysis result of the product's popularity information.

[0027] Preferably, in the statistics unit, the classification and statistical analysis of the sales data of a product on each sales platform specifically include:

[0028] Obtaining the real-time access data of the product:

[0029] TAcc i (t) = ω at-1 ×pAc 1,i (t) + ω at-2 ×pAc 2,i (t) + … + ω at-K ×pAc K,i (t)

[0030] Among them, TAcc i (t) represents the real-time access data of product i at time t; ω at-k represents the preset access volume weighting adjustment factor of sales platform k, pAc k,i (t) represents the access data of product i at time t in sales platform k, where k = 1, 2, … K, and K represents the total number of sales platforms; when sales platform k is an online platform, the access data of product i at time t represents the customer access volume of the sales page of product i within a period of time based on time t; when sales platform k is an offline platform, the access data of product i at time t represents the pedestrian flow data in front of the display stand corresponding to product i within a period of time based on time t;

[0031] Obtaining the real-time transaction data of the product;

[0032] TDea i (t) = pDe 1,i (t) + pDe 2,i (t) + … + pDe K,i (t)

[0033] Among them, TDea i (t) represents the real-time transaction data of product i at time t; pDe k,i(t) represents the number of completed orders for product i in sales platform k within a period of time based on time t, where k = 1, 2, ... K, and K represents the total number of sales platforms.

[0034] Preferably, in the heat analysis unit, intelligent analysis is performed on the heat information of the commodity according to the sales data of the commodity, including:

[0035] According to the current period of time t=t 0 ,t 1 …t N 、Access data pAc of product i on various sales platforms 1,i (t),pAc 2,i (t),…pAc K,i (t) and real-time access to data TAcc i (t) The feature vector set thre that makes up product i i (t); where t 0 represents the current time, t n The larger the subscript n in , the longer the time from the current moment is;

[0036] The obtained feature vector set is input into the trained heat analysis model, and the heat analysis model extracts features and predicts transaction conversion based on the feature vector set to obtain the current time t 0 Product heat analysis results of product i; the heat analysis model is built based on CNN neural network, and in the model training process, the historical feature vector set thre i (t B ) and the corresponding transaction data set dea i (t B-1 ) training, where the transaction data set consists of a period of time t=t B-1 ,t B …t B+N-1 、Product i's transaction order volume data pDe on each sales platform 1,i (t),pDe 2,i (t),…,pDe K,i (t) and real-time transaction data TDea i (t); the heat analysis result of the product output by the heat analysis model is dea i (t -1 ), t -1 indicates the future time, corresponding to the predicted result.

[0037] The beneficial effects of the present invention are as follows: The data collection module collects the sales data of the products sold by merchant users on different sales platforms. The sales data of the specified products can be obtained through online data interfaces or offline on-site terminals. The sales data is summarized and classified and statistically analyzed through the data middle platform, which helps to improve the comprehensiveness and real-time nature of the collection of product sales data. Further, based on the data analysis module, the popularity of products is analyzed according to the sales data of products on different channels or platforms, which helps to estimate the popularity of products based on the sales data and realize the intelligent analysis of product popularity. The data display module intuitively displays the popularity information of products, which helps the merchant management personnel to use the real-time popularity information of products as the basis for decision-making and timely adjust the sales strategies of products on different channels or platforms, improving the overall management effect of merchant users on product sales. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The present invention will be further described with reference to the accompanying drawings. However, the embodiments in the drawings do not constitute any limitation to the present invention. For those of ordinary skill in the art, other drawings can be obtained according to the following drawings without creative efforts.

[0039] Figure 1 It is a framework structure diagram of a smart e-commerce business data middle platform management system in an exemplary embodiment of the present invention;

[0040] Figure 2 It is a framework structure diagram of a smart e-commerce business data middle platform management system in another exemplary embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0041] The present invention will be further described in combination with the following application scenarios.

[0042] Refer to Figure 1 , which shows a smart e-commerce business data middle platform management system, including a data collection module, a data analysis module, and a data display module; wherein,

[0043] The data collection module is used to collect the sales data of each product sold by merchant users on different sales platforms, where the sales platforms include online platforms and offline platforms; the sales data includes browsing data, transaction data, etc.;

[0044] The data analysis module is used to analyze the popularity information of the current product according to the obtained sales data of each product, and obtain the analysis result of the product popularity information;

[0045] The data display module is used to integrate and generate a real-time popularity report according to the analysis result of the product popularity information of various products sold by merchant users, and push the initial popularity report to the merchant user terminal.

[0046] In the above embodiments of the present invention, the data collection module collects the sales data of the products sold by merchant users on different sales platforms. The sales data of specified products can be obtained through online data interfaces or offline on-site terminals, etc. The sales data is summarized and classified and statistically analyzed through the data middle platform, which helps to improve the comprehensiveness and real-time nature of the collection of product sales data. Further, based on the data analysis module, the popularity analysis of products is carried out according to the sales data of products in different channels or different platforms, which helps to estimate the popularity of products based on the sales data and realize the intelligent analysis of product popularity. The intuitive display of the popularity information of products through the data display module helps the merchant management personnel to use the real-time popularity information of products as the decision-making basis to timely adjust the sales strategies of products in different channels or different platforms, and improve the overall management effect of merchant users on product sales.

[0047] Among them, in an exemplary scenario, the management system can be built for merchant users. Through the built management system, merchants can uniformly manage the products they sell on different sales platforms, and realize the collection and unified management of product e-commerce sales data through the data middle platform.

[0048] Preferably, the system further includes a data storage module;

[0049] The data storage module is used to associate and store with the corresponding products according to the obtained sales data and the analysis results of product popularity information, build a product data file, and classify and store and manage the product data file to build a product database.

[0050] Among them, a special database module is also set in the management system to uniformly store and manage the historical sales data of products, and further build a product data file according to the sales data of products, which helps to further improve the management level of product sales data.

[0051] Preferably, referring to Figure 2 , the system further includes an intelligent shelf placement module;

[0052] The intelligent shelf placement module is used to perform intelligent product shelf placement or push according to the obtained implementation heat report and the product popularity information.

[0053] In an exemplary scenario, the management system can also be built based on the merchant user's own online mall or based on an authorized third-party online sales platform. Based on the obtained analysis results of product popularity information, the online mall or the online sales platform can perform targeted shelf placement or hot promotion on specified products (such as the products with the highest popularity), so as to adjust the key sales strategies of products according to the real-time product popularity information, which helps to improve the product sales effect.

[0054] Preferably, the data acquisition module includes an online data acquisition unit;

[0055] The online data acquisition unit is used to connect to the data interfaces of each online platform respectively, obtain the sales data of the specified products provided by the online platforms, and associate the obtained sales data with the corresponding products; where the online platforms include merchant online sales platforms, third-party e-commerce platforms, etc.

[0056] Among them, when collecting the sales data of products for an online platform, it can connect to the data interface of the corresponding online platform, and after completing the data call authorization, obtain the corresponding product sales data from the platform background data;

[0057] In an exemplary scenario, for the sales platform that lists its own products, it can directly obtain the sales data of each product from the background data, while for a third-party e-commerce platform, it is necessary to first obtain authorization from the third-party e-commerce platform before being able to obtain the corresponding product sales data.

[0058] Preferably, the data acquisition module includes an offline data acquisition unit;

[0059] The offline data acquisition unit is used to connect to the on-site terminals set up on the offline platforms, obtain the browsing data and transaction data of the specified products collected by the on-site terminals, and associate the obtained browsing data and transaction data with the corresponding products; where the offline platforms include merchant offline stores, offline shopping malls, etc., and the browsing data includes the pedestrian flow data in front of the product display stands and the customer stay time data.

[0060] Among them, the present invention also particularly collects the sales data of products in offline physical stores. Through on-site terminals, it can collect browsing data such as the pedestrian flow and stay time corresponding to the products, which helps to subsequently combine the offline sales data and the online sales data to feedback the popularity information of the products, improving the comprehensiveness of obtaining product sales data.

[0061] Preferably, the on-site terminal includes an image acquisition device, and the pedestrian flow image data in front of the product display stand is obtained through the image acquisition device;

[0062] The offline acquisition unit further includes an image analysis unit and an association unit;

[0063] The image analysis unit is used to identify and track the customers in the image based on the Yolov5 image analysis model according to the obtained pedestrian flow image data, and obtain the pedestrian flow data and customer stay time data in front of the corresponding product display stand;

[0064] The association unit is used to associate the obtained pedestrian flow data and customer stay time data with the corresponding products.

[0065] Among them, the acquisition methods of the passenger flow data and the customer stay time data can be realized by means of image analysis. By collecting the image data of the offline product display area in real time and identifying and tracking the customers in the area based on the obtained image data, the corresponding passenger flow data and customer stay time data can be extracted, so as to realize the collection of offline sales data.

[0066] In an exemplary scenario, the image analysis model based on Yolov5 can adopt the pre-trained Yolov5 model in the prior art to realize the customer identification and tracking based on the image data, and then obtain the corresponding passenger flow data and customer stay time data. The present invention does not make specific limitations here.

[0067] In an exemplary scenario, the image acquisition device is set on the product display stand, and its shooting angle is to shoot the aisle area in front of the display stand obliquely from top to bottom at a certain angle, so as to obtain the passenger flow image data in front of the corresponding product display stand.

[0068] In an exemplary scenario, according to the shooting angle and the shooting picture of the image acquisition device, the passenger flow image is pre-marked, and the area in front of the corresponding product display stand in the passenger flow image is marked as the area of interest. The image analysis unit further identifies and tracks the people entering the picture based on the YoloV5 image processing model according to the obtained passenger flow image data. When a person enters the picture from the edge of the picture, track the moving trajectory of the person. When the person enters the marked area of interest, record that the passenger flow in front of the product display stand is increased by 1, and further track the trajectory of the person in the area of interest, and count the continuous stay time of the person in the area of interest as the customer stay time data corresponding to the person. When the person leaves the area of interest and the edge of the picture, stop tracking the person. Based on the obtained passenger flow image data, the intelligent identification of the passenger flow data in front of the product display stand and the stay time of the customer helps to improve the intelligent level and reliability of obtaining the browsing data when the product is displayed in the offline store.

[0069] Among them, when considering personnel identification in the process of analyzing the number of people flow based on on-site terminals, the moving objects in the number-of-people flow image data are key feature information. However, in the actual commodity display site, for example, there are animated advertising screens in the picture, products with screens (such as: the products placed on another display stand directly opposite the target commodity display stand are electronic products with screens such as smartphones, tablets, high-definition TVs, etc. In order to display the performance of electronic products, usually the electronic products will be set to play animated display videos, etc. when in the standby state), etc., there will be a large amount of dynamic information. On the one hand, there are dynamic light changes in this dynamic information, and on the other hand, there will be changing picture content, which is likely to interfere with the recognition and tracking results (misjudged as personnel or affecting the accuracy of trajectory tracking) in the subsequent process of identifying personnel and tracking personnel trajectories based on the YoloV5 image analysis model, affecting the adaptability and reliability of personnel identification.

[0070] Therefore, in view of the above situation, the present invention particularly proposes a technical solution for first preprocessing the acquired number-of-people flow image data based on an image analysis unit to reduce the impact of the above situation and improve the effect of further analyzing the number of people flow and counting the customer stay time according to the number-of-people flow image data subsequently.

[0071] Preferably, the image analysis unit further includes an image preprocessing unit;

[0072] The image preprocessing unit is used to preprocess the acquired number-of-people flow image data before image analysis, specifically including:

[0073] Extract the frame of the picture from the acquired number-of-people flow image data A according to the time feature, and preprocess the number-of-people flow image data A={A(1), A(2),..., A(n),..., A(N)} in sequence according to the time order, where A(n) represents the frame picture of the number-of-people flow image at time t=n;

[0074] For the frame picture A(n) of the number-of-people flow image at time t=n, transform the frame picture A(n) of the number-of-people flow image from the RGB color space to the Lab color space, and obtain the luminance channel sub-picture L(n), color channel sub-picture a(n), and color channel sub-picture b(n) of the frame picture A(n) of the number-of-people flow image;

[0075] According to the obtained luminance channel sub-picture L(n), calculate the dynamic light transformation weight factor at each pixel point position, where the dynamic light transformation weight factor calculation function used is:

[0076] ω z,t(n) (x,y)=α L ×ω L,t(n) (x,y)+α D ×ω D,t(n)(x, y)

[0077]

[0078]

[0079] where ω z,t(n) (x, y) represents the dynamic light transformation weight factor at the position of the pixel point (x, y) at time t = n, ω L,t(n) (x, y) represents the luminance factor at the position of the pixel point (x, y) at time t = n, ω D,t(n) (x, y) represents the displacement factor at the position of the pixel point (x, y) at time t = n, α L and α D respectively represent preset weight values, where α L +α D = 1; 0.3 ≤ α L ≤ 0.7; represents the average luminance channel value of each pixel point within a 3×3 range centered on the pixel point (x, y) at time t = n, where L t(n) (x + a, y + b) represents the luminance channel value at the position of the pixel point (x + a, y + b) at time t = n, represents the average luminance channel value of each pixel point within a 3×3 range centered on the pixel point (x, y) at time t = n - C, where the variable C = 1, 2,..., G, and G represents the preset number of influencing frames, 3 ≤ G ≤ 5; represents the average value of each value, L t(n) (x, y) represents the luminance channel value at the position of the pixel point (x, y) at time t = n, L t(n-1) (x, y) represents the luminance channel value at the position of the pixel point (x, y) at time t = n - 1, LS represents the preset change standard value; the variables R = X-, X+, Y-, Y+ respectively represent the negative X-axis direction, the positive X-axis direction, the negative Y-axis direction, and the positive Y-axis direction, DE atoR,t(n) (x, y) represents the pixel distance between the pixel point (x, y) at time t = n and the nearest luminance edge pixel point in the R direction, DE atoR,t(n-1) (x, y) represents the pixel distance between the pixel point (x, y) at time t = n - 1 and the nearest luminance edge pixel point in the R direction; represents the sum of the values of each |DE atoR,t(n) (x, y)-DE atoR,t(n-1) (x, y)|; where DES represents the preset displacement standard value;

[0080] Adaptive brightness adjustment processing is performed according to the dynamic light transformation weight factor at the position of each pixel point, and the adopted adaptive brightness adjustment function is:

[0081]

[0082] Among them, represents the brightness channel value at the position of pixel point (x, y) at time t = n after adaptive brightness adjustment processing, and ω z,t(n) (x, y) represents the dynamic light transformation weight factor at the position of pixel point (x, y) at time t = n, represents the brightness channel value at the position of pixel point (x, y) at time t = n, represents the average brightness channel value of each pixel point within the s×s range centered on pixel point (x, y) at time t = n, where ω = ω z,t(n) (x, y), S represents the side length dimension of the screen, represents the floor operation, L t(n) (x + a, y + b) represents the brightness channel value at the position of pixel point (x + a, y + b) at time t = n, β p and β a respectively represent preset weight values, where β p + β a = 1; 0.2 ≤ α L ≤ 0.8; LV represents the set standard brightness channel value, where 50 ≤ LV ≤ 70;

[0083] According to the brightness channel sub - map L′(n), color channel sub - map a(n) and color channel sub - map b(n) after adaptive brightness adjustment processing, reconstruction is performed to obtain the pre - processed pedestrian flow image frame A′(n);

[0084] Update n = n + 1, and repeat the above steps to further pre - process the next frame of the pedestrian flow image frame until the pre - processing of all pedestrian flow image frames is completed to obtain the pre - processed pedestrian flow image data.

[0085] Preferably, the set parameters meet the following value ranges: 10 ≤ LS ≤ 20, 2 ≤ DES ≤ 10, 0 ≤ ω L,t(n) (x, y) ≤ 1, 0 ≤ ω D,t(n) (x, y) ≤ 1.

[0086] Preferably, in the image pre - processing unit, it further includes: obtaining brightness edge pixels according to the obtained brightness channel sub - map L(n), including:

[0087] According to the brightness channel value L of each pixel point t(n)(x, y), calculate the eigenvalue Sobel(x, y) of each pixel point based on the Sobel operator, and compare the obtained eigenvalue Sobel(x, y) with the set feature threshold. When the eigenvalue is greater than the threshold, mark the pixel point (x, y) as a feature edge pixel point.

[0088] Preferably, the image analysis unit identifies and tracks customers in the image based on the image analysis model of Yolov5 according to the preprocessed pedestrian flow image data, and obtains the pedestrian flow data and customer stay time data in front of the corresponding product display stand.

[0089] In the above embodiments of the present invention, the present invention also particularly proposes a preprocessing technical solution capable of adaptively adjusting the screen light pollution in the pedestrian flow image data screen. In this solution, first, the obtained pedestrian flow image data is converted to the Lab color space, the luminance channel sub-image of the image is extracted, and based on the change characteristics of the image frame, the dynamic light transformation weight factor of each pixel point position is calculated first. Among them, a dynamic light transformation weight factor calculation function is proposed, which can accurately identify the fixed screen in the picture based on the luminance change and displacement change characteristics of the pixel point position (the idea is that considering that the position of the electronic screen in the picture is usually fixed, and the picture of the electronic screen has special luminance characteristics), and can accurately distinguish the key features (such as the people and the devices carried by the people in the picture) and pollution features (advertising screens, product electronic screens, etc.) in the pedestrian flow image picture through the dynamic light transformation weight factor. According to the obtained dynamic light transformation weight factor, an adaptive luminance adjustment function is further proposed to adaptively suppress the luminance information of the pollution features to weaken the interference degree of its dynamic information (the content displayed on the screen). Among them, through the setting of the adaptive weight factor, the adaptation effect of the luminance adjustment of the pollution features can be further improved, ensuring the visual level of the suppression area and the surrounding area, avoiding the situation of additional noise features caused by distortion in traditional luminance adjustment technologies, and at the same time, the key feature information in the image can also be highlighted, which helps to improve the adaptability and reliability of subsequent identification of people and tracking of people's trajectories based on the YoloV5 image analysis model.

[0090] Preferably, the data acquisition module further includes a data security management unit;

[0091] The data security management unit is used to perform security verification on the obtained sales data of the product. After the security verification passes, the original sales data of the product is obtained.

[0092] Among them, in the process of obtaining the sales data of goods from, for example, a third-party e-commerce platform, in order to ensure the privacy of the sales data, usually the e-commerce platform will encrypt the relevant sales data to ensure that the sales data is only provided to the corresponding merchants, and to avoid the leakage of the goods sales data between different merchant users. Therefore, after obtaining the sales data of the goods, the data security management unit is further used to perform security verification on the sales data, which helps to improve the security and privacy of data extraction.

[0093] In an exemplary scenario, the data security management unit decrypts the sales data obtained from the online platform based on the identity information of the merchant user to obtain accurate goods sales data. The sales data obtained from the online platform is encrypted by the online platform, and can only be decrypted to extract relevant data after the correct merchant user identity verification is completed.

[0094] Preferably, the data analysis module includes a statistics unit and a popularity analysis unit;

[0095] The statistics unit is used to classify and count the sales data of the goods on each sales platform for a specified good to obtain the statistical information of the goods sales data;

[0096] The popularity analysis unit is used to intelligently analyze the popularity information of the goods based on the sales data of the goods to obtain the analysis result of the goods popularity information.

[0097] Through the summary statistics of the sales data, it is possible to quantify and summarize by combining the browsing or transaction situations of the goods on different platforms, and achieve accurate statistics of the goods sales data. Based on the sales data and statistical results of the goods on each platform, further analysis of the popularity of the goods helps to further evaluate the popularity of the goods based on the goods sales data and statistical data, and improve the intelligent level of goods data processing and popularity analysis.

[0098] Preferably, in the statistics unit, the classification and statistics of the sales data of the goods on each sales platform specifically include:

[0099] Obtain the real-time access data of the goods:

[0100] TAcc i (t) = ω at-1 ×pAc 1,i (t) + ω at-2 ×pAc 2,i (t) + … + ω at-K ×pAc K,i (t)

[0101] Among them, TAcc i (t) represents the real-time access data of product i at time t; ωat-k Denote the access volume weighted adjustment factor of the preset sales platform k as pAc k,i (t) represents the access data of product i at time t on sales platform k, where k = 1, 2, …, K, and K represents the total number of sales platforms; when sales platform k is an online platform, the access data of product i at time t represents the customer access volume of the sales page of product i within a period of time based on time t; when sales platform k is an offline platform, the access data of product i at time t represents the pedestrian flow data in front of the display shelf corresponding to product i within a period of time based on time t;

[0102] Obtain the real-time transaction data of the product;

[0103] TDea i (t) = pDe 1,i (t) + pDe 2,i (t) + … + pDe K,i (t)

[0104] Among them, TDea i (t) represents the real-time transaction data of product i at time t; pDe k,i (t) represents the number of transaction orders of product i within a period of time based on time t on sales platform k, where k = 1, 2, …, K, and K represents the total number of sales platforms.

[0105] For different offline platforms and online platforms, the access volume data of products can be effectively quantified and summarized, so as to obtain the implementation access data for feedback on the products. Among them, the offline access data is added as a consideration factor, which can more intuitively feedback the popularity information of the products and improve the effect and intelligent level of product sales data management.

[0106] Preferably, in the popularity analysis unit, the popularity information of the product is intelligently analyzed according to the sales data of the product, including:

[0107] According to the current period of time t = t 0 , t 1 … t N 、 the access data pAc 1,i (t), pAc 2,i (t), … pAc K,i (t) and the real-time access data TAcc i (t) of product i on each sales platform form the feature vector set thre i (t) of product i; where t 0 represents the current time, and the larger the subscript n value in t n , the longer the time distance from the current time;

[0108] The obtained set of feature vectors is input into a trained popularity analysis model. The popularity analysis model performs feature extraction and conversion prediction based on the set of feature vectors to obtain the current time t 0 The popularity analysis result of product i; the popularity analysis model is built based on a CNN neural network. During the training process of the model, a method based on a historical set of feature vectors thre i (t B ) and the corresponding deal dataset dea i (t B-1 ) are used for training. The deal dataset consists of the deal order volume data pDe B-1 ,t B …t B+N-1 、of product i on each sales platform at different times t = t 1,i (t), pDe 2,i (t), …, pDe K,i (t) and the real-time deal data TDea i (t). The popularity analysis result output by the popularity analysis model is dea i (t -1 ), where t -1 represents a future time, corresponding to the obtained prediction result.

[0109] Based on an artificial intelligence approach, the popularity information (estimated transaction volume) of products is estimated based on the popularity analysis model. The built popularity analysis model can adaptively extract the popularity correlation features and popularity change features between platforms based on a CNN neural network, and comprehensively estimate the change amount of the popularity information of products from the dimensions of single-platform feature quantity change and cross-platform feature correlation, improving the intelligent level of product popularity estimation. Using the obtained popularity analysis result of products as a reference can enable managers to adjust the product listing strategy or hot promotion strategy, improving the intelligent level and effect of product sales management.

[0110] Preferably, the data display module is used to integrate and generate a real-time popularity report based on the popularity information analysis results of various products sold by merchant users, and push the initial popularity report to the merchant user terminal.

[0111] In an exemplary scenario, during the process of live streaming sales by merchant users, they can make targeted on-site recommendations for products with high popularity according to the real-time popularity information analysis results to improve the product sales effect.

[0112] In another exemplary scenario, the merchant user can view the analysis results of real-time product popularity information through the background system, and can adjust the products in the product hot promotion positions of the online mall to improve the product sales effect.

[0113] It should be noted that in each embodiment of the present invention, each functional unit / module can be integrated in a processing unit / module, or each unit / module can exist physically alone, or two or more units / module can be integrated in one unit / module. The above integrated unit / module can be implemented in the form of hardware or in the form of a software functional unit / module.

[0114] Through the description of the above embodiments, those skilled in the art can understand that the embodiments described herein can be implemented by hardware, software, firmware, middleware, code, or any appropriate combination thereof. For hardware implementation, the processor can be implemented in one or more of the following units: application specific integrated circuit (ASIC), digital signal processor (DSP), digital signal processing device (DSPD), programmable logic device (PLD), field programmable gate array (FPGA), processor, controller, microcontroller, microprocessor, other electronic units designed to implement the functions described herein, or a combination thereof. For software implementation, part or all of the processes of the embodiments can be completed by a computer program instructing the relevant hardware. When implemented, the above program can be stored in a computer-readable medium or transmitted as one or more instructions or codes on a computer-readable medium. The computer-readable medium includes computer storage media and communication media, where the communication media includes any medium that facilitates the transfer of a computer program from one place to another. The storage media can be any available medium that can be accessed by a computer. The computer-readable medium can include, but is not limited to, RAM, ROM, EEPROM, CD-ROM or other optical disc storage, magnetic disk storage media or other magnetic storage devices, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer.

[0115] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than limiting the protection scope of the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the essence and scope of the technical solutions of the present invention.

Claims

1. A smart e-commerce business data middle-end management system, characterized in that: It includes data collection module, data analysis module and data display module; among them, The data collection module is used to collect sales data of various commodities sold by merchant users on different sales platforms, where the sales platforms include online platforms and offline platforms; sales data includes browsing data and transaction data; The data analysis module is used to analyze the popularity information of the current commodity based on the acquired sales data of each commodity, and obtain the commodity popularity information analysis result; The data display module is used to generate a real-time heat report based on the analysis results of the heat information of various commodities sold by merchant users, and push the real-time heat report to the merchant user terminal; Wherein, the data collection module includes an offline data collection unit; The offline data collection unit is used to connect with the on-site terminal set up on the offline platform, obtain the browsing data and transaction data of the specified commodity collected by the on-site terminal, and associate the acquired browsing data and transaction data with the corresponding commodity; wherein the offline platform includes the merchant's offline store, and the browsing data includes the flow of people in front of the commodity display shelf and the customer's stay time data; The on-site terminal includes an image acquisition device, through which image data of the flow of people in front of the commodity display rack is obtained; The offline acquisition unit also includes an image analysis unit and a correlation unit; The image analysis unit is used to identify and track customers in the image based on the acquired human flow image data and the image analysis model of Yolov5, and obtain the human flow data and customer stay time data in front of the corresponding product display rack; The association unit is used to associate the acquired traffic flow data and customer stay time data with the corresponding commodities; The image analysis unit also includes an image preprocessing unit; The image preprocessing unit is used to preprocess the acquired human flow image data before image analysis, specifically including: Extracting frames from the acquired human flow image data A according to the time feature, and preprocessing the human flow image data A = {A(1), A(2), ..., A(n), ..., A(N)} in chronological order, wherein A(n) represents a human flow image frame at time t = n; For the pedestrian flow image frame A(n) at time t=n, the pedestrian flow image frame A(n) is transformed from the RGB color space to the Lab color space, and the brightness channel sub-image L(n), the color channel sub-image a(n) and the color channel sub-image b(n) of the pedestrian flow image frame A(n) are obtained; According to the acquired brightness channel sub-image L(n), the dynamic light transformation weight factor of each pixel position is calculated, wherein the dynamic light transformation weight factor calculation function used is: oh z，t(n) (x, y) = a L ×ω L,t(n) (x,y)+a D ×ω D,t(n) (x, y) Among them, ω z，t(n) (x, y) represents the dynamic light transformation weight factor of the pixel point (x, y) at time t = n, ω L，t(n) (x, y) represents the brightness factor of the pixel point (x, y) at time t=n, ω D，t(n) (x, y) represents the displacement factor of the pixel point (x, y) at time t = n, α L and α D Respectively represent the preset weight values, where α L +α D =1; 0.3≤α L ≤0.7; represents the average brightness channel value of each pixel in a 3×3 range centered at the pixel point (x, y) at time t=n, where L t(n) (x+a, y+b) represents the brightness channel value of the pixel point (x+a, y+b) at time t=n. represents the average brightness channel value of each pixel in a 3×3 range centered on the pixel (x, y) at time t=nC, where the variable C=1, 2, …, G, G represents the preset number of affected frames, 3≤G≤5; Indicates each The average value of L t(n) (x, y) represents the brightness channel value of the pixel point (x, y) at time t=n, L t(n-1) (x, y) represents the brightness channel value of the pixel point (x, y) at time t=n-1, LS represents the preset change standard value; the variable R=X-, X+, Y-, Y+ represent the negative direction of the X axis, the positive direction of the X axis, the negative direction of the Y axis, and the positive direction of the Y axis respectively, DE atoR，t(n) (x, y) represents the pixel distance between the pixel point (x, y) at time t=n and the nearest brightness edge pixel in the R direction, DE atoR,t(n-1) (x, y) represents the pixel distance between the position of the pixel point (x, y) at time t=n-1 and the nearest brightness edge pixel in the R direction; Indicates each |DE atoR，t(n) (x, y)-DE atoR,t(n-1) (x, y)|; where DES represents the preset displacement standard value; Adaptive brightness adjustment is performed according to the dynamic light transformation weight factor of each pixel position, where the adaptive brightness adjustment function used is: in, represents the brightness channel value of the pixel point (x, y) at time t = n after adaptive brightness adjustment processing, ω z，t(n) (x, y) represents the dynamic light transformation weight factor of the pixel point (x, y) at time t = n, Represents the brightness channel value of the pixel point (x, y) at time t=n, represents the average brightness channel value of each pixel within the s×s range centered at the pixel point (x, y) at time t=n, where ω=ω z，t(n) (x, y), S represents the side length of the picture, Indicates floor operation. L t(n) (x+a, y+b) represents the brightness channel value of the pixel point (x+a, y+b) at time t=n, β p and β a Respectively represent the preset weight values, where β p +β a =1; LV represents the set standard brightness channel value, where 50≤LV≤70; Reconstruct the brightness channel sub-image L′(n), the color channel sub-image a(n) and the color channel sub-image b(n) after the adaptive brightness adjustment to obtain the pre-processed human flow image frame A′(n); Update n=n+1, repeat the above steps to further preprocess the next frame of the human flow image frame, until the preprocessing of all human flow image frames is completed, and the preprocessed human flow image data is obtained.

2. According to claim 1, a smart e-commerce business data middle platform management system is characterized in that: Also included is a data storage module; The data storage module is used to associate and store the acquired sales data and commodity popularity information with the corresponding commodities according to the analysis results, and to build commodity data archives, and to classify, store and manage the commodity data archives to build a commodity database.

3. According to claim 1, a smart e-commerce business data middle platform management system is characterized in that: It also includes a smart racking module; The intelligent listing module is used to intelligently list or push products based on the real-time popularity report and product popularity information.

4. According to claim 1, a smart e-commerce business data middle platform management system is characterized in that: The data collection module includes an online data collection unit; The online data collection unit is used to connect to the data interface of each online platform respectively, obtain the sales data of the specified commodity provided by the online platform, and associate the obtained sales data with the corresponding commodity; The online platforms include merchants’ online sales platforms and third-party e-commerce platforms.

5. According to claim 1, a smart e-commerce business data middle platform management system is characterized in that: The data acquisition module also includes a data security management unit; The data security management unit is used to perform security verification on the acquired commodity sales data. When the security verification passes, the original commodity sales data is obtained.

6. According to claim 1, a smart e-commerce business data middle platform management system is characterized in that: The data analysis module includes a statistical unit and a heat analysis unit; The statistical unit is used to classify and count the sales data of a specified commodity on each sales platform to obtain sales data statistical information of the commodity; The heat analysis unit is used to perform intelligent analysis on the heat information of the commodity according to the sales data of the commodity to obtain the analysis result of the heat information of the commodity.

7. According to claim 6, a smart e-commerce business data middle platform management system is characterized in that: In the statistical unit, the sales data of the products on each sales platform are classified and counted, including: Get real-time access data of products: TAcc i (t)=ω at-1 ×pAc 1，i (t)+ω at-2 ×pAc 2，i (t)+…+ω at-K ×pAc K，i (t) Among them, TAcc i (t) represents the real-time access data of product i at time t; ω at-k represents the weighted adjustment factor of the visits of the preset sales platform k, pAc k，i (t) represents the access data of product i at time t in sales platform k, where k = 1, 2, ..., K, K represents the total number of sales platforms; when sales platform k is an online platform, the access data of product i at time t represents the number of customer visits to the sales page of product i within a period of time based on time t; when sales platform k is an offline platform, the access data of product i at time t represents the flow of people in front of the display stand of corresponding product i within a period of time based on time t; Get real-time transaction data of goods; TDea i (t)=pDe 1，i (t)+pDe 2，i (t)+…+pDe K，i (t) Among them, TDea i (t) represents the real-time transaction data of commodity i at time t; pDe k，i (t) represents the number of completed orders for product i in sales platform k within a period of time based on time t, where k = 1, 2, ... K, and K represents the total number of sales platforms.

8. According to claim 7, a smart e-commerce business data middle platform management system is characterized in that: In the heat analysis unit, the heat information of the product is intelligently analyzed based on the sales data of the product, including: According to the current period of time t = t0, t1…t N , product i's access data on various sales platforms 1，i (t), pAc 2，i (t),…pAc K，i (t) and real-time access to data TAcc i (t) The feature vector set thre that makes up product i i (t); where t0 represents the current time, t n The larger the subscript n value is, the longer the time from the current moment is. The obtained feature vector set is input into the trained heat analysis model, and the heat analysis model extracts features and predicts transaction conversion based on the feature vector set to obtain the product heat analysis result of product i at the current time t0; the heat analysis model is built based on the CNN neural network, and in the model training process, the historical feature vector set thre is used. i (t B ) and the corresponding transaction data set dea i (t B-1 ) training, where the transaction data set consists of a period of time t=t B-1 , t B …t B+N-1 , the transaction order volume data of product i on each sales platform pDe 1，i (t), pDe 2，i (t),…,pDe K，i (t) and real-time transaction data TDea i (t) composition; the heat analysis result of the commodity output by the heat analysis model is dea i (t -1 ), t -1 Indicates the future moment and the corresponding prediction result.

Images