Artificial intelligence store entry collection pushing system

Abstract

The invention, which belongs to the technical field of commodity recommendation, discloses an artificial intelligence store-in collection pushing system comprising a platform terminal and a user terminal. The platform end comprises an installation service module and a user service module; the installation service module is used for serving the user to install the user side according to the application scene of the user; the user service module is used for being in butt joint with user sides of all users and receiving service requirements of the user sides in real time; performing corresponding service processing according to the service demand; the user side comprises a shop information management module, a consumption analysis module and a buffer analysis module; the shop information management module is used for performing information management on a user shop and establishing a commodity information base and a member information base; the consumption analysis module is used for determining recommended commodity information and preferential information of consumers; and the buffer analysis module is used for carrying out buffer analysis according to the recommended commodity information and the consumption bill, determining the buffer consumption time of the consumer, and displaying the buffer consumption time, the preferential information and the recommended commodity information when the consumer checks out.

Description

Technical Field

[0001] The present invention belongs to the technical field of commodity recommendation, and specifically relates to an artificial intelligence store payment collection and push system. Background Art

[0002] In the current retail industry, the payment process is often regarded as the end of the transaction process, while ignoring its potential in product recommendation and discount push. Traditional payment systems mainly focus on the rapid completion of transactions, but lack in-depth analysis of customer shopping behaviors and preferences. Therefore, the existing technology has obvious deficiencies in providing personalized product recommendations and discount information to customers when collecting payments. Based on this, in order to solve the above problems, the present invention provides an artificial intelligence store payment collection and push system. Summary of the invention

[0003] In order to solve the problems existing in the above solutions, the present invention provides an artificial intelligence in-store payment collection and push system.

[0004] The purpose of the present invention can be achieved through the following technical solutions: Artificial intelligence in-store payment push system, including platform side and user side; The platform end includes an installation service module and a user service module; the installation service module is used to serve the user to install the user end according to the user's application scenario; The user service module is used to connect to the user terminal of each user, receive the service requirements of the user terminal in real time, and perform corresponding service processing according to the service requirements.

[0005] Furthermore, application scenarios include manual checkout scenarios and unmanned checkout scenarios.

[0006] The user terminal includes a store information management module, a consumption analysis module and a buffer analysis module; The store information management module is used to manage information on user stores and establish a product information database and a member information database. The product information database is used to store information on products sold in the user store, including brand, product image, price, production date, manufacturer, and shelf location; the member information database is used to store member information of the user store, including account number, shopping record, birthday, and generate consumption portraits of corresponding members based on the stored member information, and store the consumption portraits in the member information database.

[0007] Furthermore, the generation of consumer portraits includes: The platform has a built-in target portrait generation method, analyzes the member information according to the target portrait generation method, and generates a consumption portrait of the member.

[0008] Furthermore, the determination of the target portrait generation method includes: Collecting several portrait generation methods to be selected; establishing a portrait test set, wherein the portrait test set includes several member simulation information; analyzing the portrait test set through the portrait generation method to be selected, and generating a test portrait corresponding to the portrait generation method to be selected; The platform will preset the corresponding evaluation items and the corresponding evaluation weights; Analyze the test profile according to the evaluation items to obtain test values ​​corresponding to the evaluation items; Normalize the test values ​​of each test profile according to the evaluation items; mark the normalized test values ​​as CV i , i represents the corresponding evaluation item, i=1, 2, ..., n, n is the number of evaluation items; the evaluation weight of the evaluation item is marked as β i ; Calculate the corresponding portrait screening value according to the portrait evaluation formula. The portrait evaluation formula is: ; Where: HS is the image screening value; The candidate portrait generation method with the largest portrait screening value is selected as the target portrait generation method.

[0009] The consumption analysis module is used to identify the consumer's consumption bill in real time, perform recommendation analysis based on the consumption bill, obtain the consumer's recommended product information; and identify the discount information corresponding to the recommended product information.

[0010] Furthermore, a recommendation analysis is performed based on the consumption bill, including: Identify the consumer's membership information and consumption bills, and obtain discount information from the user's store; integrate the obtained membership information, consumption bills and discount information into recommended material data; The recommended material data is analyzed through the recommendation analysis model preset by the platform to determine the recommended products and collect the recommended product information of the recommended products.

[0011] Furthermore, in the process of identifying the membership information of a consumer, when the consumer is not a member of the user's store, corresponding membership information is generated for the consumer.

[0012] Furthermore, corresponding membership information is generated for non-member consumers, including: Identify member information stored in the member information database, and classify the member information according to gender and age to form a number of member categories; Setting corresponding member representative information for the member classification pair; establishing a non-member matching library according to the member classification and member representative information; The gender and age of the consumer are obtained, and the corresponding member representative information is matched from the non-member matching library according to the obtained gender and age, and the member representative information is used as the member information of the consumer.

[0013] Furthermore, member information is classified by gender and age, including: Step SA1: Classify member information according to gender to obtain two initial classifications; Identify the age corresponding to the corresponding member information in the initial classification, classify the member information according to the age to obtain a plurality of unit classifications, set the unit information of the unit classifications, and sort the unit classifications in descending order according to the corresponding age to obtain a first sequence; Step SA2: define a base classification and an analysis classification, wherein the base classification is the unit classification ranked first in the first sequence; and the analysis classification is the unit classification ranked second in the first sequence; Step SA3: identifying the corresponding reference classification and analysis classification according to the first sequence, identifying the unit information corresponding to the reference classification and the analysis classification respectively, and calculating the similarity between the reference classification and the analysis classification according to the identified unit information; Step SA4: when the similarity is greater than the threshold value X1, the benchmark classification and the analysis classification are merged to form a new benchmark classification, and the unit information of the new benchmark classification is updated; When the similarity is not greater than the threshold value X1, the benchmark classification is marked as the member classification, and the member classification is removed from the first sequence to form a new first sequence; Step SA5: looping through steps SA3 to SA4 until there are no more analysis categories in the first sequence, marking the remaining benchmark categories as member categories, and ending the analysis.

[0014] The buffer analysis module is used to perform buffer analysis based on the recommended product information and the consumption bill to determine the consumer's buffer consumption time, where the buffer consumption time means that the consumer can enjoy corresponding discount information when purchasing the recommended product within the buffer consumption time; identify the discount information corresponding to the recommended product information; add the buffer consumption time and discount information to the consumption bill; and display the buffer consumption time, discount information and recommended product information when the consumer checks out.

[0015] Furthermore, a buffer analysis is performed based on the recommended product information and the consumption bill, including: Identify the corresponding recommended product amount and consumption amount based on the recommended product information and consumption bill; calculate the corresponding amount analysis value based on the analysis formula; the analysis formula is: PK=b1×FP+b2×TP; Where: PK is the amount analysis value; b1 and b2 are both proportionality coefficients, and the value ranges are 0 < b1 < 1 and 0 < b2 < 1; FP is the consumption amount; TP is the recommended product amount. The platform party presets a buffer matching curve. The horizontal axis of the buffer matching curve is the amount analysis value, and the vertical axis is the buffer consumption duration. Match the corresponding buffer consumption duration from the buffer matching curve according to the calculated amount analysis value, and determine the buffer consumption time according to the buffer consumption duration and the checkout time.

[0016] Compared with the prior art, the beneficial effects of the present invention are: By seamlessly integrating product information and preferential push, the present invention enables consumers to instantly obtain preferential information related to the purchased products during the payment process. This instant feedback mechanism greatly improves the shopping experience of consumers and reduces the time and energy they spend looking for preferential information; at the same time, display the recommended product information, buffer consumption time, and preferential information to consumers, enabling consumers to have shopping considerations within the buffer consumption time, prompting consumers to repurchase, and avoiding the situation where consumers do not apply preferential information due to the time pressure of settlement; at the same time, dynamically adjust the buffer consumption duration according to the actual consumption situation of consumers, which is convenient for fully considering the benefits of user shops. Description of the Drawings

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0018] Figure 1 It is a principle block diagram of the present invention. Detailed Embodiments

[0019] The following will clearly and completely describe the technical solutions of the present invention in combination with the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0020] As Figure 1 shown, the artificial intelligence in-store payment push system includes a platform end and a user end; The platform end is used and maintained by the platform party, and includes an installation service module and a user service module; The installation service module is used to serve the user to install the user end according to the application scenario of the user's actual store.

[0021] The user terminal mainly has two application scenarios. One application scenario is the manual cashier scenario, where the user terminal is connected to the corresponding device as the corresponding software and program, so that the user terminal can run the corresponding functions, such as data collection, etc. The other application scenario is the unmanned cashier scenario, that is, cashiering is performed using smart cashier equipment.

[0022] In one embodiment, the installation of the user end may be performed by manual service installation and debugging by the platform staff.

[0023] In one embodiment, the installation of the user terminal can be performed through automatic installation, that is, according to various existing practical application scenarios, the platform party presets the corresponding version of the user terminal, so that subsequent users can directly download and install it by themselves.

[0024] The user service module is used to connect to the user terminal of each user, receive the service requirements of the user terminal in real time, such as maintenance, exception handling, recommendation algorithm optimization and other preset demand functions; and perform corresponding service processing according to the received service requirements.

[0025] The user terminal includes a store information management module, a consumption analysis module and a buffer analysis module; The store information management module is used to manage information on the user's store and establish a product information database and a member information database. The product information database is used to store information on various products sold in the user's store, including brand, product image, price, production date, manufacturer, shelf location and other related information; the member information database is used to store consumer member information of the user's store, including account number, shopping record, birthday and other related information. If the store is a chain store, all member information can be stored; and consumption portraits of the corresponding members are generated based on the stored member information, and the consumption portraits are stored in the member information database.

[0026] In one embodiment, the product information and member information stored in the product information database and the member information database can be obtained by connecting to relevant systems such as the cash register system and the member system used in the user's store.

[0027] In one embodiment, the consumer portrait is generated using current portrait generation technology, such as selecting the portrait generation technology to be applied from portrait generation technologies applied on various sales platforms.

[0028] Exemplarily, various portrait generation technologies that can be applied in the user terminal of the present invention are identified, such as collecting content according to the consumption portrait preset by the platform; marking the obtained portrait generation technology as a portrait generation method to be selected; Establish a portrait test set, which includes member information simulated and set by the platform, marked as member simulation information; analyze the portrait test set through the candidate portrait generation method to generate a test portrait corresponding to the corresponding candidate portrait generation method; The platform party shall preset the corresponding evaluation items and the corresponding evaluation weights of the evaluation items, such as evaluation accuracy (recommendation accuracy), evaluation efficiency of identifying and understanding the test profile in the subsequent recommendation analysis process (recommendation efficiency), memory size and other corresponding evaluation items, and shall set them according to the actual needs of the platform party; Analyze the test profile according to the evaluation items to obtain the test values ​​corresponding to each evaluation item, such as evaluation accuracy, evaluation efficiency, memory usage, etc.; simulate subsequent analysis according to the test profile to determine the test value of the evaluation item; Normalize the test values ​​of each test profile according to the evaluation items to facilitate subsequent calculations. For example, if the evaluation item of evaluation accuracy is normalized and has corresponding values ​​of 1, 0.5, and 0.8, then normalize it according to [0, 1], and correspond the evaluation accuracy 1 to 1, and the evaluation accuracy 0.5 to 0. Other evaluation accuracies are converted according to the corresponding ratio, such as the evaluation accuracy 0.8 corresponds to 0.6; The normalized test value is labeled as CV i , i represents the corresponding evaluation item, i=1, 2, ..., n, n is the number of evaluation items; the evaluation weight of the evaluation item is marked as β i ; Calculate the corresponding portrait screening value according to the portrait evaluation formula. The portrait evaluation formula is: ; Where: HS is the image screening value; The candidate portrait generation method with the largest portrait screening value is selected as the target portrait generation method.

[0029] Subsequently, member information is analyzed according to the target portrait generation method to generate corresponding consumption portraits.

[0030] The consumption analysis module is used to identify the consumer's consumption bill in real time, perform recommendation analysis based on the identified consumption bill, and determine the recommended product information of the consumer. Identifying the preferential information corresponding to the recommended product information refers to the relevant preferential information for the recommended product and the purchased product.

[0031] The consumption bill includes the information of the goods purchased by the consumer, quantity, price and other relevant information.

[0032] In one embodiment, the recommended products may be determined based on existing recommendation technologies.

[0033] In one embodiment, performing recommendation analysis based on the consumption bill includes: Identify the consumer's membership information and consumption bills, and obtain discount information from the user's store, such as discounts on certain types of goods, discounts on purchases above a certain amount, and other related information; integrate the obtained membership information, consumption bills, and discount information into recommended material data; The platform establishes a recommendation analysis model, which is used to analyze the recommendation material data and determine the recommended products.

[0034] The recommendation analysis model is established based on existing intelligent recommendation technology.

[0035] Exemplary: Feature extraction: Extract consumer feature labels from consumer portrait data, such as age group, gender preference, regional characteristics, etc.

[0036] Extract consumers’ purchasing characteristics from consumption bill data, such as purchase frequency, purchase category preference, purchase time period, etc.

[0037] Extract discount features from store discount information, such as discount strength, discount type, discount validity period, etc.

[0038] Feature Selection: According to the requirements of the recommendation algorithm, select features that have a significant impact on the recommendation effect.

[0039] Use feature selection algorithms, such as chi-square test, mutual information, recursive feature elimination, etc., to screen and optimize features.

[0040] Selection and optimization of recommendation algorithms: Content-based recommendation algorithm: Recommend products similar to consumers’ interests based on their purchase history and browsing behavior. Use text mining technology to analyze text information such as product descriptions and user comments to extract product feature tags. Based on the consumer’s feature tags and product feature tags, calculate the similarity between the consumer and the product and make recommendations.

[0041] Collaborative filtering recommendation algorithm: Use the historical behavior data of consumers to find other consumers with similar behaviors, and then recommend products based on the behaviors of these similar consumers. It can be divided into user-based collaborative filtering and product-based collaborative filtering. By calculating the similarity between consumers or products, a user-product matrix or product-product matrix is ​​constructed to make recommendations.

[0042] Deep learning recommendation algorithm: Use deep learning technologies such as convolutional neural network (CNN), recurrent neural network (RNN), deep neural network (DNN), etc. to model and analyze consumers' historical behaviors. Through training models, learn the potential relationship between consumers and products, and make recommendations. It can be combined with attention mechanisms, memory networks and other technologies to improve the accuracy and personalization of recommendations.

[0043] Hybrid recommendation algorithm: Combine the advantages of multiple recommendation algorithms to perform hybrid recommendation. The results of different recommendation algorithms can be integrated and optimized using weighted fusion, serial fusion, parallel fusion and other methods.

[0044] Model evaluation and optimization: Evaluation indicators: Use accuracy, recall, F1 score, AUC and other indicators to evaluate the performance of the recommendation algorithm. Use methods such as A / B testing to compare the effects of different recommendation strategies and select the best recommendation solution.

[0045] Optimization method: According to the evaluation results, the recommendation algorithm is tuned, such as adjusting parameters, optimizing feature selection, etc. New consumer data and store discount information are continuously collected to update and optimize the recommendation model. Using online learning technology, the recommendation model can be updated and optimized in real time to improve the accuracy and real-time performance of the recommendation.

[0046] In one embodiment, if the consumer is not a member of the user's store, the member information is representative unified member information set according to the information of each member, such as the member information determined by the average value, mode, etc. as the member information of the non-member user.

[0047] In one embodiment, if the consumer is not a member of the user's store, the member information is obtained as follows: Identify member information stored in the member information database, and classify the member information by gender and age to form several member categories; Identify the member information corresponding to the member classification, and then set the representative member information corresponding to the member classification, such as the average value, mode, etc.; mark as the member representative information of the member classification; Establish a non-member matching database based on member classification and member representative information; When the consumer is not a member, the consumer's gender and age can be determined by video recognition, manual estimation, inquiry, consumer input, etc.; the corresponding member representative information is matched from the non-member matching library based on the obtained gender and age, and the obtained member representative information is used as the member information of the consumer.

[0048] In one embodiment, the member information is classified according to gender and age, and the classification can be performed in an existing manner, such as classification according to preset gender and age groups.

[0049] In one embodiment, member information is classified according to gender and age, including: Step SA1: Classify member information according to gender to obtain two initial classifications; Identify the age corresponding to the corresponding member information in the initial classification, classify the member information according to age, and obtain several unit classifications, that is, assign member information of the same age to one classification, mark the classification as a unit classification, set the unit information of the unit classification, and use the member information corresponding to the average value, mode, etc. as the representative member information of the unit classification, and mark it as the unit information; sort the unit classifications in descending order according to the corresponding age to obtain a first sequence; Step SA2: define a base classification and an analysis classification, wherein the base classification is the first unit classification in the first sequence; and the analysis classification is the second unit classification in the first sequence; Step SA3: identifying the corresponding benchmark classification and analysis classification according to the first sequence, identifying the unit information corresponding to the benchmark classification and the analysis classification respectively, and calculating the similarity between the benchmark classification and the analysis classification according to the identified unit information; calculating the similarity between the corresponding unit information according to the weight coefficient of the preset content item, and the similarity can be calculated according to the existing cosine similarity algorithm; Step SA4: When the similarity is greater than the threshold value X1, the benchmark classification and the analysis classification are merged to form a new benchmark classification, and the unit information of the benchmark classification is updated according to the setting method of the original unit information; When the similarity is not greater than the threshold value X1, the benchmark classification is marked as the member classification, and the member classification is removed from the first sequence to form a new first sequence; Step SA5: looping through steps SA3 to SA4 until there are no more analysis categories in the first sequence, marking the remaining benchmark categories as member categories, and ending the analysis.

[0050] The buffer analysis module is used to perform buffer analysis based on the recommended product information and the consumption bill to determine the consumer's buffer consumption time; the buffer consumption time means that the consumer can enjoy corresponding preferential information when purchasing the recommended product within the buffer consumption time, such as superimposed discounts, etc.; identify the preferential information corresponding to the recommended product information; add the buffer consumption time and preferential information to the consumption bill, which refers to the consumption bill that has not yet been printed; display the buffer consumption time, preferential information and recommended product information when the consumer checks out.

[0051] In one embodiment, a buffer analysis is performed based on recommended product information and consumption bills, and the buffer consumption time can be determined based on the consumption amount and the recommended product amount. The larger the amount, the longer the buffer consumption time, and the buffer consumption time is then determined based on the buffer consumption time.

[0052] Exemplarily, the consumption amount and the recommended product amount are identified; and the corresponding amount analysis value is calculated according to the analysis formula; the analysis formula is: PK=b1×FP+b2×TP; Where: PK is the amount analysis value; b1 and b2 are both proportionality coefficients, and the value ranges are 0 < b1 < 1 and 0 < b2 < 1; FP is the consumption amount; TP is the recommended product amount; The platform presets a buffer matching curve. The horizontal axis of the buffer matching curve is the amount analysis value, and the vertical axis is the buffer consumption duration; Match the corresponding buffer consumption duration according to the calculated amount analysis value, and determine the buffer consumption time according to the buffer consumption duration and the current time.

[0053] In one embodiment, the buffer matching curve can be established by existing curve establishment methods, fitting, etc.

[0054] For example, the platform presets several groups of amount analysis values and buffer consumption durations; use the preset several groups of data for fitting to form a buffer matching curve.

[0055] In one embodiment, buffer analysis is performed according to the recommended product information and the consumption bill. A buffer analysis model can be established based on a deep neural network, etc., and the platform sets a corresponding training set for training. The training set includes input data and output data. The input data is the consumption amount and the recommended product amount, and the output data is the buffer consumption duration.

[0056] The above formulas are all calculated by removing the dimension and taking their numerical values. The formulas are obtained by collecting a large amount of data for software simulation to get a formula closest to the actual 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 simulating a large amount of data.

[0057] The above embodiments are only used to illustrate the technical method of the present invention and not to limit it. 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 method of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical method of the present invention.

Claims

1. The artificial intelligence store payment push system is characterized by: Including platform side and user side; The platform end includes an installation service module and a user service module; the installation service module is used to serve the user to install the user end according to the user's application scenario; The user service module is used to connect to the user terminal of each user, receive the service requirements of the user terminal in real time, and perform corresponding service processing according to the service requirements; The user terminal includes a store information management module, a consumption analysis module and a buffer analysis module; The store information management module is used to manage the information of the user's store, establish a product information database and a member information database, the product information database is used to store the product information sold in the user's store, the product information includes brand, product image, price, production date, manufacturer, shelf location; the member information database is used to store the member information of the user's store, the member information includes account number, shopping record, birthday, and generate the consumption portrait of the corresponding member according to the stored member information, and store the consumption portrait in the member information database; The consumption analysis module is used to identify the consumer's consumption bill in real time, perform recommendation analysis based on the consumption bill, obtain the consumer's recommended product information; and identify the preferential information corresponding to the recommended product information; The buffer analysis module is used to perform buffer analysis based on the recommended product information and the consumption bill to determine the consumer's buffer consumption time, where the buffer consumption time means that the consumer can enjoy corresponding discount information when purchasing the recommended product within the buffer consumption time; identify the discount information corresponding to the recommended product information; add the buffer consumption time and discount information to the consumption bill; and display the buffer consumption time, discount information and recommended product information when the consumer checks out.

2. The artificial intelligence store payment collection and push system according to claim 1 is characterized in that: Application scenarios include manual checkout scenarios and unmanned checkout scenarios.

3. The artificial intelligence store payment collection and push system according to claim 1 is characterized in that: Generation of consumer portraits, including: The platform has a built-in target portrait generation method, analyzes the member information according to the target portrait generation method, and generates a consumption portrait of the member.

4. The artificial intelligence store payment collection and push system according to claim 3 is characterized in that: Determination of the target profile generation method, including: Collecting several portrait generation methods to be selected; establishing a portrait test set, wherein the portrait test set includes several member simulation information; analyzing the portrait test set through the portrait generation method to be selected, and generating a test portrait corresponding to the portrait generation method to be selected; The platform will preset the corresponding evaluation items and the corresponding evaluation weights; Analyze the test profile according to the evaluation items to obtain test values ​​corresponding to the evaluation items; Normalize the test values ​​of each test profile according to the evaluation items; mark the normalized test values ​​as CV i , i represents the corresponding evaluation item, i=1, 2, ..., n, n is the number of evaluation items; the evaluation weight of the evaluation item is marked as β i ; The corresponding portrait screening value is calculated according to the portrait evaluation formula. The portrait evaluation formula is: ; Where: HS is the image screening value; The candidate portrait generation method with the largest portrait screening value is selected as the target portrait generation method.

5. The artificial intelligence store payment collection and push system according to claim 1 is characterized in that: Recommended analysis based on consumption bills, including: Identify the consumer's membership information and consumption bills, and obtain discount information from the user's store; integrate the obtained membership information, consumption bills and discount information into recommended material data; The recommended material data is analyzed through the recommendation analysis model preset by the platform to determine the recommended products and collect the recommended product information of the recommended products.

6. The artificial intelligence store payment collection and push system according to claim 5 is characterized in that: In the process of identifying the membership information of a consumer, when the consumer is not a member of the user's store, corresponding membership information is generated for the consumer.

7. The artificial intelligence store payment collection and push system according to claim 6 is characterized in that: Generate corresponding membership information for non - member consumers, including: Identify the membership information stored in the membership information database, classify the membership information by gender and age to form several membership classifications; Set corresponding membership representative information for the membership classifications; establish a non - member matching database based on the membership classifications and membership representative information; Obtain the gender and age of the consumer, match the corresponding membership representative information from the non - member matching database according to the obtained gender and age, and use the membership representative information as the membership information of the consumer.

8. The artificial intelligence store payment collection and push system according to claim 7 is characterized in that: Classify the membership information by gender and age, including: Step SA1: Classify the membership information by gender to obtain two initial classifications; Identify the ages corresponding to the membership information within the initial classifications, classify the membership information by age to obtain several unit classifications, set the unit information of the unit classifications, and sort the unit classifications in descending order according to the corresponding ages to obtain the first sequence; Step SA2: Define a reference classification and an analysis classification, where the reference classification is the unit classification ranked first in the first sequence; the analysis classification is the unit classification ranked second in the first sequence; Step SA3: Identify the corresponding reference classification and analysis classification according to the first sequence, identify the unit information corresponding to the reference classification and the analysis classification respectively, and calculate the similarity between the reference classification and the analysis classification according to the identified unit information; Step SA4: When the similarity is greater than the threshold X1, merge the reference classification and the analysis classification to form a new reference classification, and update the unit information of the new reference classification; When the similarity is not greater than the threshold X1, mark the reference classification as a membership classification, remove the membership classification from the first sequence to form a new first sequence; Step SA5: Loop through steps SA3 to SA4 until there is no analysis classification in the first sequence, mark the remaining reference classifications as membership classifications, and end the analysis.

9. The artificial intelligence store payment collection and push system according to claim 1 is characterized in that: Conduct buffer analysis based on the recommended product information and consumption bills, including: Identify the corresponding recommended product amount and consumption amount according to the recommended product information and consumption bills; calculate the corresponding amount analysis value according to the analysis formula; the analysis formula is: PK = b1×FP + b2×TP; In the formula: PK is the amount analysis value; b1 and b2 are both proportionality coefficients, and the value range is 0 < b1 < 1, 0 < b2 < 1; FP is the consumption amount; TP is the recommended product amount; The platform party presets a buffer matching curve, where the horizontal axis of the buffer matching curve is the amount analysis value and the vertical axis is the buffer consumption duration; Match the corresponding buffer consumption duration from the buffer matching curve according to the calculated amount analysis value, and determine the buffer consumption time according to the buffer consumption duration and the checkout time.

Images