Multi-terminal transaction platform based on data mining

Abstract

The invention relates to the technical field of data processing and electronic commerce, in particular to a multi-terminal transaction platform based on data mining, which comprises a data acquisition and preprocessing module used for acquiring, integrating, cleaning and storing user data and commodity data; the security and privacy protection module is used for encrypting storage and setting an access control strategy and an authentication mechanism; the personalized recommendation and prediction module is used for carrying out personalized recommendation on commodities and predicting future demands of users; the supply chain management module is used for predicting a sales result and optimizing inventory management; the commodity bidding module is used for auction bidding; the risk early warning module is used for monitoring abnormal behaviors in real time and dividing early warning signals into different grades; the optimization and improvement module is used for carrying out regular evaluation and summarization and is also used for improving and optimizing a model and an algorithm; through data mining and analysis, personalized transaction experience is provided for different terminal users, and the platform operation efficiency and the user experience are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing and e-commerce, and in particular to a multi-terminal transaction platform based on data mining. Background Art

[0002] With the rapid development of e-commerce, trading platforms have accumulated huge amounts of transaction data. However, traditional data processing methods make it difficult to effectively mine the hidden information in these data, limiting the platform's service quality and user experience. In addition, with the popularization of mobile Internet, users' demands for the convenience, real-timeness and personalization of trading platforms are growing. Therefore, it is particularly important to develop a trading platform that can span multiple terminals, efficiently process and analyze data, and provide personalized services.

[0003] In related technologies, the problem of being difficult to fully mine and utilize valuable information hidden in data has not yet been solved. Summary of the invention

[0004] To this end, the present invention provides a multi-terminal transaction platform based on data mining to overcome the problem in the prior art that it is difficult to fully mine and utilize the valuable information hidden in the data.

[0005] To achieve the above object, the present invention provides a multi-terminal transaction platform based on data mining, the platform comprising:

[0006] The data collection and preprocessing module is used to collect user data and product data; it is also used to integrate, clean and store the collected user data and product data;

[0007] The security and privacy protection module is used to encrypt and store all sensitive data based on the collected user data and product data, and to set access control policies and authentication mechanisms to verify and authorize user identities;

[0008] Personalized recommendation and prediction module, which is used to extract useful data information based on the collected user data and product data, use recommendation algorithms to make personalized recommendations for products, and predict future user needs based on time series analysis;

[0009] Supply chain management module, which is used to predict sales results based on the predicted future demand of users, determine whether the current inventory can meet future sales demand based on sales trends, and optimize inventory management;

[0010] The commodity bidding module is used to conduct auction bidding for commodities with scarce inventory according to the inventory management results;

[0011] The risk warning module is used to build a transaction risk assessment model based on the commodity purchase structure, monitor abnormal behaviors in the transaction process in real time, and classify warning signals into different levels according to the severity of abnormal behaviors;

[0012] The optimization and improvement module is used to regularly evaluate and summarize the operation of the risk warning module to obtain evaluation results, and is also used to improve and optimize the transaction risk assessment model and abnormal behavior detection algorithm based on the evaluation results.

[0013] Furthermore, the security and privacy protection module encrypts and stores all sensitive data based on the collected user data and product data, desensitizes the sensitive data, and records all access and operation behaviors on the sensitive data.

[0014] Furthermore, the security and privacy protection module sets access control policies and authentication mechanisms to verify and authorize user identities, wherein:

[0015] Assign corresponding permissions based on the user's role. Users with different roles can only access data within their permission range.

[0016] Attribute-based access control dynamically determines the user's access rights based on the user's attributes, where the user attributes include geographic location and time;

[0017] Adopt a strong password policy and require users to set passwords that are complex and difficult to guess.

[0018] Furthermore, the personalized recommendation and prediction module inputs user data and product data into a decision tree model, and the decision tree model outputs user behavior patterns and product association rules. When establishing the decision tree model, 70% of the keyword association data is divided into a training set, and 30% of the keyword association data is divided into a test set. The G ini coefficient G of the keyword is calculated based on the keyword association data of the training set, and the G ini coefficient G of the keyword is set. k is the sum of categories, i is a category, i=0 means that the category is 0, P(i) is the probability that l abe l=i, the keywords are sorted in order of G in ascending order according to the G ini coefficient G, and the keyword corresponding to the smallest G ini coefficient G is used as the root node, the G ini coefficient of the remaining keywords after the root node is set is calculated, and the keyword corresponding to the minimum G ini coefficient among the remaining keywords is used as the internal node, and the calculation is repeated until the remaining keywords are 0, and the decision tree model to be tested is obtained, and the decision tree model to be tested is tested according to the keyword association data of the test set, and the test accuracy Q is compared with the preset test accuracy Q0, and the establishment result of the decision tree model is judged according to the comparison result, where:

[0019] When Q≥Q0, it is determined that the decision tree model to be tested is output as a decision tree model;

[0020] When Q<Q0, it is determined that the keyword association data of the expanded training set continues to train the decision tree model to be tested until the judgment condition that the test accuracy Q is greater than or equal to the preset test accuracy Q0 is met, and the decision tree model to be tested is output as a decision tree model.

[0021] Furthermore, the personalized recommendation and prediction module generates an algorithm based on the frequent item sets and association rules in the decision tree model, identifies the association relationship between products and calculates the support and confidence Determine the strength of the association rule, where:

[0022] When the confidence value is closer to 1, the probability that B will appear under the condition that A appears is higher, and the strength of the association rule is greater;

[0023] When the confidence value is closer to 0, the probability of B appearing under the condition that A appears is lower, and the strength of the association rule is weaker.

[0024] Furthermore, the supply chain management module analyzes historical sales data according to the decision tree model, monitors inventory data in real time, obtains sales forecast results, comprehensively analyzes sales forecast results and inventory data, compares inventory quantity K with preset inventory quantity K0, and determines whether the current inventory can meet future sales demand based on the comparison results, wherein:

[0025] When K ≥ K0, it is determined that the current inventory can meet future sales demand;

[0026] When K<K0, it is determined that the current inventory cannot meet future sales demand, and a replenishment request is automatically triggered, sending a replenishment order to the supplier or adjusting the production plan.

[0027] Furthermore, the commodity bidding module conducts statistical auctions of commodities that are determined to be in current inventory to meet future sales demand based on the inventory management results. The merchant sets the commodity starting price P0 and the auction period S0. The user fills in the purchase price P based on the commodity starting price P0, and selects the maximum purchase price Pmax within the auction period S0. The maximum purchase price Pmax is used as the final price of the commodity for sale.

[0028] Furthermore, the risk warning module determines risk assessment indicators based on analyzing the commodity purchase structure and combining the user purchase behavior pattern, inputs the risk assessment indicators into the transaction risk assessment model, and obtains the results of the transaction risk assessment model as a basis for monitoring abnormal behavior during the transaction process.

[0029] Furthermore, the risk warning module establishes a risk warning level system and sends a risk warning signal based on the detected abnormal behavior, wherein:

[0030] When the risk level is low-risk abnormal behavior, no risk warning signal is sent;

[0031] When the risk level is high-risk abnormal behavior, emergency measures are taken immediately and risk warning signals are sent.

[0032] Further, the optimization and improvement module records the number of false alarms W, the number of missed alarms L, the response time X and the number of correct warnings Z of the risk warning module, calculates the false alarm rate W0=false alarm number W / (false alarm number W+correct warning number Z)×100%, the missed alarm rate L0=missed alarm number L / (missed alarm number L+correct warning number Z)×100%, the average response time X1=the sum of all warning response times XS / warning number WZ and the warning hit rate M0=correct warning number Z / actual risk number R×100% of the risk warning module, and compares the false alarm rate W0, the missed alarm rate L0, the average response time X1 and the warning hit rate M0 of the risk warning module with the preset false alarm rate YW0, ​​the preset missed alarm rate YL0, the preset average response time YX1 and the preset warning hit rate YM0 respectively, and according to the comparison results, regularly improves and optimizes the transaction risk assessment model and the abnormal behavior detection algorithm, wherein:

[0033] When W0>YW0, adjust the transaction risk assessment model parameters;

[0034] When L0>YL0, a dynamic adjustment mechanism is introduced;

[0035] When X1>YX1, the transaction risk assessment model algorithm is improved;

[0036] When M0<YM0, optimize the transaction risk assessment model feature selection and adjust the transaction risk assessment model parameters.

[0037] Compared with the prior art, the beneficial effects of the present invention are that the platform is set up in various terminals, and through in-depth mining of user data and product data, it provides users with personalized and intelligent shopping experience, and helps merchants to achieve precise marketing and efficient operation; the platform collects user data and product data through a data collection and preprocessing module, and integrates, cleans and stores the collected user data and product data, so as to facilitate subsequent preprocessing based on the real-time collected data; the platform encrypts and stores all sensitive data based on the collected user data and product data through a security and privacy protection module, and sets access control policies and authentication mechanisms to verify and authorize user identities, so as to ensure the security of user data and product data; the platform extracts useful data information based on the collected user data and product data through a personalized recommendation and prediction module, uses a recommendation algorithm to make personalized recommendations for products, and predicts future user needs based on time series analysis, so as to ensure that each user can Get customized product recommendations and services; the platform predicts sales results based on the predicted future needs of users through the supply chain management module, determines whether the current inventory can meet future sales needs based on sales trends, and optimizes inventory management to reduce the risks of out-of-stock and excess inventory of merchants; the platform auctions and bids for scarce goods based on inventory management results through the product bidding module to achieve a win-win situation for users and merchants; the platform constructs a transaction risk assessment model based on the purchase structure of goods through the risk warning module, monitors abnormal behaviors in the transaction process in real time, and divides warning signals into different levels according to the severity of abnormal behaviors, so that relevant personnel can take timely measures; the platform regularly evaluates and summarizes the operation of the risk warning module through the optimization and improvement module, obtains evaluation results, and improves and optimizes the transaction risk assessment model and abnormal behavior detection algorithm based on the evaluation results, so as to improve the accuracy, timeliness and adaptability of the transaction risk assessment model and effectively prevent potential risks.

[0038] In particular, the security and privacy protection module encrypts sensitive data based on the collected user data and product data, sets access control policies and authentication mechanisms to verify and authorize user identities, and prevents user data and product data from being intercepted and tampered with, thereby ensuring the security and privacy of user data and product data.

[0039] In particular, the decision module inputs user data and product data into a decision tree model, and the decision tree model outputs user behavior patterns and product association rules to identify user interest preferences and purchase intentions to formulate targeted marketing strategies, thereby improving user experience and business operating efficiency.

[0040] In particular, the personalized recommendation and prediction module extracts useful data information based on the collected user data and product data, uses recommendation algorithms to make personalized recommendations for products, and predicts users' future needs based on time series analysis. Based on the data analysis results, it provides platform managers and merchants with intuitive reporting and visualization tools to help users make data-driven decisions quickly, so as to predict sales trends, identify potential market opportunities, and evaluate the effectiveness of marketing activities.

[0041] In particular, the supply chain management module optimizes inventory management based on predicted future user demand to ensure that the platform has a clear understanding of inventory status and can automatically trigger replenishment requests to reduce the risks of out-of-stock and excess inventory.

[0042] In particular, the commodity bidding module conducts auction bidding for commodities with scarce inventory according to the inventory management results, which can quickly attract the attention of many potential buyers and expand the exposure of commodities. Sellers can find buyers willing to pay a higher price for commodities in a shorter period of time. Buyers can understand the bids of other participants in real time during the bidding process and make adjustments according to their own needs and budgets, thereby improving sales efficiency.

[0043] In particular, the risk warning module constructs a transaction risk assessment model based on the commodity purchase structure, monitors abnormal behavior during the transaction process in real time, ensures the safety of platform operation and user use, and divides the warning signals into different levels according to the severity of the abnormal behavior so that relevant personnel can take timely measures.

[0044] In particular, the optimization and improvement module regularly evaluates and summarizes the operation of the risk warning module to obtain evaluation results, and based on the evaluation results, improves and optimizes the transaction risk assessment model and abnormal behavior detection algorithm, adjusts the feature selection, parameter setting, algorithm and dynamic adjustment mechanism of the transaction risk assessment model, and conducts simulation testing and continuous monitoring and optimization to continuously improve the accuracy, timeliness and adaptability of the transaction risk assessment model and effectively prevent potential risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is a schematic diagram of the structure of a multi-terminal transaction platform based on data mining in this embodiment. DETAILED DESCRIPTION

[0046] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0047] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0048] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the drawings. This is merely for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0049] In addition, it should be noted that in the description of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0050] See also Figure 1 As shown, it is a schematic diagram of the structure of a multi-terminal transaction platform based on data mining in this embodiment, and the platform includes:

[0051] The data collection and preprocessing module is used to collect user data and product data; it is also used to integrate, clean and store the collected user data and product data;

[0052] The security and privacy protection module is used to encrypt and store all sensitive data based on the collected user data and product data, and is also used to set access control policies and authentication mechanisms to verify and authorize user identities. The security and privacy protection module is connected to the data collection and preprocessing module;

[0053] The personalized recommendation and prediction module is used to extract useful data information based on the collected user data and product data, use the recommendation algorithm to make personalized recommendations for products, and predict future user needs based on time series analysis. The personalized recommendation and prediction module is connected to the security and privacy protection module;

[0054] Supply chain management module, which is used to predict sales results based on the predicted future needs of users, determine whether the current inventory can meet future sales needs based on sales trends, and optimize inventory management. The supply chain management module is connected with the personalized recommendation and prediction module;

[0055] The commodity bidding module is used to conduct auction bidding for commodities with scarce inventory according to the inventory management results. The commodity bidding module is connected with the supply chain management module;

[0056] The risk warning module is used to build a transaction risk assessment model based on the commodity purchase structure, monitor abnormal behaviors in the transaction process in real time, and classify warning signals into different levels according to the severity of abnormal behaviors. The risk warning module is connected with the security and privacy protection module, the supply chain management module and the commodity bidding module;

[0057] The optimization and improvement module is used to regularly evaluate and summarize the operation of the risk warning module to obtain the evaluation results. It is also used to improve and optimize the transaction risk assessment model and abnormal behavior detection algorithm based on the evaluation results, and to connect the optimization and improvement module with the risk warning module.

[0058] Specifically, the platform is set up in various terminals, and through in-depth mining of user data and product data, it provides users with personalized and intelligent shopping experience, and helps merchants achieve precise marketing and efficient operations; the platform collects user data and product data through a data collection and preprocessing module, and integrates, cleans and stores the collected user data and product data, so as to facilitate subsequent preprocessing based on the real-time collected data; the platform encrypts and stores all sensitive data based on the collected user data and product data through a security and privacy protection module, and sets access control policies and authentication mechanisms to verify and authorize user identities, so as to ensure the security of user data and product data; the platform extracts useful data information based on the collected user data and product data through a personalized recommendation and prediction module, uses a recommendation algorithm to make personalized recommendations for products, and predicts future user needs based on time series analysis, so as to ensure that each user can obtain customized Product recommendation and service; the platform predicts sales results based on the predicted future demand of users through the supply chain management module, determines whether the current inventory can meet future sales demand based on sales trends, and optimizes inventory management to reduce the risk of out-of-stock and excess inventory of merchants; the platform auctions scarce commodities based on inventory management results through the commodity bidding module to achieve a win-win situation for users and merchants; the platform constructs a transaction risk assessment model based on the commodity purchase structure through the risk warning module, monitors abnormal behaviors in the transaction process in real time, and divides warning signals into different levels according to the severity of abnormal behaviors, so that relevant personnel can take timely measures; the platform regularly evaluates and summarizes the operation of the risk warning module through the optimization and improvement module, obtains evaluation results, and improves and optimizes the transaction risk assessment model and abnormal behavior detection algorithm based on the evaluation results, so as to improve the accuracy, timeliness and adaptability of the transaction risk assessment model and effectively prevent potential risks.

[0059] Specifically, the user data includes basic user information and historical shopping information. The basic user information includes the user's age, gender and occupation. The historical shopping information includes the purchased goods, purchase time, purchase quantity and price. The product data includes basic product information and user evaluation information on the goods. The basic product information includes the product name, product category, product price and product specifications. The user evaluation information on the goods includes the score, evaluation content and comment time. The data integration refers to the platform's support for real-time collection of user data and product data from various data sources. The data cleaning refers to the process of reviewing and verifying user data and product data using data cleaning technology. The data storage refers to the use of distributed storage systems to achieve efficient storage and backup of massive data, while using MongoDB and PostgreSQL databases for data sharding and distributed deployment to improve data access efficiency.

[0060] It can be understood that the present embodiment does not limit the setting type of the data source, and those skilled in the art can freely set it according to the actual situation, and only need to meet the collection requirements of transaction data, such as setting the data source to an e-commerce platform, social media, and logistics system; the present embodiment does not limit the setting type of the data cleaning technology, and those skilled in the art can freely set it according to the actual situation, and only need to meet the cleaning of user data and product data, such as setting the data cleaning technology to missing value filling, noise removal, and data type conversion; the present embodiment does not limit the setting type of the distributed storage system, and those skilled in the art can freely set it according to the actual situation, and only need to meet the storage requirements of user data and product data, such as setting the distributed storage system to Ceph.

[0061] Specifically, the data collection and preprocessing module collects user data and product data, and performs data integration, data cleaning and data storage on the collected user data and product data to ensure the accuracy and consistency of the data to form a unified data warehouse.

[0062] Specifically, the security and privacy protection module encrypts and stores all sensitive data based on the collected user data and product data, desensitizes the sensitive data, and records all access and operation behaviors on the sensitive data.

[0063] Specifically, the sensitive data includes user personal information, payment information and commodity transaction information; the desensitization processing refers to the deformation, replacement, hiding or encryption of sensitive data. This embodiment does not limit the setting method of the desensitization processing. Technical personnel in this field can set it freely as long as the desensitization requirements for sensitive data are met. For example, the desensitization processing method can be set to hide some numbers and replace sensitive words.

[0064] Specifically, the security and privacy protection module sets access control policies and authentication mechanisms to verify and authorize user identities, where:

[0065] Assign corresponding permissions based on the user's role. Users with different roles can only access data within their permission range.

[0066] Attribute-based access control dynamically determines the user's access rights based on the user's attributes, where the user attributes include geographic location and time;

[0067] Adopt a strong password policy and require users to set passwords that are complex and difficult to guess.

[0068] Specifically, the security and privacy protection module encrypts sensitive data based on the collected user data and product data, sets access control policies and authentication mechanisms to verify and authorize user identities, prevents user data and product data from being intercepted and tampered with, and ensures the security and privacy of user data and product data.

[0069] Specifically, the personalized recommendation and prediction module inputs user data and product data into a decision tree model, and the decision tree model outputs user behavior patterns and product association rules. When establishing the decision tree model, 70% of the keyword association data is divided into a training set, and 30% of the keyword association data is divided into a test set. The G ini coefficient G of the keyword is calculated based on the keyword association data of the training set, and the G ini coefficient G of the keyword is set. k is the sum of categories, i is a category, i=0 means that the category is 0, P(i) is the probability that l abe l=i, the keywords are sorted in order of G in ascending order according to the G ini coefficient G, and the keyword corresponding to the smallest G ini coefficient G is used as the root node, the G ini coefficient of the remaining keywords after the root node is set is calculated, and the keyword corresponding to the minimum G ini coefficient among the remaining keywords is used as the internal node, and the calculation is repeated until the remaining keywords are 0, and the decision tree model to be tested is obtained, and the decision tree model to be tested is tested according to the keyword association data of the test set, and the test accuracy Q is compared with the preset test accuracy Q0, and the establishment result of the decision tree model is judged according to the comparison result, where:

[0070] When Q≥Q0, it is determined that the decision tree model to be tested is output as a decision tree model;

[0071] When Q<Q0, it is determined that the keyword association data of the expanded training set continues to train the decision tree model to be tested until the judgment condition that the test accuracy Q is greater than or equal to the preset test accuracy Q0 is met, and the decision tree model to be tested is output as a decision tree model.

[0072] Specifically, the keyword-related data includes user data and commodity data; the remaining keywords refer to keywords that are not set after being set as root nodes or internal nodes.

[0073] Specifically, the decision module inputs user data and product data into a decision tree model, and the decision tree model outputs user behavior patterns and product association rules to identify user interest preferences and purchase intentions to formulate targeted marketing strategies, thereby improving user experience and business operating efficiency.

[0074] Specifically, the personalized recommendation and prediction module generates an algorithm based on the frequent item sets and association rules in the decision tree model, identifies the association relationship between products and calculates the support and confidence Determine the strength of the association rule, where:

[0075] When the confidence value is closer to 1, the probability that B will appear under the condition that A appears is higher, and the strength of the association rule is greater;

[0076] When the confidence value is closer to 0, the probability of B appearing under the condition that A appears is lower, and the strength of the association rule is weaker.

[0077] Specifically, the support refers to the frequency of the association rule appearing in the data set, the confidence refers to the strength and credibility of the association rule, and the confidence value is between 0 and 1. The count(A∪B) represents the number of transaction records containing both item sets A and B, the count(T) represents the total number of transaction records, and the count(A) represents the number of transaction records containing item set A.

[0078] It can be understood that this embodiment does not limit the setting type of the generation algorithm. Those skilled in the art can freely set it according to actual conditions, as long as it meets the requirement of identifying the relationship between commodities. For example, the generation algorithm can be set to April algorithm.

[0079] Specifically, the personalized recommendation and prediction module extracts useful data information based on the collected user data and product data, uses the recommendation algorithm to make personalized recommendations for products, and predicts users' future needs based on time series analysis. Based on the data analysis results, it provides intuitive reporting and visualization tools for platform managers and merchants to help users make data-driven decisions quickly, so as to predict sales trends, identify potential market opportunities, and evaluate the effectiveness of marketing activities.

[0080] Specifically, the supply chain management module analyzes historical sales data according to the decision tree model, monitors inventory data in real time, obtains sales forecast results, comprehensively analyzes sales forecast results and inventory data, compares inventory quantity K with preset inventory quantity K0, and determines whether the current inventory can meet future sales demand based on the comparison results, where:

[0081] When K ≥ K0, it is determined that the current inventory can meet future sales demand;

[0082] When K<K0, it is determined that the current inventory cannot meet future sales demand, and a replenishment request is automatically triggered, sending a replenishment order to the supplier or adjusting the production plan.

[0083] Specifically, the historical sales data includes the quantity of goods sold, the types of goods sold and the time of goods sold; the monitored inventory data includes inventory quantity, inventory turnover rate and inventory cost; the preset inventory quantity refers to the preset value of the remaining quantity of goods. This embodiment does not limit the set quantity of the preset inventory quantity. Technical personnel in this field can freely set it according to actual conditions, as long as it meets the prediction of continuous supply of goods. For example, the preset inventory quantity can be set to 20.

[0084] Specifically, the supply chain management module optimizes inventory management based on predicted future user demand to ensure that the platform has a clear understanding of inventory status and can automatically trigger replenishment requests to reduce the risk of out-of-stock and excess inventory.

[0085] Specifically, the commodity bidding module will conduct statistical auctions on commodities that are determined to be in current inventory to meet future sales demand based on the inventory management results. The merchant will set the commodity starting price P0 and the auction period S0. The user will fill in the purchase price P based on the commodity starting price P0, and select the maximum purchase price Pmax within the auction period S0. The maximum purchase price Pmax will be used as the final price of the commodity for sale.

[0086] Specifically, the starting price of the commodity refers to the basic price set by the merchant for the auction of the commodity. This embodiment does not limit the setting price of the starting price of the commodity. Technical personnel in this field can freely set it according to actual conditions, and only need to meet the bidding requirements of the commodity auction, such as being set to no less than 1 yuan; the auction period refers to the latest deadline for the merchant to finally sell the auctioned commodity. This embodiment does not limit the setting time of the auction period. Technical personnel in this field can freely set it according to actual conditions, and only need to meet the selling requirements of the auctioned commodity, such as setting the auction period to 3 days; the purchase price refers to the price offered by the user for the auctioned commodity that is higher than the starting price. This embodiment does not limit the setting format of the purchase price. Technical personnel in this field can freely set it according to actual conditions, and only need to meet the bidding requirements that the purchase price is greater than the starting price, such as the purchase price format can be set to XX.XX yuan; the maximum purchase price refers to the maximum value among all prices filled in by users for the commodity.

[0087] Specifically, the commodity bidding module conducts auctions for commodities with scarce inventory based on inventory management results, which can quickly attract the attention of many potential buyers and expand the exposure of commodities. Sellers can find buyers who are willing to pay a higher price for commodities in a shorter period of time. Buyers can understand the bids of other participants in real time during the bidding process and make adjustments based on their own needs and budgets, thereby improving sales efficiency.

[0088] Specifically, the risk warning module determines risk assessment indicators based on analyzing the commodity purchase structure and combining user purchase behavior patterns, inputs the risk assessment indicators into the transaction risk assessment model, and obtains the results of the transaction risk assessment model as a basis for monitoring abnormal behavior during the transaction process.

[0089] Specifically, the commodity purchase structure includes the frequency of commodity purchases, the quantity of commodity purchases and the amount of commodity purchases; the user purchase behavior pattern includes the time, place and method of purchase; the risk assessment indicator refers to abnormal fluctuations in transaction amount, abnormal changes in purchase frequency and abnormal user behavior. The transaction risk assessment model is a neural network model that identifies whether there is abnormal behavior based on the thresholds and weights set in advance for the risk assessment indicators. This embodiment does not limit the construction method of the transaction risk assessment model. Those skilled in the art can freely set it according to actual conditions, as long as the judgment requirements for the risk assessment indicators are met. For example, the historical risk assessment indicators can be set as the transaction risk assessment model construction data, and the transaction risk assessment model construction data can be divided to 70% of the transaction risk assessment indicators. The data for constructing the transaction risk assessment model is used as the training set of the transaction risk assessment model, and 30% of the data for constructing the transaction risk assessment model is used as the test set of the transaction risk assessment model. The convolutional neural network model is trained according to the transaction risk assessment model training set, and the trained convolutional neural network model is tested according to the transaction risk assessment model test set. The trained convolutional neural network model with a test accuracy of 98% is output as the transaction risk assessment model, and the trained convolutional neural network model with a test accuracy that does not reach 98% is continued to be trained until the test accuracy reaches 98%, and then the trained convolutional neural network model is output as the transaction risk assessment model. The abnormal behavior refers to abnormal transaction amounts, frequent transaction cancellations, and multiple purchases of the same product by the same user in a short period of time.

[0090] Specifically, the risk warning module establishes a risk warning level system and sends a risk warning signal based on the detected abnormal behavior, wherein:

[0091] When the risk level is low-risk abnormal behavior, no risk warning signal is sent;

[0092] When the risk level is high-risk abnormal behavior, emergency measures are taken immediately and risk warning signals are sent.

[0093] Specifically, the risk warning signal refers to a warning signal issued after the potential risks faced by an enterprise, project or individual are identified through certain methods and means. The present embodiment does not limit the method for sending the risk warning signal. Technical personnel in this field can freely set it according to actual conditions, and only need to meet the requirements for sending the risk warning signal, such as setting the risk warning signal sending method to SMS, email or pop-up window; the present embodiment does not limit the types of high-risk abnormal behaviors. Technical personnel in this field can freely set it according to actual conditions, and only need to meet the risk requirements that can endanger the interests of others, such as setting high-risk abnormal behaviors to frequent transaction cancellations and the same user purchasing the same product multiple times in a short period of time; the present embodiment does not limit the types of emergency measures. Technical personnel in this field can freely set it according to actual conditions, and only need to suppress the occurrence of high-risk abnormal behaviors, such as setting emergency measures to suspend transactions and investigate and verify.

[0094] Specifically, the risk warning module builds a transaction risk assessment model based on the commodity purchase structure, monitors abnormal behavior during the transaction process in real time, ensures the safety of platform operation and user use, and divides the warning signals into different levels according to the severity of the abnormal behavior so that relevant personnel can take timely measures.

[0095] Specifically, the optimization and improvement module records the number of false alarms W, the number of missed alarms L, the response time X and the number of correct warnings Z of the risk warning module, calculates the false alarm rate W0=false alarms W / (false alarms W+correct warnings Z)×100%, the missed alarm rate L0=missed alarms L / (missed alarms L+correct warnings Z)×100%, the average response time X1=the sum of all warning response times XS / warning times WZ and the warning hit rate M0=correct warning times Z / actual risk times R×100% of the risk warning module, and compares the false alarm rate W0, missed alarm rate L0, average response time X1 and warning hit rate M0 of the risk warning module with the preset false alarm rate YW0, ​​preset missed alarm rate YL0, preset average response time YX1 and preset warning hit rate YM0 respectively, and according to the comparison results, regularly improves and optimizes the transaction risk assessment model and abnormal behavior detection algorithm, wherein:

[0096] When W0>YW0, adjust the transaction risk assessment model parameters;

[0097] When L0>YL0, a dynamic adjustment mechanism is introduced;

[0098] When X1>YX1, the transaction risk assessment model algorithm is improved;

[0099] When M0<YM0, optimize the transaction risk assessment model feature selection and adjust the transaction risk assessment model parameters.

[0100] Specifically, the false alarm rate refers to the probability of misjudging a normal situation as a risk situation and issuing a warning; the false alarm rate refers to the probability of a risk situation actually existing but no warning being issued; the average response time is the average value of all time intervals from the occurrence of a risk event to the issuance of a warning signal; the warning hit rate refers to the probability that the risk warning module accurately identifies the actual risk; the preset false alarm rate refers to the preset value of the probability of misjudging a normal situation as a risk situation and issuing a warning; the preset false alarm rate refers to the preset value of the probability of a risk situation actually existing but no warning being issued; the preset average response time is the preset value of the average value of all time intervals from the occurrence of a risk event to the issuance of a warning signal; the preset warning hit rate refers to the preset value of the probability that the risk warning module accurately identifies the actual risk; the risk assessment model parameters include thresholds and weights.

[0101] It can be understood that the present embodiment does not limit the periodic setting duration, and those skilled in the art can freely set it according to actual conditions, and only need to meet the regularity of evaluating and optimizing the risk warning module, such as setting the periodic setting duration to weekly; the present embodiment does not limit the adjustment method of the risk assessment model parameters, and those skilled in the art can freely set it according to actual conditions, and only need to meet the adjustment requirements of the risk assessment model parameters, such as setting the adjustment method of the risk assessment model parameters to increase the threshold; the present embodiment does not limit the type of the introduced mechanism, and those skilled in the art can freely set it according to actual conditions, and only need to meet the introduction requirements of the dynamic mechanism, such as setting the introduction of the dynamic adjustment mechanism to use an online learning algorithm; the present embodiment does not limit the improved algorithm of the transaction risk assessment model, and those skilled in the art can freely set it according to actual conditions, and only need to meet the improvement requirements of the risk assessment model, such as setting the improved algorithm of the risk assessment model to a deep learning algorithm; the present embodiment does not limit the optimization method of the feature selection of the transaction risk assessment model, and those skilled in the art can freely set it according to actual conditions, and only need to meet the optimization requirements of the feature selection of the transaction risk assessment model, such as setting the optimization method of the feature selection of the transaction risk assessment model to propose feature variables with weak correlation with transaction risks.

[0102] Specifically, the optimization and improvement module regularly evaluates and summarizes the operation of the risk warning module to obtain evaluation results, and based on the evaluation results, improves and optimizes the transaction risk assessment model and abnormal behavior detection algorithm, adjusts the feature selection, parameter setting, algorithm and dynamic adjustment mechanism of the transaction risk assessment model, and conducts simulation testing and continuous monitoring and optimization to continuously improve the accuracy, timeliness and adaptability of the transaction risk assessment model and effectively prevent potential risks.

[0103] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

Claims

1. A multi-terminal trading platform based on data mining, characterized in that: The platform includes: The data collection and preprocessing module is used to collect user data and product data; it is also used to integrate, clean and store the collected user data and product data; The security and privacy protection module is used to encrypt and store all sensitive data based on the collected user data and product data, and to set access control policies and authentication mechanisms to verify and authorize user identities; Personalized recommendation and prediction module, which is used to extract useful data information based on the collected user data and product data, use recommendation algorithms to make personalized recommendations for products, and predict future user needs based on time series analysis; Supply chain management module, which is used to predict sales results based on the predicted future demand of users, determine whether the current inventory can meet future sales demand based on sales trends, and optimize inventory management; The commodity bidding module is used to conduct auction bidding for commodities with scarce inventory according to the inventory management results; The risk warning module is used to build a transaction risk assessment model based on the commodity purchase structure, monitor abnormal behaviors in the transaction process in real time, and classify warning signals into different levels according to the severity of abnormal behaviors; The optimization and improvement module is used to regularly evaluate and summarize the operation of the risk warning module to obtain evaluation results, and is also used to improve and optimize the transaction risk assessment model and abnormal behavior detection algorithm based on the evaluation results.

2. The multi-terminal transaction platform based on data mining according to claim 1, characterized in that: The security and privacy protection module encrypts and stores all sensitive data based on the collected user data and product data, desensitizes the sensitive data, and records all access and operation behaviors on the sensitive data.

3. The multi-terminal transaction platform based on data mining according to claim 2, characterized in that: The security and privacy protection module sets access control policies and authentication mechanisms to verify and authorize user identities, including: Assign corresponding permissions based on the user's role. Users with different roles can only access data within their permission range. Attribute-based access control dynamically determines the user's access rights based on the user's attributes, where the user attributes include geographic location and time; Adopt a strong password policy and require users to set passwords that are complex and difficult to guess.

4. The multi-terminal transaction platform based on data mining according to claim 3, characterized in that: The personalized recommendation and prediction module inputs user data and product data into a decision tree model, and the decision tree model outputs user behavior patterns and product association rules. When establishing the decision tree model, 70% of the keyword association data is divided into a training set, and 30% of the keyword association data is divided into a test set. The Gini coefficient G of the keyword is calculated based on the keyword association data of the training set, and the Gini coefficient G of the keyword is set. k is the sum of categories, i is a category, i=0 means that the category is 0, P(i) is the probability of label=i, the keywords are sorted in order of Gini coefficient G from small to large, and the keyword corresponding to the smallest Gini coefficient G is used as the root node, the Gini coefficient of the remaining keywords after the root node is set is calculated, and the keyword corresponding to the minimum Gini coefficient of the remaining keywords is used as the internal node, and the calculation is repeated until the remaining keywords are 0, and the decision tree model to be tested is obtained. The decision tree model to be tested is tested according to the keyword association data of the test set, and the test accuracy Q is compared with the preset test accuracy Q0. The establishment result of the decision tree model is judged according to the comparison result, where: When Q≥Q0, it is determined that the decision tree model to be tested is output as a decision tree model; When Q<Q0, it is determined that the keyword association data of the expanded training set continues to train the decision tree model to be tested until the judgment condition that the test accuracy Q is greater than or equal to the preset test accuracy Q0 is met, and the decision tree model to be tested is output as a decision tree model.

5. The multi-terminal transaction platform based on data mining according to claim 4, characterized in that: The personalized recommendation and prediction module generates an algorithm based on the frequent item sets and association rules in the decision tree model, identifies the association relationship between products and calculates the support and confidence Determine the strength of the association rule, where: When the confidence value is closer to 1, the probability of B appearing under the condition that A appears is higher, and the strength of the association rule is greater; When the confidence value is closer to 0, the probability of B appearing under the condition that A appears is lower, and the strength of the association rule is weaker.

6. The multi-terminal transaction platform based on data mining according to claim 5, characterized in that: The supply chain management module analyzes historical sales data according to the decision tree model, monitors inventory data in real time, obtains sales forecast results, comprehensively analyzes sales forecast results and inventory data, compares inventory quantity K with preset inventory quantity K0, and determines whether the current inventory can meet future sales demand based on the comparison results, wherein: When K ≥ K0, it is determined that the current inventory can meet future sales demand; When K<K0, it is determined that the current inventory cannot meet future sales demand, and a replenishment request is automatically triggered, sending a replenishment order to the supplier or adjusting the production plan.

7. The multi-terminal transaction platform based on data mining according to claim 6, characterized in that: The commodity bidding module conducts statistical auctions on commodities that are determined to be in current inventory to meet future sales demand based on the inventory management results. The merchant sets the commodity starting price P0 and the auction period S0. The user fills in the purchase price P based on the commodity starting price P0 and selects the maximum purchase price Pmax within the auction period S0. The maximum purchase price Pmax is used as the final price of the commodity for sale.

8. The multi-terminal transaction platform based on data mining according to claim 7, characterized in that: The risk warning module determines risk assessment indicators based on analyzing the commodity purchase structure and combining the user's purchase behavior pattern, inputs the risk assessment indicators into the transaction risk assessment model, and obtains the results of the transaction risk assessment model as a basis for monitoring abnormal behavior during the transaction process.

9. The multi-terminal transaction platform based on data mining according to claim 8, characterized in that: The risk warning module establishes a risk warning level system and sends a risk warning signal based on the detected abnormal behavior, wherein: When the risk level is low-risk abnormal behavior, no risk warning signal is sent; When the risk level is high-risk abnormal behavior, emergency measures are taken immediately and risk warning signals are sent.

10. The multi-terminal transaction platform based on data mining according to claim 9, characterized in that: The optimization and improvement module records the number of false alarms W, the number of missed alarms L, the response time X and the number of correct warnings Z of the risk warning module, calculates the false alarm rate W0=false alarm number W / (false alarm number W+correct warning number Z)×100%, the missed alarm rate L0=missed alarm number L / (missed alarm number L+correct warning number Z)×100%, the average response time X1=the sum of all warning response times XS / warning number WZ and the warning hit rate M0=correct warning number Z / actual risk number R×100% of the risk warning module, and compares the false alarm rate W0, the missed alarm rate L0, the average response time X1 and the warning hit rate M0 of the risk warning module with the preset false alarm rate YW0, ​​the preset missed alarm rate YL0, the preset average response time YX1 and the preset warning hit rate YM0 respectively, and according to the comparison results, regularly improves and optimizes the transaction risk assessment model and the abnormal behavior detection algorithm, wherein: When W0>YW0, adjust the transaction risk assessment model parameters; When L0>YL0, a dynamic adjustment mechanism is introduced; When X1>YX1, the transaction risk assessment model algorithm is improved; When M0<YM0, optimize the transaction risk assessment model feature selection and adjust the transaction risk assessment model parameters.

Images