E-commerce platform numerical value splitting and price interval dynamic sorting and returning system

Abstract

The invention discloses an e-commerce platform numerical value splitting and price interval dynamic sorting and returning system, and relates to the technical field of data processing and analysis. Comprising a back-end data storage and management module, a system logic and algorithm execution module, a back-end service architecture module, a front-end display module, a data statistics and analysis module, a safety and stability guarantee module and a monitoring and feedback mechanism module. The system logic and algorithm execution module splits transaction order prices based on anchor point values, and realizes price interval dynamic sorting and rebate return through accumulative storage, fund pool management and asynchronous scheduling; and the system logic and algorithm execution module comprises a price splitting sub-module, an accumulative storage sub-module, a fund pool management sub-module and a rebate triggering and execution sub-module. A system logic and algorithm execution module is constructed, and the problems that an existing e-commerce platform is extensive in price processing, simple in fund return logic, lack of flexibility in a rebating mechanism, easy to copy a business mode and insufficient in user stickiness can be solved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing and analysis, and in particular to a system for splitting numerical values ​​and dynamically sorting and returning price ranges on an e-commerce platform. Background Art

[0002] The traditional e-commerce platform return system cannot subdivide price ranges according to product characteristics, and cannot accurately adapt to the sales patterns and cost structures of products in different price ranges, resulting in difficulty in achieving efficient resource allocation in fund processing and rebate settings. In addition, due to the complex amount-based triggered return model, it is easily affected by price fluctuations and decimal calculation accuracy, increasing financial risks, heavy computational burden, low processing efficiency, and making it difficult for users to understand the rebate rules, affecting user experience. The e-commerce platform return system of this application can solve the problems of existing e-commerce platforms' extensive price processing, simple fund return logic, lack of flexibility in rebate mechanism, easy-to-copy business model, and insufficient user stickiness.

[0003] The defects of the existing e-commerce platform return system are:

[0004] 1. Patent document CN108537382B discloses a method and system for predicting e-commerce price trends. This document primarily addresses the issue of manual investigation and complex data analysis required to determine price trends when e-commerce product prices change, resulting in high labor costs and hindering automated price adjustments. It does not address the issues of crude price processing, simple fund return logic, inflexible rebate mechanisms, easily replicable business models, and insufficient user engagement on existing e-commerce platforms.

[0005] 2. Patent document CN113689263B discloses a blockchain-based e-commerce historical price query system. This document primarily addresses the issues of snapshots failing to preserve every price change and the platform deleting or tampering with snapshot data. It fails to address the issues of data storage being easily lost, slow to query, insecure, and difficult to retrieve unstructured data on existing e-commerce platforms.

[0006] 3. Patent document CN113869987B discloses a computer system for monitoring commodity prices on e-commerce platforms. This document primarily addresses the issue of improving the efficiency and accuracy of commodity price monitoring on e-commerce platforms, but does not address the issues of existing e-commerce platform architectures being unable to cope with high concurrent traffic, slow service response times, untimely data processing, and poor service communication security.

[0007] 4. Patent document CN113313531B discloses an e-commerce trial evaluation and recommendation system based on user needs. The document mainly considers selecting product trial users through the Internet, and then mailing the products to the trial users. The trial users fill out the trial report according to the pre-set answer sheet to obtain product trial data. There is uncertainty in the entire process, and the output evaluation results may not be consistent with the actual situation. It does not consider how to solve the problems of slow front-end page loading, single interactive form and non-intuitive data display of existing e-commerce platforms, which are difficult to meet the diverse needs of users. Summary of the Invention

[0008] The purpose of the present invention is to provide an e-commerce platform numerical value splitting and price range dynamic sorting and return system to solve the problems raised in the above background technology.

[0009] To achieve the above objectives, the present invention provides the following technical solutions: a system for numerical value splitting and price range dynamic sorting and return on an e-commerce platform, comprising a back-end data storage and management module, a system logic and algorithm execution module, a back-end service architecture module, a front-end display module, a data statistics and analysis module, a security and stability assurance module, and a monitoring and feedback mechanism module;

[0010] The backend data storage and management module uses multiple database types and blockchain technology to achieve efficient storage, integrity assurance, and semantic retrieval of data such as product prices and orders;

[0011] The system logic and algorithm execution module splits the executed order price based on the anchor value, and realizes dynamic sorting of price ranges and rebate returns through cumulative storage, fund pool management and asynchronous scheduling;

[0012] The system logic and algorithm execution module includes the price splitting submodule, the accumulation storage submodule, the fund pool management submodule, and the rebate triggering and execution submodule;

[0013] The price splitting submodule splits the order transaction price using 10, 30, 50, and 100 as anchor points. It first allocates the price to the price intervals of 10-30, 30-50, 50-100, and >100. It then performs a subtraction operation on the maximum value of the price interval, removing the remainder and retaining the integer part. If the split value still falls within another interval, it is assigned to the corresponding interval for further splitting, with the minimum splitting number being 10.

[0014] The accumulation storage submodule is used to store the integer part obtained after the price splitting submodule is processed into the integer accumulation pool of the corresponding price range in the order of transaction time;

[0015] The fund pool management submodule sets up a fractional fund pool and an integer fund pool. The integer fund pool is divided into a static pool accounting for 30% and a dynamic pool accounting for 70%. The static pool is used for fixed rules, and the dynamic pool is used for recommendation and repurchase weights.

[0016] The rebate triggering and execution submodule uses asynchronous scheduling rules to manage the fund pool, stipulating that rebates are only paid from the integer fund pool. When the cumulative number of integers in the price range meets 5 or 6 orders, rebates are calculated based on retained earnings. Only one triggering return condition is selected in the same range. According to the first-in-first-out principle, every time 5 or 6 orders trigger the return condition, the first order is returned. The return is made in sequence, and the cycle continues, returning funds from the integer fund pool.

[0017] Preferably, the backend service architecture module adopts microservices, message queues and real-time data processing technologies, combined with machine learning algorithms to achieve traffic control and data intelligent analysis;

[0018] The front-end display module is based on the Vue3 technology stack and integrates mixed reality, dynamic tables, and intelligent loading technologies to achieve price and rebate visualization and real-time interaction with user data;

[0019] The data statistics and analysis module uses time series analysis and clustering algorithms to deeply explore the correlation between rebate trends and user behavior, supporting personalized recommendations and strategy optimization;

[0020] The security and stability assurance module ensures high system availability and data security through scheduled backup, circuit breaker and current limiting, blockchain evidence storage, and search engine optimization;

[0021] The monitoring and feedback mechanism module builds a Prometheus+Grafana monitoring system, combining multi-channel feedback and intelligent traceability to achieve rebate progress tracking and multi-dimensional user queries.

[0022] Preferably, the backend data storage and management module includes a commodity price data storage submodule, a user order data recording submodule, a table structure design submodule and an index strategy submodule;

[0023] The commodity price data storage submodule uses the time series database InfluxDB to store historical price change data and uses time series indexes to improve query efficiency;

[0024] The user order data recording submodule uses the Apache Cassandra database and adopts a dual partitioning strategy based on user ID and order time to store data;

[0025] The table structure design submodule adds a blockchain hash field to the transaction information table, implements tamper-proof data storage through smart contracts, and introduces a layered blockchain evidence storage system to ensure data integrity through a three-level structure, including transaction layer, summary layer, and verification layer.

[0026] The transaction layer adds a new blockchain_hash field in the relational database;

[0027] The summary layer writes the transaction hash value to the Hyperledger Fabric consortium chain every day;

[0028] The verification layer provides a real-time verification interface through smart contracts;

[0029] The index strategy submodule introduces vector indexing technology to establish indexes for unstructured data and support semantic retrieval.

[0030] Preferably, the backend service architecture module includes a microservice architecture submodule, a real-time data processing submodule and an offline data analysis submodule;

[0031] The microservice architecture submodule uses Istio service mesh technology to achieve refined control of traffic between services and secure communication, and implements service circuit breaking, degradation, and retry mechanisms through the Envoy proxy;

[0032] The message queue submodule uses the Pulsar message queue and introduces a content-based routing strategy to route different types of transaction data to corresponding processing channels.

[0033] The batch processing submodule uses Flink-CDC technology to capture data changes in real time, combines it with the Doris real-time data warehouse for data aggregation and analysis, and uses the random forest algorithm to predict and sort price ranges and time sequences.

[0034] Preferably, the front-end display module includes a product price and rebate rule integrated display sub-module, a dynamic data interaction sub-module, a paging and intelligent loading sub-module, and a personal account data dashboard sub-module;

[0035] The integrated display submodule of product prices and rebate rules uses mixed reality technology to achieve simultaneous display of 3D visualization of product prices and simulation of rebate rules on mobile devices.

[0036] The dynamic data interaction submodule integrates the original dynamic table and chart display functions, and uses VantTable and EChartsGL to implement data table filtering, sorting and 3D chart analysis;

[0037] The paging and smart loading submodule uses infinite scrolling loading combined with a priority queue strategy to optimize data loading efficiency through a global data cache pool, ensuring the coordination of real-time data updates and paging loading.

[0038] The personal account data dashboard sub-module associates the transaction information table with the blockchain hash value based on the user IP, and displays in real time the price range corresponding to the integer part of each order after splitting, as well as the arrangement order of the integer in the corresponding interval queue. It is updated in real time using dynamic table technology.

[0039] Preferably, the data statistics and analysis module includes a rebate statistics submodule and a user behavior analysis submodule;

[0040] At 2:00 a.m. every Monday, the rebate statistics submodule uses Python's Pandas, Numpy, and Matplotlib libraries to conduct in-depth statistical analysis of the previous week's rebate data. The analysis indicators include the number of rebates in different price ranges, the total rebate amount, the number of benefited users, the average amount of a single rebate, and the standard deviation of the rebate amount. The SARIMA time series analysis method is used to predict the rebate trend in each price range for the next four weeks. The analysis results are displayed in the form of a visual dashboard. The dashboard includes a bar chart showing the number of rebates and the number of benefited users in different price ranges, a line chart showing the time trend of the total rebate amount, and a predicted curve of the rebate trend in the next four weeks. Users can view data and switch between different price ranges for comparative analysis by clicking on the chart elements.

[0041] Preferably, the user behavior analysis submodule combines user purchase order data and rebate participation, integrates the FP-Growth algorithm and the DBSCAN algorithm, uses the FP-Growth algorithm to mine the association rules of user purchases, sets the minimum support to 0.05 and the minimum confidence to 0.6, finds the associations between products in different price ranges, inputs these association rules as features into the DBSCAN algorithm, uses the elbow method to determine the number of clusters to be 5, performs cluster analysis on users, and formulates personalized product recommendation strategies and rebate mechanisms for each user group based on the clustering results, monitors changes in user behavior in real time, and the monitoring indicators include user purchase frequency, price range distribution of purchased products, and number of rebate participations. The monitoring frequency is once a day, and the clustering results and recommendation strategies are updated every two weeks.

[0042] Preferably, the security and stability assurance module includes a data backup and recovery submodule, a concurrent processing and performance optimization submodule, a search engine technology submodule, and a blockchain evidence storage submodule;

[0043] The data backup and recovery submodule uses MySQL's mysqldump command to perform a full database backup at 3:00 a.m. every day and stores the backup file in an off-site cloud storage service. An incremental backup is performed every Saturday at 4:00 a.m. At the same time, a data disaster recovery plan is formulated. Within one hour after a database failure, data is restored using the most recent full and incremental backup files. In the event of a disaster in an off-site data center, service is switched to the backup data center within three hours. During each full and incremental backup process, the backup speed and completion percentage are monitored in real time. If the backup speed drops below 1MB / s for five consecutive minutes, an alarm is sent to notify the operation and maintenance personnel.

[0044] The concurrent processing and performance optimization submodule uses the Spring Cloud distributed architecture combined with service grid technology to implement system microservices and traffic management. It uses the Druid database connection pool, setting the maximum number of connections to 200 and the minimum number of idle connections to 20.

[0045] Use Prometheus and Grafana to build a system performance monitoring platform to monitor the system's concurrent request number, response time, and database connection number indicators in real time. When the number of concurrent requests exceeds 1000 times / second, the circuit breaker and current limiting mechanism are automatically triggered. The current limiting ratio is 50%, that is, only 50% of the requests are allowed to pass, and the circuit breaker recovery time is 5 minutes.

[0046] Preferably, when processing query and sorting scenarios involving more than 10,000 records, the search engine technology submodule enables Elasticsearch search engine technology and configures 3 nodes for the Elasticsearch cluster, each equipped with an 8-core CPU, 16GB of memory, and a 500GB hard drive;

[0047] Check the Elasticsearch index weekly. When the index fragmentation rate exceeds 30% or the index space usage exceeds 80%, optimize and rebuild the index.

[0048] The blockchain evidence storage submodule adopts a hybrid on-chain and off-chain architecture. The original system calculation is retained off-chain to ensure high concurrent processing efficiency. On-chain, the split integer amount, queue position and rebate trigger time data are written into the private blockchain through smart contracts, generating an unalterable transaction hash and constructing a five-sequence rebate chain to record split records, queue status, rebate execution, behavior weight and fund pool dynamic information.

[0049] Preferably, the monitoring and feedback mechanism module includes a system monitoring submodule, a user feedback processing submodule, a rebate progress tracking submodule and a user query submodule;

[0050] The system monitoring submodule is used to deploy Prometheus and combine it with Grafana to build a monitoring system, set specific monitoring indicators for different links, and synchronize data to the intelligent problem tracing system;

[0051] The user feedback processing submodule is used to establish a multi-channel feedback system, including online customer service, a real-time open user forum, and a 24-hour complaint mailbox. It also introduces intelligent problem tracing capabilities, automatically correlating system monitoring data from the same period to generate analysis reports to assist in problem location.

[0052] The rebate progress tracking submodule queries the integer accumulation pool data of the accumulation storage submodule through the shared Redis cache. Combined with the asynchronous scheduling rules of the rebate trigger and execution submodule, it calculates the matching status of each rebate trigger condition and displays the current status of each condition in parallel on the interface. At the same time, it displays the progress of the integer accumulation amount in the current price range relative to the target value as a percentage.

[0053] The user query submodule provides three query methods: transaction hash, mobile phone number, and blockchain address. Users can enter the blockchain hash associated with the order number to view the complete transaction split and return link, aggregate the progress of rebate orders in progress by linking the mobile phone number to the account, and query the on-chain fund flow record based on the blockchain address.

[0054] The display content includes the queue position and estimated rebate time calculated by combining historical queue digestion speed with progress prediction algorithms such as exponential smoothing, the real-time behavior weight coefficient and the equivalent queue position after acceleration, and the real-time balances of the static pool and dynamic pool presented in dynamic charts, as well as the proportion of the daily distribution amount.

[0055] Compared with the prior art, the present invention has the following beneficial effects:

[0056] 1. The present invention constructs a system logic and algorithm execution module. The price splitting submodule uses 10, 30, 50 and 100 as anchor points to perform fine splitting of order transaction prices. Through cumulative storage, fund pool management and asynchronous scheduling, dynamic sorting of price ranges and rebate returns are realized. The rebate is only paid from the integer fund pool. When the integer cumulative number of the price range meets 5 or 6 orders, the selection of 5 or 6 orders is calculated and allocated based on retained profits. Only one triggering return condition is selected in the same range. According to the first-in-first-out principle, every time 5 or 6 orders trigger the return condition, the first order is returned, and the return is returned in sequence. Going forward, funds will be returned from the integer fund pool. This refined numerical splitting and dynamic rebate mechanism has changed the situation of the existing e-commerce platform's rough price processing and simple fund return logic, making the rebate mechanism more flexible and not easy to be simply copied. At the same time, through dynamic sorting and personalized rebate methods, it can effectively improve user participation and stickiness. Therefore, it can solve the problems of the existing e-commerce platform's rough price processing, simple fund return logic, lack of flexibility in the rebate mechanism, easy copying of the business model and insufficient user stickiness, and comprehensively enhance the comprehensive strength of the e-commerce platform in fund management, user experience and market competition.

[0057] 2. The present invention constructs a back-end data storage and management module that adopts multiple database types and blockchain technology. The back-end data storage and management module uses the time series database InfluxDB to store historical price change data to improve query efficiency, and utilizes the Apache Cassandra database to store user order data based on a dual partitioning strategy of user ID and order time. By adding a blockchain hash field to the transaction information table and introducing a layered blockchain evidence storage system to ensure data immutability and integrity, the present invention also introduces vector indexing technology to support semantic retrieval of unstructured data. Therefore, it can solve the problems of easy data loss, slow query, low security, and difficult unstructured data retrieval on existing e-commerce platforms.

[0058] 3. The present invention constructs a back-end service architecture module, which adopts microservices, message queues and real-time data processing technologies, combined with machine learning algorithms. The microservice architecture sub-module uses Istio service grid technology to achieve refined control of inter-service traffic and secure communication, and implements circuit breaking, degradation and retry mechanisms through Envoy proxy. The message queue sub-module uses Pulsar message queue and adopts content-based routing strategy. The batch processing sub-module adopts Flink-CDC technology to capture data changes in real time and combines Doris real-time data warehouse and random forest algorithm to perform data aggregation analysis and predictive sorting. The application of this series of technologies can effectively cope with high concurrent traffic, improve service response speed, realize timely data processing and intelligent analysis, and ensure the security of service communication. Therefore, it can solve the problems that the existing e-commerce platform architecture is difficult to cope with high concurrent traffic, slow service response, untimely data processing and poor service communication security.

[0059] 4. The present invention constructs a front-end display module, which is based on the Vue3 technology stack and integrates mixed reality, dynamic tables and intelligent loading technology. The integrated display sub-module of commodity prices and rebate rules adopts mixed reality technology to realize the three-dimensional visualization of commodity prices on mobile terminals and the simultaneous display of rebate rule simulation. The dynamic data interaction sub-module uses VantTable and EChartsGL to realize data table screening, sorting and three-dimensional chart analysis. The paging and intelligent loading sub-module adopts infinite scrolling loading combined with priority queue strategy and optimizes efficiency through the global data cache pool. The personal account data dashboard sub-module displays order splitting and queue arrangement order in real time and updates dynamically. Therefore, it can solve the problems of slow front-end page loading, single interaction form and non-intuitive data display of existing e-commerce platforms, which are difficult to meet the diverse needs of users. DETAILED DESCRIPTION

[0060] The technical solutions in the embodiments of the present invention are clearly and completely described. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0061] Example 1:

[0062] A system for numerical value splitting and price range dynamic sorting and return on e-commerce platforms, including a back-end data storage and management module, a system logic and algorithm execution module, a back-end service architecture module, a front-end display module, a data statistics and analysis module, a security and stability assurance module, and a monitoring and feedback mechanism module;

[0063] The system logic and algorithm execution module splits the executed order price based on the anchor value, and realizes dynamic sorting of price ranges and rebate returns through cumulative storage, fund pool management and asynchronous scheduling;

[0064] The system logic and algorithm execution module includes the price splitting submodule, the accumulation storage submodule, the fund pool management submodule, and the rebate triggering and execution submodule;

[0065] The price splitting submodule splits the order transaction price using 10, 30, 50, and 100 as anchor points. It first allocates the price to the price intervals of 10-30, 30-50, 50-100, and >100. It then performs a subtraction operation on the maximum value of the price interval, removing the remainder and retaining the integer part. If the split value still falls within another interval, it is assigned to the corresponding interval for further splitting, with the minimum splitting number being 10.

[0066] The accumulation storage submodule is used to store the integer part obtained after the price splitting submodule is processed into the integer accumulation pool of the corresponding price range in the order of transaction time;

[0067] The fund pool management submodule sets up a fractional fund pool and an integer fund pool. The integer fund pool is divided into a static pool accounting for 30% and a dynamic pool accounting for 70%. The static pool is used for fixed rules, and the dynamic pool is used for recommendation and repurchase weights.

[0068] The rebate triggering and execution submodule uses asynchronous scheduling rules to manage the fund pool, stipulating that rebates are only paid out from the integer fund pool. When the cumulative number of integers in the price range meets 5 or 6 orders, according to the first-in-first-out principle, every time 5 or 6 orders are generated to trigger the rebate conditions, the first order will be rebated, and the rebates will be returned in sequence. This cycle will continue, and funds will be returned from the integer fund pool.

[0069] Furthermore, if 1 order is returned for every 5 orders generated, the following situation will occur:

[0070]

[0071]

[0072] When the accumulated order reaches the 5th, the 0th order will be returned; when the accumulated order reaches the 10th, the 1st order will be returned; when the accumulated order reaches the 15th, the 2nd order will be returned, and so on;

[0073] The situation where 1 order is returned for every 6 orders generated is:

[0074]

[0075]

[0076] When the accumulated order reaches the 6th, the 0th order will be returned; when the accumulated order reaches the 12th, the 1st order will be returned; when the accumulated order reaches the 18th, the 2nd order will be returned, and so on;

[0077] For every n orders generated and 1 order returned, the formula for the returned order number k is:

[0078]

[0079] Where i is the current accumulated number of orders, n is the accumulated number of orders that trigger the return condition, Indicates rounding down;

[0080] When n=5:

[0081] i=5: Return order 0;

[0082] i=10: Return the first order;

[0083] i=15: Return the second order;

[0084] When n=6:

[0085] i=6: Return order 0;

[0086] i=12: Return the first order;

[0087] i=18: Return the 2nd order.

[0088] When the cumulative number of integer orders within a certain price range reaches i (i=5 or 6), a rebate is triggered. Each time the rebate is triggered, the earliest unrefunded order in the price range is refunded, excluding the orders in the currently triggered i orders. The rebate amount is only paid out of the integer fund pool to ensure the independence of fund management.

[0089] When n orders are accumulated, i=i. At this time, the n orders include the latest nth order. The system needs to wait for the next order to be generated, that is, i=i+1, to ensure that the i order at the time of triggering is a historical order, rather than the order currently being processed.

[0090] The rebate will not be triggered when i=n. If it is triggered when i=n, ​​the latest n-th order will be included in the historical orders, which may cause data consistency issues. Waiting for i=n+1 to trigger ensures that all n orders are completed historical orders.

[0091] The price splitting submodule splits the order transaction price using 10, 30, 50, and 100 as anchor points. For example, for an order price of 86.3 yuan, which is between 50 and 100, 50 is first subtracted to get 36.3 yuan. 36.3 yuan is between 30 and 50, so 30 is subtracted to get 6.3 yuan. After removing the remaining 6.3 yuan, the final price is split into two integers, 50 and 30. Another example is 27.5 yuan, which is between 10 and 30. It is split into two 10s, and the remaining 7.5 yuan is removed. This splitting method accurately maps prices to different price ranges, facilitating subsequent statistics and rebate operations.

[0092] The accumulation storage submodule stores the integer parts obtained after processing by the price splitting submodule into the integer accumulation pool of the corresponding price range in the order of transaction time. For example, the split 50 and 30 will be stored in the integer accumulation pools of the 50-100 and 30-50 ranges respectively, and are arranged according to the first-in-first-out principle to ensure the orderliness of transaction records.

[0093] First-in-first-out (FIFO) is a scheduling strategy based on time sequence, meaning that tasks or data that enter the system first are processed first. In the e-commerce rebate scenario, its core logic is: user orders are split by price and enter the corresponding interval queue. Orders in the queue trigger rebates in the order of entry time, and orders that enter the queue first receive priority in receiving fund refunds.

[0094] If user A's order enters the queue at 9:00 and user B's order enters the queue at 9:10, when the accumulated amount in the queue reaches the target, user A's order will complete the rebate before user B's order.

[0095] In the e-commerce platform's numerical splitting and price range dynamic sorting and return system, the core value of first-in-first-out is reflected in triggering rebates by strictly following the time order in which orders enter the corresponding interval queue after being split by price, ensuring that all users enjoy fair rebate priority and avoiding unfair resource allocation caused by human intervention or random processing; at the same time, through transparent time sorting rules, users can estimate the rebate progress based on the order in which orders enter the queue, improving operational predictability and user trust; in addition, the orderly queue processing mechanism can reduce the probability of task conflicts in concurrent scenarios, ensure the consistency of capital flow and data flow, and thereby enhance system stability, providing underlying order guarantees for the implementation of differentiated rebate strategies.

[0096] The fund pool management submodule sets up a fractional fund pool and an integer fund pool. The integer fund pool is further divided into a static pool accounting for 30% and a dynamic pool accounting for 70%. This allocation mechanism not only ensures the stability of funds, but also can flexibly adjust the use of funds according to actual trading conditions.

[0097] The rebate triggering and execution sub-module uses asynchronous scheduling rules to manage the fund pool, stipulating that rebates are only paid out from the integer fund pool. When the cumulative number of integers in a certain price range meets five orders, funds are returned from the integer fund pool according to the first-in-first-out principle. For example, when there are 5 50s in the cumulative pool of integers in the range of 50 to 100, the rebate operation will be triggered, and the corresponding funds will be paid out from the integer fund pool and returned to the user, increasing the number and probability of returns and promoting consumer consumption.

[0098] Example 2:

[0099] A system for splitting numerical values ​​and dynamically sorting and returning price ranges on an e-commerce platform. The backend data storage and management module uses multi-type databases and blockchain technology to achieve efficient storage, integrity assurance, and semantic retrieval of product prices, order data, and other data.

[0100] The backend data storage and management module includes a commodity price data storage submodule, a user order data recording submodule, a table structure design submodule, and an index strategy submodule;

[0101] The commodity price data storage submodule uses the time series database InfluxDB to store historical price change data and uses time series indexes to improve query efficiency;

[0102] The user order data recording submodule uses the Apache Cassandra database and adopts a dual partitioning strategy based on user ID and order time to store data;

[0103] The table structure design submodule adds a blockchain hash field to the transaction information table, implements tamper-proof data storage through smart contracts, and introduces a layered blockchain evidence storage system to ensure data integrity through a three-level structure, including transaction layer, summary layer, and verification layer.

[0104] The transaction layer adds a new blockchain_hash field in the relational database;

[0105] The summary layer writes the transaction hash value to the Hyperledger Fabric consortium chain every day;

[0106] The verification layer provides a real-time verification interface through smart contracts;

[0107] The index strategy submodule introduces vector indexing technology to establish indexes for unstructured data and support semantic retrieval.

[0108] Furthermore, the commodity price data storage submodule uses the time-series database InfluxDB to store historical data on commodity price changes. InfluxDB is a high-performance database designed specifically for time-series data, capable of efficiently processing timestamped price data. Using time-series indexing technology, this submodule optimizes and organizes price data according to the time dimension, significantly improving the efficiency of querying historical price data by time range. For example, when querying the price fluctuations of a particular commodity over the past week, the time-series index can quickly locate the relevant data interval, reducing the data scan range and significantly shortening query response time. Furthermore, InfluxDB supports efficient aggregation calculations on price data, such as calculating statistical indicators such as price averages, maximums, and minimums, providing strong support for the platform's price analysis and decision-making.

[0109] The user order data recording submodule uses the Apache Cassandra database to store user order data, employing a dual partitioning strategy based on user ID and order date. This dual partitioning strategy first partitions order data by user ID, ensuring that order data for the same user is stored in similar physical locations, improving the efficiency of querying orders by user. Furthermore, within each user ID partition, data is further partitioned by order date, enabling fast access to recent orders from the same user. This partitioning approach leverages Cassandra's distributed architecture to achieve horizontal scalability and load balancing of order data. When faced with a large number of concurrent order query requests, the dual partitioning strategy evenly distributes the requests across multiple nodes, avoiding single-point performance bottlenecks and ensuring stable system operation in high-concurrency scenarios.

[0110] The table structure design submodule ensures the integrity of transaction data through an innovative layered blockchain evidence storage system. At the transaction layer, a new field called "blockchain_hash" is added to the transaction information table in the relational database to store the hash value of each transaction. This hash value is a unique identifier derived by cryptographically calculating key transaction information, ensuring that transaction data cannot be tampered with while stored. The summary layer aggregates the hash values ​​of all transactions daily and writes them to the Hyperledger Fabric consortium blockchain. Maintained by multiple trusted nodes, the consortium blockchain is decentralized and tamper-proof, ensuring the authenticity and permanence of transaction summaries. The verification layer provides a real-time verification interface through smart contracts, allowing external systems or internal modules to verify the integrity of a transaction at any time. When verifying a transaction, simply recalculate the hash value of the transaction's key information and compare it with the hash value stored on the blockchain to quickly determine if the transaction data has been tampered with. This three-tiered blockchain evidence storage system retains the transaction processing and query advantages of relational databases while leveraging blockchain technology to ensure data immutability and traceability.

[0111] The indexing strategy submodule introduces vector indexing technology to handle unstructured data. On e-commerce platforms, unstructured data such as product descriptions and user reviews takes up a significant amount of storage space, and traditional indexing methods struggle to meet the requirements for semantic retrieval of this data. Vector indexing technology converts text data into vector representations in a high-dimensional vector space and uses similarity calculations between these vectors to enable semantic retrieval. For example, when a user searches for "comfortable sneakers," the system converts the search keyword into a vector and compares its similarity with the product description vectors, returning the products that best match the search intent. This indexing method not only understands the semantic meaning of the text but also handles semantically related queries such as synonyms and antonyms, significantly improving the accuracy and efficiency of retrieval of unstructured data. Furthermore, vector indexing technology supports efficient indexing and fast querying of massive amounts of unstructured data, ensuring that the system's retrieval performance remains unaffected even as data volumes continue to grow.

[0112] Example 3,

[0113] A system for splitting numerical values ​​and dynamically sorting and returning price ranges on an e-commerce platform. The backend service architecture module uses microservices, message queues, and real-time data processing technologies, combined with machine learning algorithms to achieve traffic control and intelligent data analysis.

[0114] The backend service architecture module includes a microservice architecture submodule, a real-time data processing submodule, and an offline data analysis submodule;

[0115] The microservice architecture submodule uses Istio service mesh technology to achieve refined control of traffic between services and secure communication, and implements service circuit breaking, degradation, and retry mechanisms through the Envoy proxy;

[0116] The message queue submodule uses the Pulsar message queue and introduces a content-based routing strategy to route different types of transaction data to corresponding processing channels.

[0117] The batch processing submodule uses Flink-CDC technology to capture data changes in real time, combines it with the Doris real-time data warehouse for data aggregation and analysis, and uses the random forest algorithm to predict and sort price ranges and time sequences.

[0118] Furthermore, the cloud-native design based on the Serverless architecture combines AWS Lambda and API Gateway to achieve serverless deployment of services. The platform can automatically and elastically scale computing resources according to request traffic. When traffic peaks, it automatically increases resources to handle a large number of requests. When traffic is low, it reduces resource usage and operating costs. At the same time, local access is achieved through edge computing nodes, which reduces response latency and improves user experience.

[0119] The microservice architecture submodule adopts Istio service mesh technology to achieve refined control of traffic and secure communication between services, and implements service circuit breaking, degradation and retry mechanisms through Envoy proxy. When a service fails or is overloaded, it automatically blocks or reduces request traffic to prevent the spread of the fault and ensure system stability.

[0120] The message queue submodule uses the Pulsar message queue and introduces a content-based routing strategy to route different types of transaction data to corresponding processing channels. For example, transaction data in different price ranges can be sent to different processing modules for processing through the routing strategy, improving the efficiency and targeting of data processing.

[0121] The batch processing submodule uses Flink-CDC technology to capture data changes in real time, combines it with the Doris real-time data warehouse for data aggregation and analysis, and uses the random forest algorithm to predict and sort price ranges and time sequences, providing data support for the platform's operational decisions, such as predicting sales trends in different price ranges and rationally arranging inventory.

[0122] Example 4:

[0123] A system for splitting numerical values ​​and dynamically sorting and returning prices within e-commerce platforms. The front-end display module is based on the Vue3 technology stack and integrates mixed reality, dynamic tables, and intelligent loading technologies to achieve price and rebate visualization and real-time interaction with user data.

[0124] The front-end display module includes a sub-module for integrated display of product prices and rebate rules, a sub-module for dynamic data interaction, a sub-module for paging and intelligent loading, and a sub-module for personal account data dashboard.

[0125] The integrated display submodule of product prices and rebate rules uses mixed reality technology to achieve simultaneous display of 3D visualization of product prices and simulation of rebate rules on mobile devices.

[0126] The dynamic data interaction submodule integrates the original dynamic table and chart display functions, and uses VantTable and EChartsGL to implement data table filtering, sorting and 3D chart analysis;

[0127] The paging and smart loading submodule uses infinite scrolling loading combined with a priority queue strategy to optimize data loading efficiency through a global data cache pool, ensuring the coordination of real-time data updates and paging loading.

[0128] The personal account data dashboard sub-module associates the transaction information table with the blockchain hash value based on the user IP, and displays in real time the price range corresponding to the integer part of each order after splitting, as well as the arrangement order of the integer in the corresponding interval queue. It is updated in real time using dynamic table technology.

[0129] Furthermore, the integrated display submodule for product prices and rebate rules leverages mobile mixed reality technology, using the AR.js and Three.js frameworks, to present product prices and rebate rules in an immersive 3D visualization. When users scan a product QR code or click on a product image with their phone camera, the system overlays a 3D virtual model of the product onto the real-world scene, dynamically generating visual elements around the model to display price and rebate information. For example, the price value appears suspended above the model in a bright, 3D font, with different colors corresponding to different price ranges. The rebate rules are animated to simulate the splitting process, such as splitting a 128 yuan product price into 100 yuan and 28 yuan, with the 100 yuan entering the corresponding price range's accumulation pool. Users can rotate and zoom the model using gestures, and clicking on the price or rebate rule visualization triggers a pop-up window with detailed instructions and calculation logic, helping them intuitively understand the complex price splitting and rebate mechanisms.

[0130] The dynamic data interaction submodule builds a highly interactive data display system based on VantTable and EChartsGL. At the data table level, VantTable implements powerful filtering and sorting capabilities. Users can use the filter bar above the table to perform multiple filters on fields such as price range, rebate status, and order time, supporting various filtering methods such as drop-down selection and date range selection. Clicking a column header sorts data in ascending or descending order, and custom sorting rules can be used through the right-click menu. For 3D chart analysis, EChartsGL transforms data into intuitive 3D visualizations. For example, a 3D bar chart displays the distribution of rebate amounts across different time periods and price ranges. Users can rotate the chart by dragging the mouse to observe data changes from multiple angles. A 3D line chart presents rebate trends, allowing key data points to be directly annotated on the chart. Clicking an annotation reveals detailed information. Furthermore, the table and chart feature bidirectional linkage. Selecting a data item in the table automatically highlights the corresponding data point in the chart, and vice versa, facilitating multi-dimensional data exploration and analysis.

[0131] The paging and smart loading submodule utilizes a strategy that combines infinite scrolling with a priority queue to optimize data loading performance. As users browse a page, the system monitors the scroll bar position in real time. When the scroll bar reaches 200 pixels from the bottom of the page, it automatically triggers a request to load the next page of data. Data loading is prioritized based on timeliness and importance: newly generated order data and orders with upcoming rebates are prioritized and retrieved and displayed first; less important data, such as historical order data, is loaded sequentially. The global data cache pool uses a least-repeated (LRU) algorithm to manage the cache, automatically eliminating the least recently used data when the cache is full. For data updates, the system establishes a persistent connection via WebSocket, receiving new data pushed by the server in real time and automatically updating the page display. This, in conjunction with the infinite scrolling mechanism, ensures that new data is displayed promptly without affecting the smoothness of page loading. Furthermore, to address network instability, the system includes a data preload function that pre-loads data likely to be needed for the next page while the user is browsing the current page, further enhancing the user experience.

[0132] The personal account data dashboard submodule uses user IP addresses and blockchain hash values ​​to accurately correlate and display personal account data in real time. When a user logs in, the system uses their IP address as a unique identifier. This unique identifier, combined with the blockchain hash algorithm, generates a secure key. This key is then matched against the blockchain_hash field in the transaction information table, ensuring that only authorized users can access their account data. The dashboard page utilizes dynamic table technology to display detailed, broken-down information for each order in real time. Each row in the table corresponds to an order and includes fields such as the order number, original price, split-up integer portion, price range, order within the corresponding range queue, and estimated rebate time. Data updates utilize virtual DOM technology, so when new data is generated, only the changed data is updated, reducing page redraw overhead. For example, when the rebate status of an order changes, the table automatically refreshes the row's data and notifies the user with a color change. The dashboard also offers a data export function, allowing users to export table data as a CSV file for local analysis and archiving.

[0133] Example 5,

[0134] A system for splitting values ​​and dynamically sorting and rebating price ranges on e-commerce platforms. The data statistics and analysis module uses time series analysis and clustering algorithms to deeply explore the correlation between rebate trends and user behavior, supporting personalized recommendations and strategy optimization.

[0135] The data statistics and analysis module includes the rebate statistics submodule and the user behavior analysis submodule;

[0136] The rebate statistics submodule uses Python's Pandas, Numpy, and Matplotlib libraries to conduct in-depth statistical analysis of the previous week's rebate data at 2:00 AM every Monday. The analysis metrics include the number of rebates, total rebate amount, number of users benefiting, average single rebate amount, and standard deviation of rebate amount across different price ranges. Using the SARIMA time series analysis method, it predicts rebate trends for each price range over the next four weeks. The analysis results are presented in the form of a visual dashboard, which includes a bar chart showing the number of rebates and the number of users benefiting across different price ranges, a line chart showing the time trend of the total rebate amount, and a predicted curve for the rebate trend over the next four weeks. Users can click on chart elements to view data and switch between different price ranges for comparative analysis.

[0137] The user behavior analysis submodule combines user purchase order data and rebate participation, integrates the FP-Growth algorithm and the DBSCAN algorithm, and uses the FP-Growth algorithm to mine the association rules of user purchases. The minimum support is set to 0.05 and the minimum confidence is set to 0.6. The associations between products in different price ranges are found, and these association rules are input into the DBSCAN algorithm as features. The elbow method is used to determine the number of clusters to be 5, and cluster analysis is performed on users. Based on the clustering results, personalized product recommendation strategies and rebate mechanisms are formulated for each user group, and changes in user behavior are monitored in real time. The monitoring indicators include user purchase frequency, price range distribution of purchased products, and number of rebate participations. The monitoring frequency is once a day, and the clustering results and recommendation strategies are updated every two weeks.

[0138] Furthermore, the rebate statistics submodule has built a complete periodic data analysis process. At 2 a.m. every Monday, a scheduled task extracts detailed rebate data from the data warehouse for the previous week, including fields such as order ID, price range, rebate amount, user ID, and timestamp. In the data preprocessing stage, Pandas' dropna() method is used to handle missing values, and the z-score method is used to detect and replace outliers. In the statistical analysis stage, groupby() is used to group by price range. Core indicators are calculated, including count() to count the number of rebates, sum() to calculate the total amount, nunique() to count the number of benefited users, and mean() and std() to calculate the mean and standard deviation of a single rebate.

[0139] The SARIMA model is used in time series analysis to predict future rebate trends. First, the data stationarity is confirmed by the ADF test. If it is not stationary, differential processing is performed. The auto_arima function of the pmdarima library is used to automatically search for the optimal parameter combination (p, d, q) (P, D, Q, s). For example, for data in the range of "100-200 yuan", the SARIMA (1, 1, 1) (1, 1, 1, 7) model is fitted. The rebate amount series for the next 28 days is generated during the forecast, and the 95% confidence interval is calculated. The interactive dashboard is constructed using Matplotlib and Plotly for visualization.

[0140] The interactive design of the dashboard allows users to click on the legend to switch price ranges, hover to view specific values, and select local areas to zoom in on details. The system also provides a data export function, supports downloading raw data in CSV format, and saving charts in PNG format. To ensure the timeliness of analysis results, an email is automatically sent to the operations team after each report is generated, and the internal BI system dashboard is updated.

[0141] The user behavior analysis submodule implements a complete analysis chain from association rule mining to user clustering. In the data preparation stage, user order data is converted into a transaction database format. Each transaction contains the user ID, a list of purchased items, and the corresponding price range.

[0142] When the FP-Growth algorithm is executed, the FP tree is first constructed, and the minimum support is set to 0.05 through the pyfpgrowth library to generate frequent item sets. Then, the association rules are calculated, and the minimum confidence is set to 0.6 to filter out meaningful rules.

[0143] During the feature engineering phase, association rules are converted into user feature vectors: each user corresponds to a multidimensional vector, with the dimensions representing all association rules and the vector value representing the user's support for the rule. DBSCAN clustering is implemented using the sklearn library. The silhouette coefficient is calculated under different EPS values ​​using the elbow method to determine the optimal parameter combination. Users are then divided into five categories: high-value, affordable, balanced, potential, and active. High-value users frequently purchase high-priced goods and have rich association rules; affordable users primarily purchase low-priced goods; balanced users are evenly distributed across all price ranges; potential users occasionally purchase high-priced goods; and active users purchase frequently but in low amounts.

[0144] Exclusive rebate coupons for high-priced items are pushed to high-value users, while budget-conscious users are recommended bundles of low-priced items. The real-time monitoring module calculates user behavior metrics daily: purchase frequency = number of orders in the past 30 days / 30, price range distribution entropy = -Σ(p_i*log(p_i)) (p_i is the proportion of user spending in the i-th price range), and rebate participation rate = number of participating rebate orders / total number of orders. The clustering model is retrained every two weeks, and the stability of the user group is assessed by comparing the cosine similarity of the new and old cluster centers. When the similarity falls below the threshold of 0.8, the recommendation strategy is updated to ensure the accuracy of personalized service.

[0145] Example 6,

[0146] A system for splitting numerical values ​​and dynamically sorting and returning price ranges on e-commerce platforms. Its security and stability assurance module ensures high system availability and data security through scheduled backups, circuit breakers, current limiting, blockchain evidence storage, and search engine optimization.

[0147] The security and stability assurance module includes the data backup and recovery sub-module, the concurrent processing and performance optimization sub-module, the search engine technology sub-module, and the blockchain evidence storage sub-module;

[0148] The data backup and recovery submodule uses MySQL's mysqldump command to perform a full database backup at 3:00 a.m. every day and stores the backup file in an off-site cloud storage service. An incremental backup is performed every Saturday at 4:00 a.m. At the same time, a data disaster recovery plan is formulated. Within one hour after a database failure, data is restored using the most recent full and incremental backup files. In the event of a disaster in an off-site data center, service is switched to the backup data center within three hours. During each full and incremental backup process, the backup speed and completion percentage are monitored in real time. If the backup speed drops below 1MB / s for five consecutive minutes, an alarm is sent to notify the operation and maintenance personnel.

[0149] The concurrent processing and performance optimization submodule uses the Spring Cloud distributed architecture combined with service grid technology to implement system microservices and traffic management. It uses the Druid database connection pool, setting the maximum number of connections to 200 and the minimum number of idle connections to 20.

[0150] Use Prometheus and Grafana to build a system performance monitoring platform to monitor the system's concurrent request number, response time, and database connection number indicators in real time. When the number of concurrent requests exceeds 1000 times / second, the circuit breaker and current limiting mechanism are automatically triggered. The current limiting ratio is 50%, that is, only 50% of the requests are allowed to pass, and the circuit breaker recovery time is 5 minutes.

[0151] The search engine technology submodule uses Elasticsearch search engine technology when processing query and sorting scenarios involving more than 10,000 records. The Elasticsearch cluster is configured with three nodes, each equipped with an 8-core CPU, 16GB of memory, and a 500GB hard drive.

[0152] Check the Elasticsearch index weekly. When the index fragmentation rate exceeds 30% or the index space usage exceeds 80%, optimize and rebuild the index.

[0153] The blockchain evidence storage submodule adopts a hybrid on-chain and off-chain architecture. The original system calculation is retained off-chain to ensure high concurrent processing efficiency. On-chain, the split integer amount, queue position and rebate trigger time data are written into the private blockchain through smart contracts, generating an unalterable transaction hash and constructing a five-sequence rebate chain to record split records, queue status, rebate execution, behavior weight and fund pool dynamic information.

[0154] Furthermore, the data backup and recovery submodule has established a multi-layered data protection system. Daily full backups are performed by exporting the database physical files using the mysqldump command, using the single-transaction parameter to ensure data consistency. After compression, the files are transferred to off-site cloud storage via an SSL-encrypted channel. Cloud storage utilizes multi-region redundant storage and sets access control lists to restrict access to authorized IP addresses only. Weekly incremental backups utilize the MySQL binary log, extracting change records using the mysqlbinlog tool and generating incremental SQL scripts.

[0155] In the event of a database failure, operations personnel first confirm the failure using the ping command and database service status check, then initiate the backup and recovery script. The script automatically downloads the latest full backup file, uses commands to restore the underlying data, and then applies the most recent incremental backup file to complete the changes. When switching to a remote data center, the preconfigured backup IP address in DNS is quickly pointed to the disaster recovery center, and the application service dynamically loads the disaster recovery environment parameters through the configuration center. The backup monitoring system collects I / O rate metrics of the backup process through Prometheus, and Grafana draws real-time curve charts. If the backup speed is detected to be less than 1MB / s for five consecutive minutes, an alarm rule is triggered, and the operations team is notified through multiple channels such as WeChat, SMS, and email.

[0156] The concurrency processing and performance optimization submodule utilizes the Spring Cloud microservices framework to achieve service decomposition, dividing the core business into independent modules such as price splitting, rebate calculation, and fund pool management. Inter-service communication utilizes a unified Spring Cloud Gateway portal, integrated with Sentinel for flow control. Nacos is used for service registration and discovery, enabling automatic service registration, health checks, and load balancing. When a service instance times out (the default threshold is 2 seconds), Nacos automatically removes it from the available list.

[0157] The database connection pool configuration adheres to performance optimization principles: the maximum number of connections is set to 200, calculated based on an empirical formula: the number of CPU cores on the database server multiplied by 12.5. The minimum number of idle connections is set to 20 to ensure sufficient connections during cold starts. PSCache is enabled in the Druid connection pool, with a cache size of 50 to reduce SQL compilation overhead. The monitoring platform collects connection pool metrics via Prometheus's JMX exporter, and Grafana sets alert thresholds: an alert is triggered when the number of active connections exceeds 180, representing 90% utilization, and a circuit breaker is triggered when the number exceeds 200.

[0158] When the number of concurrent requests exceeds 1000 per second, isolation mode is triggered, queuing requests by business priority. 70% of tokens are allocated to real-time rebate calculations for core services, and 30% to historical data exports for non-core services. Rejected requests return to a custom downgrade page with the message "System busy, please try again later." Circuit breaker recovery uses a gradual strategy: first reducing the throttling ratio to 80%, then gradually restoring it to 100% after observing no abnormalities for 30 seconds. The entire process is automatically controlled by a state machine.

[0159] The Elasticsearch cluster in the search engine technology submodule utilizes a three-node hot standby architecture, with nodes automatically discovered and clustered using the ZenDiscovery mechanism. Each node is configured with dedicated master, data, and ingest roles to avoid single points of failure. The indexing strategy is optimized for business characteristics: the product price range index uses the keyword type to support exact matching; the rebate amount field uses the scaled_float type with a scaling_factor of 100 to ensure accuracy while reducing storage space.

[0160] The index maintenance process includes automated checks and optimizations: Index health checks are performed every Sunday at 2:00 AM, using the _cat / indices API to obtain fragmentation and space usage. When the fragmentation rate exceeds 30%, a _forcemerge operation with max_num_segments=1 is executed to merge fragments. When space usage exceeds 80%, an index rollover strategy is initiated: a new index is created, an alias pointing to the new index is created, and the old index is marked as read-only. Regarding performance optimization, inverted indexes are created for frequently queried fields, and BKD tree indexes are created for range query fields, improving query performance by 30% to 50%.

[0161] The blockchain evidence storage submodule's hybrid on-chain and off-chain architecture achieves a balance between performance and security. The off-chain system retains an efficient relational database to handle high-concurrency transactions, while the on-chain Hyperledger Fabric consortium chain records key data. Smart contracts are developed in Go and deployed within a consortium chain network comprised of platform operators, regulators, and third-party auditors. When an order price is split, the system generates JSON data containing the order number, split amount, and timestamp, calculates its SHA-256 hash, and writes the hash to the blockchain by calling the on-chain contract's saveSplitRecord method via the Fabric SDK.

[0162] The data structure of the Five-Sequence Rebate Chain is designed as follows: the split record block contains a Merkle tree structure, aggregating the hash values ​​of multiple split records from the same batch; the queue status block maintains a FIFO queue through a smart contract, with each node containing the hash and timestamp of the previous node; the rebate execution block records the flow of funds, linking the fund pool account and the user's wallet address; the behavior weight block stores user behavior feature values, implementing the weight calculation logic through a Solidity contract; and the fund pool dynamic block records the details of each fund inflow and outflow, ensuring fund traceability through the UTXO model. All blocks use the PBFT consensus algorithm to ensure that the system can still reach consensus even when 3f+1 nodes are functioning properly, and transaction confirmation time is controlled within 2 seconds. Users can query on-chain data and verify transaction authenticity through the Web3.js library, and auditors can use the verifyRecord method provided by the smart contract to compare on-chain and off-chain data consistency.

[0163] Embodiment 7,

[0164] A system for splitting numerical values ​​and dynamically sorting price ranges for rebates on e-commerce platforms. The monitoring and feedback mechanism module builds a Prometheus+Grafana monitoring system, combining multi-channel feedback with intelligent traceability to achieve rebate progress tracking and multi-dimensional user queries.

[0165] The monitoring and feedback mechanism module includes a system monitoring submodule, a user feedback processing submodule, a rebate progress tracking submodule, and a user query submodule;

[0166] The system monitoring submodule is used to deploy Prometheus and combine it with Grafana to build a monitoring system, set specific monitoring indicators for different links, and synchronize data to the intelligent problem tracing system;

[0167] The user feedback processing submodule is used to establish a multi-channel feedback system, including online customer service, a real-time open user forum, and a 24-hour complaint mailbox. It also introduces intelligent problem tracing capabilities, automatically correlating system monitoring data from the same period to generate analysis reports to assist in problem location.

[0168] The rebate progress tracking submodule queries the integer accumulation pool data of the accumulation storage submodule through the shared Redis cache. Combined with the asynchronous scheduling rules of the rebate trigger and execution submodule, it calculates the matching status of each rebate trigger condition and displays the current status of each condition in parallel on the interface. At the same time, it displays the progress of the integer accumulation amount in the current price range relative to the target value as a percentage.

[0169] The user query submodule provides three query methods: transaction hash, mobile phone number, and blockchain address. Users can enter the blockchain hash associated with the order number to view the complete transaction split and return link, aggregate the progress of rebate orders in progress by linking the mobile phone number to the account, and query the on-chain fund flow record based on the blockchain address.

[0170] The display content includes the queue position and estimated rebate time calculated by combining historical queue digestion speed with progress prediction algorithms such as exponential smoothing, the real-time behavior weight coefficient and the equivalent queue position after acceleration, and the real-time balances of the static pool and dynamic pool presented in dynamic charts, as well as the proportion of the daily distribution amount.

[0171] Furthermore, the system monitoring submodule builds a multi-layered monitoring system based on Prometheus and Grafana. At the data collection layer, Prometheus's exporter components collect various metrics: Nodeexporter monitors server CPU, memory, and disk I / O; JMXexporter collects JVM metrics for Java applications; and MySQLexporter captures database slow queries and connection pool status. Custom business metrics are integrated into Spring Boot applications using the Micrometer library and exposed as Prometheus-formatted endpoints.

[0172] The infrastructure layer monitors server load and network latency; the application layer monitors service response time and error rates, with a service response time SLA target of P99 < 500ms and an error rate threshold of 0.5%. Core business indicators monitor the number of accumulated pools in each price range and the frequency of rebate triggers. Data storage uses Thanos extended with Prometheus, retaining 180 days of historical data. Multi-dimensional dashboards configured with Grafana allow operations and maintenance personnel to monitor the overall health of the system.

[0173] The intelligent problem tracing system uses ELKStack for log aggregation and analysis, aligning time-series data collected by Prometheus with log data timestamps. When an abnormal metric is detected, it automatically correlates the corresponding logs from the same period. The Pinpoint distributed link tracing system then locates the specific microservice instance and call chain, generating an analysis report that includes abnormal metric trends, associated log fragments, and service dependencies.

[0174] The user feedback processing submodule integrates online customer service, forums, and email systems through a multi-channel feedback system. Online customer service uses the WebSocket protocol for full-duplex communication, supporting text, image, and file transfers. An integrated intelligent reply bot automatically responds to frequently asked questions. The user forum, built on Discourse, manages posts using categories and tags, with pinning, highlighting, and voting options. The operations team regularly purges invalid posts daily. The complaint mailbox is connected to the internal ticket system via the IMAP protocol, allowing for automatic keyword categorization.

[0175] When receiving user feedback, the system automatically extracts system monitoring data for 30 minutes before and after the feedback timestamp, including Prometheus indicators and Grafana dashboard screenshots. It then combines the user ID with the behavior log to analyze the user operation path. The generated analysis report includes a system status snapshot, suspicious anomalies, and historical solutions to similar problems, helping customer service personnel quickly locate the problem.

[0176] The rebate progress tracking submodule uses Redis caching for high-performance data querying. The cumulative storage submodule writes the integer amount and its corresponding price range into a Redis SortedSet data structure after completing each order price split, using timestamps as fractions to ensure FIFO order. The rebate progress calculation service reads the cumulative amount for each range from Redis every 5 minutes and compares it with the target value of 5 orders to calculate the completion percentage. The front-end interface uses WebSocket to push progress updates in real time. The price range cards use TailwindCSS to implement a dynamic progress bar, automatically switching to an orange warning state when the completion rate exceeds 80%.

[0177] When the cumulative volume reaches 5 or 6 orders within a certain price range, payments are calculated based on retained earnings. Within each range, only one triggering condition is selected to trigger the rebate execution task. Available funds are retrieved from the fund pool management submodule, and the rebate operation is executed in queue order. The interface displays real-time status, including: current cumulative volume / target value (e.g., 3 / 5), queue head waiting time (e.g., "Waiting for 2 hours and 15 minutes"), and historical average processing speed (e.g., "The historical average processing time per order in this range is 4 hours"). The system uses exponential smoothing to predict rebate times, using the formula: Estimated time = current time + (queue length × historical average processing time × smoothing coefficient 0.7).

[0178] The user query submodule implements multi-dimensional data retrieval. Transaction hash queries use Web3.js to call blockchain nodes, verify the authenticity of order split records, and return complete JSON data containing the original price, split value, and queue position. Mobile phone number queries link user accounts and aggregate all in-transit rebate orders: a list view displays the order number, split amount, and current queue position; a chart view displays the rebate progress distribution by price range. Blockchain address queries call the smart contract's getTransactionHistory method to obtain on-chain fund flow records, including fund pool allocation details and rebate receipt records.

[0179] The system maintains the historical queue digestion speed of each price range, calculates the weighted moving average through the exponential smoothing method, predicts the future processing speed, and dynamically adjusts the behavior weight coefficient based on the user's purchase frequency, repurchase rate, etc. The front-end uses Chart.js to draw the fund pool balance change curve, distinguishes static pools from dynamic pools by different colors, and updates the proportion of the daily allocation amount in real time. All query results can be exported to CSV files to facilitate users to save records.

[0180] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations that come within the meaning and range of equivalents of the claims be embraced therein.

Claims

1. A system for splitting numerical values ​​and dynamically sorting and returning price ranges on an e-commerce platform, characterized by: It includes back-end data storage and management module, system logic and algorithm execution module, back-end service architecture module, front-end display module, data statistics and analysis module, security and stability assurance module and monitoring and feedback mechanism module; The backend data storage and management module uses multiple database types and blockchain technology to achieve efficient storage, integrity assurance, and semantic retrieval of data such as product prices and orders; The system logic and algorithm execution module splits the executed order price based on the anchor value, and realizes dynamic sorting of price ranges and rebate returns through cumulative storage, fund pool management and asynchronous scheduling; The system logic and algorithm execution module includes the price splitting submodule, the accumulation storage submodule, the fund pool management submodule, and the rebate triggering and execution submodule; The price splitting submodule splits the order transaction price using 10, 30, 50, and 100 as anchor points. It first allocates the price to the price intervals of 10-30, 30-50, 50-100, and >100. It then performs a subtraction operation on the maximum value of the price interval, removing the remainder and retaining the integer part. If the split value still falls within another interval, it is assigned to the corresponding interval for further splitting, with the minimum splitting number being 10. The accumulation storage submodule is used to store the integer part obtained after the price splitting submodule is processed into the integer accumulation pool of the corresponding price range in the order of transaction time; The fund pool management submodule sets up a fractional fund pool and an integer fund pool. The integer fund pool is divided into a static pool accounting for 30% and a dynamic pool accounting for 70%. The static pool is used for fixed rules, and the dynamic pool is used for recommendation and repurchase weights. The rebate triggering and execution submodule uses asynchronous scheduling rules to manage the fund pool, stipulating that rebates are only paid from the integer fund pool. When the cumulative number of integers in the price range meets 5 or 6 orders, rebates are calculated based on retained earnings. Only one triggering return condition is selected in the same range. According to the first-in-first-out principle, every time 5 or 6 orders trigger the return condition, the first order is returned. The return is made in sequence, and the cycle continues, returning funds from the integer fund pool.

2. The e-commerce platform numerical value splitting and price range dynamic sorting and return system according to claim 1 is characterized by: The backend service architecture module uses microservices, message queues, and real-time data processing technologies, combined with machine learning algorithms to achieve traffic control and intelligent data analysis; The front-end display module is based on the Vue3 technology stack and integrates mixed reality, dynamic tables, and intelligent loading technologies to achieve price and rebate visualization and real-time interaction with user data; The data statistics and analysis module uses time series analysis and clustering algorithms to deeply explore the correlation between rebate trends and user behavior, supporting personalized recommendations and strategy optimization; The security and stability assurance module ensures high system availability and data security through scheduled backup, circuit breaker and current limiting, blockchain evidence storage, and search engine optimization; The monitoring and feedback mechanism module builds a Prometheus+Grafana monitoring system, combining multi-channel feedback and intelligent traceability to achieve rebate progress tracking and multi-dimensional user queries.

3. The e-commerce platform numerical value splitting and price range dynamic sorting and return system according to claim 1 is characterized by: The backend data storage and management module includes a commodity price data storage submodule, a user order data recording submodule, a table structure design submodule, and an index strategy submodule; The commodity price data storage submodule uses the time series database InfluxDB to store historical price change data and uses time series indexes to improve query efficiency; The user order data recording submodule uses the Apache Cassandra database and adopts a dual partitioning strategy based on user ID and order time to store data; The table structure design submodule adds a blockchain hash field to the transaction information table, implements tamper-proof data storage through smart contracts, and introduces a layered blockchain evidence storage system to ensure data integrity through a three-level structure, including transaction layer, summary layer, and verification layer. The transaction layer adds a new blockchain_hash field in the relational database; The summary layer writes the transaction hash value to the Hyperledger Fabric consortium chain every day; The verification layer provides a real-time verification interface through smart contracts; The index strategy submodule introduces vector indexing technology to establish indexes for unstructured data and support semantic retrieval.

4. The e-commerce platform numerical value splitting and price range dynamic sorting and return system according to claim 1 is characterized by: The backend service architecture module includes a microservice architecture submodule, a real-time data processing submodule, and an offline data analysis submodule; The microservice architecture submodule uses Istio service mesh technology to achieve refined control of traffic between services and secure communication, and implements service circuit breaking, degradation, and retry mechanisms through the Envoy proxy; The message queue submodule uses the Pulsar message queue and introduces a content-based routing strategy to route different types of transaction data to corresponding processing channels. The batch processing submodule uses Flink-CDC technology to capture data changes in real time, combines it with the Doris real-time data warehouse for data aggregation and analysis, and uses the random forest algorithm to predict and sort price ranges and time sequences.

5. The e-commerce platform numerical value splitting and price range dynamic sorting and return system according to claim 1 is characterized by: The front-end display module includes a sub-module for integrated display of product prices and rebate rules, a sub-module for dynamic data interaction, a sub-module for paging and intelligent loading, and a sub-module for personal account data dashboard. The integrated display submodule of product prices and rebate rules uses mixed reality technology to achieve simultaneous display of 3D visualization of product prices and simulation of rebate rules on mobile devices. The dynamic data interaction submodule integrates the original dynamic table and chart display functions, and uses VantTable and EChartsGL to implement data table filtering, sorting and 3D chart analysis; The paging and smart loading submodule uses infinite scrolling loading combined with a priority queue strategy to optimize data loading efficiency through a global data cache pool, ensuring the coordination of real-time data updates and paging loading. The personal account data dashboard sub-module associates the transaction information table with the blockchain hash value based on the user IP, and displays in real time the price range corresponding to the integer part of each order after splitting, as well as the arrangement order of the integer in the corresponding interval queue. It is updated in real time using dynamic table technology.

6. The e-commerce platform numerical value splitting and price range dynamic sorting and return system according to claim 1 is characterized by: The data statistics and analysis module includes the rebate statistics submodule and the user behavior analysis submodule; At 2:00 a.m. every Monday, the rebate statistics submodule uses Python's Pandas, Numpy, and Matplotlib libraries to conduct in-depth statistical analysis of the previous week's rebate data. The analysis indicators include the number of rebates in different price ranges, the total rebate amount, the number of benefited users, the average amount of a single rebate, and the standard deviation of the rebate amount. The SARIMA time series analysis method is used to predict the rebate trend in each price range for the next four weeks. The analysis results are displayed in the form of a visual dashboard. The dashboard includes a bar chart showing the number of rebates and the number of benefited users in different price ranges, a line chart showing the time trend of the total rebate amount, and a predicted curve of the rebate trend in the next four weeks. Users can view data and switch between different price ranges for comparative analysis by clicking on the chart elements.

7. The e-commerce platform numerical value splitting and price range dynamic sorting and return system according to claim 6 is characterized by: The user behavior analysis submodule combines user purchase order data and rebate participation, integrates the FP-Growth algorithm and the DBSCAN algorithm, and uses the FP-Growth algorithm to mine the association rules of user purchases. The minimum support is set to 0.05 and the minimum confidence is set to 0.

6. The associations between products in different price ranges are found, and these association rules are input into the DBSCAN algorithm as features. The elbow method is used to determine the number of clusters to be 5, and cluster analysis is performed on users. Based on the clustering results, personalized product recommendation strategies and rebate mechanisms are formulated for each user group, and changes in user behavior are monitored in real time. The monitoring indicators include user purchase frequency, price range distribution of purchased products, and number of rebate participations. The monitoring frequency is once a day, and the clustering results and recommendation strategies are updated every two weeks.

8. The e-commerce platform numerical value splitting and price range dynamic sorting and return system according to claim 1 is characterized by: The security and stability assurance module includes the data backup and recovery sub-module, the concurrent processing and performance optimization sub-module, the search engine technology sub-module, and the blockchain evidence storage sub-module; The data backup and recovery submodule uses MySQL's mysqldump command to perform a full database backup at 3:00 a.m. every day and stores the backup file in an off-site cloud storage service. An incremental backup is performed every Saturday at 4:00 a.m. At the same time, a data disaster recovery plan is formulated. Within one hour after a database failure, data is restored using the most recent full and incremental backup files. In the event of a disaster in an off-site data center, service is switched to the backup data center within three hours. During each full and incremental backup process, the backup speed and completion percentage are monitored in real time. If the backup speed drops below 1MB / s for five consecutive minutes, an alarm is sent to notify the operation and maintenance personnel. The concurrent processing and performance optimization submodule uses the Spring Cloud distributed architecture combined with service grid technology to implement system microservices and traffic management. It uses the Druid database connection pool, setting the maximum number of connections to 200 and the minimum number of idle connections to 20. Use Prometheus and Grafana to build a system performance monitoring platform to monitor the system's concurrent request number, response time, and database connection number indicators in real time. When the number of concurrent requests exceeds 1000 times / second, the circuit breaker and current limiting mechanism are automatically triggered. The current limiting ratio is 50%, that is, only 50% of the requests are allowed to pass, and the circuit breaker recovery time is 5 minutes.

9. The e-commerce platform numerical value splitting and price range dynamic sorting and return system according to claim 8, characterized in that: The search engine technology submodule uses Elasticsearch search engine technology when processing query and sorting scenarios involving more than 10,000 records. The Elasticsearch cluster is configured with three nodes, each equipped with an 8-core CPU, 16GB of memory, and a 500GB hard drive. Check the Elasticsearch index weekly. When the index fragmentation rate exceeds 30% or the index space usage exceeds 80%, optimize and rebuild the index. The blockchain evidence storage submodule adopts a hybrid on-chain and off-chain architecture. The original system calculation is retained off-chain to ensure high concurrent processing efficiency. On-chain, the split integer amount, queue position and rebate trigger time data are written into the private blockchain through smart contracts, generating an unalterable transaction hash and constructing a five-sequence rebate chain to record split records, queue status, rebate execution, behavior weight and fund pool dynamic information.

10. The e-commerce platform numerical value splitting and price range dynamic sorting and return system according to claim 1 is characterized by: The monitoring and feedback mechanism module includes a system monitoring submodule, a user feedback processing submodule, a rebate progress tracking submodule, and a user query submodule; The system monitoring submodule is used to deploy Prometheus and combine it with Grafana to build a monitoring system, set specific monitoring indicators for different links, and synchronize data to the intelligent problem tracing system; The user feedback processing submodule is used to establish a multi-channel feedback system, including online customer service, a real-time open user forum, and a 24-hour complaint mailbox. It also introduces intelligent problem tracing capabilities, automatically correlating system monitoring data from the same period to generate analysis reports to assist in problem location. The rebate progress tracking submodule queries the integer accumulation pool data of the accumulation storage submodule through the shared Redis cache. Combined with the asynchronous scheduling rules of the rebate trigger and execution submodule, it calculates the matching status of each rebate trigger condition and displays the current status of each condition in parallel on the interface. At the same time, it displays the progress of the integer accumulation amount in the current price range relative to the target value as a percentage. The user query submodule provides three query methods: transaction hash, mobile phone number, and blockchain address. Users can enter the blockchain hash associated with the order number to view the complete transaction split and return link, aggregate the progress of rebate orders in progress by linking the mobile phone number to the account, and query the on-chain fund flow record based on the blockchain address. The display content includes the queue position and estimated rebate time calculated by combining historical queue digestion speed with progress prediction algorithms such as exponential smoothing, the real-time behavior weight coefficient and the equivalent queue position after acceleration, and the real-time balances of the static pool and dynamic pool presented in dynamic charts, as well as the proportion of the daily distribution amount.

Images