Cross-border electronic commerce and trade platform based on Internet mining

Abstract

The invention discloses a cross-border electronic commerce and trade platform based on Internet mining, and belongs to the technical field of electronic commerce and trade, and the platform specifically comprises a parameter obtaining module which is used for obtaining real-time network quality data of a geographic area where a user terminal is located; the data processing module is used for matching a compression configuration scheme in a preset processing strategy library according to the network quality data, and the compression configuration scheme comprises a combination condition of a text simplification rule, an image conversion parameter and a video fragmentation standard; the edge computing module is used for creating an isolated operation environment at an edge computing node to which a user access base station belongs, and loading a basic data set and a logic judgment rule required by platform service processing; the real-time monitoring module is used for continuously collecting network quality data and dynamically updating a compression configuration scheme and resource allocation of edge computing nodes; the method provides reliable technical support for popularization and application of cross-border electronic commerce and trade.

Description

Technical Field

[0001] The present invention relates to the technical field of electronic commerce, and in particular to a cross-border electronic commerce platform based on Internet mining. Background Art

[0002] With the continued advancement of global trade, cross-border e-commerce platforms have become a crucial vehicle for international trade. However, network infrastructure development is uneven across different regions. Some regions are constrained by low bandwidth, high latency, and poor stability, resulting in slow platform access and inefficient data transmission. Particularly in emerging markets such as Africa and Southeast Asia, users of cross-border e-commerce services often experience page loading delays and slow transaction responses, directly impacting user experience and conversion rates. Therefore, optimizing platform performance within a global network environment has become a pressing technical challenge in the cross-border e-commerce sector.

[0003] Currently, the industry generally uses content delivery networks (CDNs) and distributed server deployments to mitigate the impact of network disparities. CDNs cache static resources at multiple nodes around the world, reducing data transmission distances and improving access speeds. Alternatively, edge computing can be used to process some business logic at nodes close to users, reducing the burden on central servers.

[0004] However, traditional CDN and data compression typically use fixed strategies and cannot dynamically adjust to real-time network conditions, resulting in potential lag or load failures even during network fluctuations. Existing edge computing solutions are mostly used for caching static data or simple queries, failing to fully integrate with e-commerce business characteristics such as dynamic product recommendations and real-time transaction verification. As a result, some key operations still rely on central servers, failing to fully reduce latency. Existing solutions typically only optimize single points, failing to establish a global network awareness and dynamic adjustment system from user terminals to edge nodes to central servers, limiting optimization effectiveness. Summary of the Invention

[0005] The purpose of the present invention is to provide a cross-border e-commerce platform based on Internet mining to solve the following technical problems:

[0006] Traditional CDN and data compression usually adopt fixed strategies and cannot be dynamically adjusted according to real-time network conditions and e-commerce business characteristics.

[0007] The purpose of the present invention can be achieved through the following technical solutions:

[0008] A cross-border e-commerce platform based on Internet mining, comprising:

[0009] A parameter acquisition module is used to obtain real-time network quality data of the geographical area where the user terminal is located, the data including the current uplink bandwidth, average data packet round trip time, and packet loss rate statistics during continuous transmission;

[0010] A data processing module is used to match a compression configuration scheme in a preset processing strategy library according to network quality data, wherein the compression configuration scheme includes a combination of text simplification rules, image conversion parameters, and video segmentation standards;

[0011] The edge computing module is used to create an isolated operating environment on the edge computing node to which the user accesses the base station and load the basic data sets and logical judgment rules required for platform business processing;

[0012] The real-time monitoring module is used to continuously collect network quality data and dynamically update the compression configuration scheme and resource allocation of edge computing nodes.

[0013] As a further solution of the present invention: the creation of an isolated operating environment specifically includes:

[0014] When a user connects to the base station for the first time, an environment initialization request is sent to the edge computing node to allocate independent memory storage space and processor operation threads;

[0015] Synchronize user historical behavior data from regional data centers, including frequently searched keywords, past order product category preferences, and payment method selection records;

[0016] Preload a collection of basic product information related to the user's current browsing session, including a product name library, price range index, and seller geographic coordinates;

[0017] Configure a session persistence mechanism to maintain the environment alive during a temporary network outage until the maximum reconnection wait time is reached.

[0018] As a further solution of the present invention: the platform business processing specifically includes:

[0019] Receive the search keyword entered by the user, perform multi-field matching in the local pre-loaded product library, and return a result set containing the intersection of product name, price range, and shipping region;

[0020] Verify the format compliance of the order information submitted by the user, check whether the product quantity meets the inventory limit, and whether the delivery address contains a valid administrative division code;

[0021] Pre-generate the framework structure of the order confirmation page, pre-load the interface verification module of the user's commonly used payment channels and the list of historical delivery addresses;

[0022] Compare the local processing results with the central server and only upload the changed data to reduce the network transmission burden.

[0023] As a further solution of the present invention: the compression configuration solution specifically includes:

[0024] Perform semantic analysis on product descriptions, extract product names, model specifications, and price ranges as retained content, and delete adjective-modifying statements and repetitive parameter descriptions;

[0025] Decompose the product display image into a basic outline layer and a detailed supplementary layer. Convert the basic outline layer into a monochrome vector graphic format, and save the detailed supplementary layer as a compressed bitmap divided into blocks by area.

[0026] Segment product demonstration videos, identify segments that showcase core features, and generate independent video streams. Replace remaining transitional frames with keyframe screenshots and audio streams.

[0027] During data processing, a hierarchical identifier is added to each content element to ensure that the terminal device can reorganize the page elements according to the original structure.

[0028] As a further solution of the present invention: the generation rule of the hierarchical identifier is:

[0029] Add semantic type tags to each reserved field in the text content, including product name tags, specification parameter tags, and price information tags;

[0030] Assign a basic display level code to the image base outline layer, and add secondary layer numbers to the detailed supplementary layer according to regional divisions;

[0031] Mark the timeline position information of the video key frame screenshot and establish a synchronized timestamp with the corresponding audio stream segment;

[0032] During the page reorganization phase, the original content tree structure is reconstructed based on the hierarchical identifier to ensure the integrity of the business logic.

[0033] As a further solution of the present invention: the dynamic update of the compression configuration scheme specifically includes:

[0034] When the uplink bandwidth remains below the threshold for a period of time exceeding the set time, the high-density compression mode is activated to increase the compression rate of the image detail supplement layer to the preset maximum value.

[0035] When the round-trip time of a data packet exceeds the upper limit of tolerance, forward predictive coding is enabled for the video stream to generate differential patch data packets of subsequent frames based on the content of the transmitted picture;

[0036] When a sharp increase in packet loss rate is detected, dynamic summary generation is implemented for the text content, retaining only key parameter items and hiding the expandable control for detailed descriptions;

[0037] After each policy adjustment, a configuration update instruction is sent to the user terminal, and the resource loading order and content parsing logic of the page rendering engine are modified synchronously.

[0038] As a further solution of the present invention: the forward predictive coding specifically includes:

[0039] Extract key frame features from the transmitted video stream and record the position coordinates and movement trajectory of the product in the picture;

[0040] Establish a motion vector prediction model based on the difference analysis of adjacent frames to generate a set of displacement compensation parameters for subsequent frames;

[0041] The predicted parameters are combined with the current frame image to reconstruct the approximate image content on the terminal device; when the network quality is restored, the original high-definition frame data is retransmitted to replace the predicted transition image.

[0042] As a further solution of the present invention: the continuous collection of network quality data specifically includes:

[0043] After a user session is established, a dedicated monitoring thread is created to continuously record the latency and throughput fluctuations of each node in the data transmission path.

[0044] Calculate the network quality degradation index based on real-time monitoring data, and trigger the compression strategy upgrade process when the index exceeds the warning threshold;

[0045] Establish a heartbeat detection mechanism between edge computing nodes and central servers, and count the success rate of command responses to assess the degree of backbone network congestion;

[0046] Regularly generate network quality change trend maps as a source of training data sets for subsequent policy library updates.

[0047] As a further solution of the present invention: it also includes a transmission interruption processing module, specifically including:

[0048] When the network connection is abnormally interrupted, the edge computing node saves a snapshot of the current session state and records the checksum of the transmitted data block;

[0049] After the monitoring network recovers the signal, it will give priority to retransmitting the difference data packets lost due to the interruption to keep the status synchronized with the central server;

[0050] Perform integrity checks on the incomplete pages cached by the user terminal and re-request the list of identifiers of the missing content blocks;

[0051] During the recovery transmission process, the incremental update mode is adopted to synchronize only the business data that has changed during the interruption.

[0052] Beneficial effects of the present invention:

[0053] This invention addresses the global network infrastructure disparities faced by cross-border e-commerce platforms. By acquiring network parameters in real time in the region where the user terminal is located and dynamically matching them with the optimal data processing mode, it effectively addresses the issue of platform access lag in low-bandwidth, high-latency environments. It also employs a combined compression strategy that combines text semantic structured processing, progressive image loading, and prioritized transmission of key frames in videos, significantly reducing data transmission volume while ensuring the integrity of business information. By deploying an edge computing environment and performing localized preprocessing at user access base stations, real-time operations such as search queries and transaction verification are handled locally, significantly reducing reliance on central servers and network round-trip latency. The established network status monitoring channel and dynamic update mechanism continuously optimize compression parameters and resource allocation strategies, ensuring stable service quality even when network conditions fluctuate. A specially designed hierarchical identifier system and interrupt recovery mechanism safeguard data integrity and consistency during compression transmission and edge processing, enabling the platform to provide a smooth user experience even in harsh network environments. Through end-to-end network awareness and adaptive adjustments, the entire solution achieves global optimization from data transmission to business processing, providing reliable technical support for the promotion and application of cross-border e-commerce in areas with weak infrastructure. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The present invention will be further described below with reference to the accompanying drawings.

[0055] Figure 1 It is a module schematic diagram of the present invention. DETAILED DESCRIPTION

[0056] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0057] See also Figure 1 As shown, the present invention is a cross-border e-commerce platform based on Internet mining, comprising:

[0058] The parameter acquisition module interacts with the user network module to obtain real-time network quality data for the user terminal's geographic area. This module utilizes network monitoring technology to accurately measure the current uplink bandwidth. For example, after multiple sampling and calculations, the current uplink bandwidth is determined to be 1 Mbps. Simultaneously, by sending test packets and recording the round-trip time, the average round-trip time is calculated to be 300 milliseconds. During a continuous data transmission period (e.g., 5 minutes), the number of lost packets is counted, resulting in a packet loss rate of 5% during this period. This data serves as the basis for processing in subsequent modules.

[0059] Next, the data processing module receives the network quality data from the parameter acquisition module. For example, the previously acquired data shows an uplink bandwidth of 1 Mbps, an average round-trip time of 300 milliseconds, and a packet loss rate of 5%. The data processing module matches this data with a library of preset processing policies. This library contains a variety of compression configurations tailored to different network quality conditions. These configurations are pre-defined based on extensive network testing and user experience data.

[0060] In this embodiment, based on the current network quality data, the matching compression configuration scheme is as follows: Regarding text simplification rules, for product descriptions, core product information such as product name, model, key technical parameters, and price range is retained, while modifying adjectives, repetitive marketing language, and parameter descriptions that have no substantial impact on performance improvement are deleted. For example, for a mobile phone description, "Stylish, lightweight smartphone with an ultra-high-pixel camera and fast charging function" would be simplified to "Smartphone, [specific model], [number of pixels] camera, supports [charging speed] fast charging."

[0061] Regarding image conversion parameters, product display images are broken down into a base outline layer and a detailed supplementary layer. The base outline layer is converted to a monochrome vector graphic format, significantly reducing file size while ensuring the image's essential shape and structure can be quickly loaded and displayed. The detailed supplementary layer is saved as a compressed bitmap segmented by region. Different compression ratios are applied to different regions based on the importance and visual appeal of the image content, balancing image quality and loading speed.

[0062] Regarding video segmentation standards, product demonstration videos are segmented into scenes, identifying segments showcasing core features, such as the phone's camera effects and system operation demonstrations. These segments are then generated into independent video streams. Other transitional scenes are replaced with keyframe screenshots combined with an audio stream, reducing the video data volume while ensuring users can understand the general content of the video.

[0063] Once a user successfully connects to the platform, the edge computing module begins operating. For example, if a user connects to a base station in Southeast Asia, the edge computing module creates an isolated operating environment on the edge computing node associated with that base station. When a user first connects to the base station, the platform sends an environment initialization request to the edge computing node. Upon receiving the request, the edge computing node allocates independent memory storage space for the user—assuming 512MB of memory—and a dedicated processor thread to ensure rapid response to user operations.

[0064] Next, the user's historical behavior data is synchronized from the regional data center. For example, the user's frequently searched keywords include "electronic products" and "outdoor sports equipment." Past order category preferences indicate frequent purchases of mobile phones, sports watches, and other items. Their payment method selection history indicates that they typically use e-wallets. This historical behavior data will be used for subsequent personalized services and business processing.

[0065] Next, a collection of basic product information related to the user's current browsing session is preloaded. For example, if a user enters an electronics category page, the edge computing node will preload a product name library containing names of various mobile phones, computers, headphones, and other products; a price range index to facilitate quick query of price ranges for different electronics products; and the seller's geographic coordinates so that the user can understand the product's shipping location information.

[0066] In addition, a session persistence mechanism is configured. If a user experiences a temporary network interruption while browsing, the isolated runtime environment will remain alive as long as the maximum reconnection wait time (assuming it is set to 60 seconds) is met. This ensures that when the user reconnects to the network, they can continue their previous operation without having to reload all data.

[0067] The real-time monitoring module runs continuously throughout the user's use of the platform. It creates a dedicated monitoring thread to continuously record the delay time and throughput fluctuations of each node in the data transmission path. For example, the network nodes are monitored every 10 seconds, and the delay time of each node from the user terminal to the edge computing node and then to the central server is recorded. By analyzing this data, the network quality degradation index is calculated. Assuming that the calculation of the network quality degradation index comprehensively considers factors such as delay time, packet loss rate and throughput, when the index exceeds the warning threshold (such as set to 0.8), the compression strategy upgrade process is triggered.

[0068] At the same time, a heartbeat detection mechanism is established between the edge computing node and the central server. At regular intervals (e.g., 30 seconds), the edge computing node sends a heartbeat command to the central server, which then returns a response. By counting the success rate of command responses, the degree of backbone network congestion can be assessed in real time. For example, if three out of ten consecutive heartbeat detections fail to respond, this indicates possible backbone network congestion.

[0069] The real-time monitoring module also periodically (e.g., hourly) generates network quality trend graphs that visually demonstrate how network quality changes over time, including bandwidth fluctuations, latency changes, and packet loss rate trends. These graphs serve as training datasets for subsequent policy library updates, enabling continuous optimization of the pre-set processing policy library and improving the platform's adaptability to diverse network environments.

[0070] In a preferred embodiment of the present invention, the step of creating an isolated operating environment specifically includes:

[0071] When a user first connects to a base station, the system quickly sends an environment initialization request to the edge computing node. This request acts as a gateway to a customized service channel for the user. Upon receiving the request, the edge computing node immediately allocates dedicated memory storage and processor threads for the user. The system dynamically adjusts memory allocation based on historical user behavior data and the general needs of the platform's services. For users who frequently browse a wide variety of products and perform frequent operations, a relatively large amount of memory, such as 512MB or even 1GB, is allocated. This ensures that the system can quickly load and process relevant data when browsing product details or performing complex searches, without experiencing lags due to insufficient memory. Regarding processor threads, an appropriate number of threads is allocated based on the resource availability of the edge computing node and the expected complexity of the user's operations. If users in the area of ​​the base station typically engage in high-in-time interactive operations, such as frequent searches and rapid switching between product pages, each user will be allocated 4-6 dedicated processing threads to ensure timely processing of user commands.

[0072] Subsequently, the user's historical behavior data is synchronized from the regional data center. This process plays a vital role in improving the user experience. The system will obtain the user's frequently searched keyword set. For example, if user A's past search keywords were concentrated in the fields of "smart wearable devices" and "outdoor camping supplies", the platform will prioritize displaying recommended information about related products based on these keywords the next time the user enters the platform, thereby improving the efficiency of users finding their desired products. At the same time, the user's past order product category preferences are obtained. If the user frequently purchases sports watches, smart bracelets and other products, the platform can push information about new sports watches in a targeted manner, or recommend matching peripheral products such as straps and chargers. In addition, the payment method selection record will also be synchronized. For example, if the user is accustomed to using a certain e-wallet to pay, the platform will prioritize displaying that payment method during the payment process, simplifying the payment process and improving the user's payment experience.

[0073] Next, the system preloads basic product information related to the user's current browsing session. For example, if a user enters a smart wearable device's browsing page, the system quickly preloads a library of product names, covering common smartwatches, smart bracelets, smart glasses, and other products on the market. A price range index is also preloaded, allowing users to quickly filter out products that fit their budget. For example, users can quickly see that smartwatch prices range from a few hundred to several thousand yuan. The seller's geographic coordinates are also preloaded, allowing users to intuitively see the product's shipping location when browsing products. For users who require fast shipping, they can prioritize products with a closer shipping location.

[0074] Finally, configure the session persistence mechanism. During a temporary network outage, this mechanism acts like a daemon, maintaining the environment until the maximum reconnection wait time is reached. The system sets a reasonable maximum reconnection wait time, such as 60 seconds. During this 60-second period, the user's previous browsing history, search history, and shopping cart information are temporarily stored in memory. Once the network is restored, the user can continue browsing products, placing orders, and more without having to redo previous operations. This significantly improves the user experience and reduces potential user churn caused by network outages.

[0075] In a preferred embodiment of this invention, the platform service processing specifically includes:

[0076] The first step is to receive the search keywords entered by the user. When the user enters a keyword in the search box, such as "smart watch with long battery life", the system will perform a multi-field match in the local pre-loaded product library. This matching process will not only compare the product name, but also go deep into the detailed description, technical parameters and other fields of the product. Through an efficient algorithm, the system will quickly return an intersection result set containing the product name, price range, and shipping area. For example, the returned results may include smart watches of brand A, priced between 1,000 and 1,500 yuan, and shipped from a certain city in China; smart watches of brand B, priced between 800 and 1,200 yuan, and shipped from a certain overseas region, etc., which makes it convenient for users to quickly understand the overview of products that meet their needs.

[0077] The order information submitted by the user is then verified for format compliance. This rigorous and comprehensive verification process checks whether the product quantity meets the inventory limit, preventing users from placing orders for items that exceed the available inventory and causing subsequent transaction disputes. Furthermore, the delivery address is checked to ensure that it contains a valid administrative code, ensuring that the order is delivered accurately and without error. For example, if the delivery address entered by the user lacks a key administrative code, the system will promptly prompt the user to complete it to ensure the validity of the order.

[0078] Next, the framework for the order confirmation page is pre-generated. When the user is about to submit their order, the system pre-loads the interface verification module for their frequently used payment channels and a list of historical delivery addresses. If the user frequently uses an e-wallet, the system pre-loads the interface verification module for that e-wallet to ensure quick payment processing. A list of historical delivery addresses is also pre-loaded, allowing users to directly select a previously used delivery address without having to manually re-enter it, saving time and improving shopping efficiency.

[0079] Finally, the local processing results are compared with the central server for differences. To reduce network transmission burden, the system intelligently uploads only the modified data. For example, if a user adds or removes several items from their shopping cart, the system compares the local shopping cart data with the shopping cart data stored on the central server and uploads only the added or removed items to the central server, rather than uploading the entire shopping cart data. This significantly reduces data transmission, improves system response speed, and provides users with a smoother shopping experience.

[0080] In another preferred embodiment of the present invention, the compression configuration scheme specifically includes:

[0081] For product description texts, the system uses advanced natural language processing technology to perform semantic analysis. Taking a smart sweeping robot as an example, when processing its product description text, it first uses vocabulary recognition and grammatical analysis to accurately extract the product name "smart sweeping robot", model specifications "[specific model, such as X5Pro]", and price range "[for example, 2000-3000 yuan]" as retained content. Then, it uses semantic understanding algorithms to identify and delete adjective-modified sentences, such as descriptions such as "super smart" and "super efficient cleaning", as well as repetitive parameter descriptions, such as repeated introductions to cleaning modes. Doing so not only greatly reduces the amount of text data, but also highlights the key information of the product, allowing users to obtain core content more quickly, while also improving the efficiency of data transmission in the network.

[0082] For product display images, the system decomposes them into a base outline layer and a detailed supplementary layer. The base outline layer mainly outlines the general shape and basic structure of the product. For the smart vacuum robot, the base outline layer is converted into a monochrome vector graphics format. The advantages of this format are small file size, fast network transmission speed, and no distortion regardless of scaling, ensuring that users can quickly see the basic form of the product even on low network bandwidth. The detailed supplementary layer is saved as a compressed bitmap divided into blocks by area. For example, the cleaning brush area, dust collection area, sensor area, etc. of the smart vacuum robot are divided into different blocks, and different compression ratios are used according to the importance of each area to displaying the product characteristics. For key functional areas, such as the cleaning brush area, the compression ratio is appropriately reduced to preserve more details; for relatively less important exterior decorative areas, the compression ratio is increased to further reduce the file size.

[0083] For product demonstration videos, the system will perform scene segmentation. Taking the demonstration video of a smart sweeping robot as an example, the system will use image recognition and motion analysis technology to identify screen segments that contain core functional demonstrations, such as the sweeping robot's cleaning process and automatic recharging process, and generate independent video streams for these segments. The remaining transitional scenes, such as the process of the sweeping robot moving from one room to another, are replaced with a combination of keyframe screenshots and audio streams. This ensures that users can fully understand the core functions of the product while greatly reducing the amount of video data and improving video loading speed.

[0084] In a preferred embodiment of this invention, the generation rule of the hierarchical identifier is:

[0085] For text content, semantic type tags are added to each reserved field. For the product description of the smart sweeping robot, a "product name tag" is added to the product name "Smart Sweeping Robot" so that the terminal device can quickly identify this as the core product identifier when parsing the text; a "specification parameter tag" is added to the model specification "X5Pro" to facilitate terminal devices to classify and display the specific product specifications; and a "price information tag" is added to the price range "2000-3000 yuan" so that price information can be quickly located and presented to users.

[0086] In terms of imagery, the base outline layer is assigned a base display level code. For example, the assigned code is "L1," indicating that this layer is the foundation for product image display and should be loaded and displayed first. For detail supplementary layers, secondary layer numbers are added by region. For example, the detail supplementary layer for the cleaning roller brush area is numbered "L2-1," and the one for the dust box area is numbered "L2-2." This numbering method allows the terminal device to clearly understand the location and importance of each detail layer, allowing it to load and combine them in the correct order.

[0087] For videos, keyframe screenshots are annotated with their timeline positions and synchronized with corresponding audio stream segments using timestamps. For example, in a keyframe screenshot from a demonstration video of a smart robot vacuum cleaner, if a keyframe shows the robot vacuuming a carpet, its position on the video timeline is annotated, such as "0:15-0:18." The corresponding audio stream segment is also timestamped. This ensures precise synchronization of image and sound when played on a device, providing a superior viewing experience for users.

[0088] During the page reorganization phase, the terminal device reconstructs the original content tree structure based on these hierarchical identifiers. It first loads the basic outline layer, then sequentially loads the detailed supplementary layers, according to the hierarchical code and number. For text content, it properly formats and displays product names, specifications, pricing information, and more based on semantic type tags. For videos, it accurately plays keyframe screenshots and video streams based on timeline position information and synchronization timestamps, ensuring audio and video synchronization. This ensures the integrity of business logic, ensuring that the page content users see is consistent with the original design, and providing users with accurate and complete product information.

[0089] In another preferred embodiment of the present invention, the dynamic update of the compression configuration scheme specifically includes:

[0090] The platform continuously monitors the real-time status of the uplink bandwidth. When the system detects that the uplink bandwidth remains below a pre-set threshold (e.g., 512kbps) for a period exceeding a set time (assuming it is set to 10 seconds), the system will activate high-density compression mode to reduce data transmission and ensure that key content can be delivered to the user terminal in a timely manner. In this mode, the compression rate of the image detail supplement layer will be increased to the preset maximum value. Taking a product image showing a high-end electronic product as an example, the original detail supplement layer may use a medium compression rate to balance image quality and loading speed, but in high-density compression mode, its compression rate will be greatly improved. This means that the image file size will be further reduced. Although some subtle image details may be lost, under poor network conditions, it can ensure that users can see the general appearance and key features of the product more quickly, avoiding user loss due to long loading times.

[0091] The round-trip time of packets is also a key network metric that impacts user experience. When the round-trip time exceeds the upper tolerance limit (e.g., set to 500 milliseconds), the system uses forward predictive coding (FPC) on the video stream to reduce video transmission latency. This technology generates differential patch packets for subsequent frames based on the content of the already transmitted video. For example, in a demonstration video of fitness equipment, the movement of the equipment is continuous and regular. FPC uses this regularity to analyze the transmitted video stream.

[0092] A sudden increase in packet loss can severely impact the integrity of data transmission. Once a sharp increase in packet loss is detected (e.g., a jump from 2% to 10% in a short period of time), the system implements a dynamic summary generation strategy for the text content. For product descriptions, only key parameters such as the product name, core functional parameters, and price are retained, while the expandable controls for detailed descriptions are hidden. For example, a smart home appliance's detailed functional description might include a variety of complex operating modes and technical principles. In the event of a high packet loss rate, these details are hidden, leaving only key information such as "smart home appliance [specific model], with [core functions such as intelligent temperature control and remote operation], priced at [X] yuan" displayed. This significantly reduces the amount of text data and mitigates the risk of data loss or errors due to packet loss, ensuring that users can access key product information and avoid being left in the dark due to network issues.

[0093] After each compression configuration strategy adjustment, the system promptly sends a configuration update command to the user terminal. This command acts like a key, synchronously modifying the resource loading order and content parsing logic of the page rendering engine. When high-density compression mode is enabled, the page rendering engine prioritizes loading the highly compressed image base layer, then attempts to load the detailed supplementary layer (if network conditions permit). For text display, content is displayed according to the rules generated by dynamic summaries.

[0094] In a preferred embodiment of the present invention, the forward predictive coding specifically includes:

[0095] 1. Keyframe Feature Extraction: First, the system extracts keyframe features from the transmitted video stream. For example, in a dynamic video showing a car, keyframes might be scenes showing representative actions such as the car starting, accelerating, and turning. The system records the position coordinates and motion trajectory of the car (the main product) in the image. Using image recognition technology, the system determines the car's specific position in the image, such as the car's center coordinates at (x1, y1) in the image. It then tracks its motion trajectory, finding that the car moves from the left to the right of the image over a period of time. The speed and direction are also recorded. This information provides the foundational data for subsequent predictions.

[0096] 2. Build a motion vector prediction model: The system builds a motion vector prediction model based on analysis of differences between adjacent frames. In car videos, the position and posture of the car will vary between frames. By comparing the changes in the car's position and shape in adjacent frames, the system calculates the motion vector. Using this motion vector data, combined with mathematical algorithms and machine learning models, it generates a set of displacement compensation parameters for subsequent frames. These parameters act like a set of instructions, telling the terminal device how to predict the content of subsequent frames based on the current frame.

[0097] 3. Reconstructing Approximate Image Content: The prediction parameters are combined with the current frame to reconstruct an approximate image content on the terminal device. When a user is watching an in-car video and network quality is poor, forward predictive coding is initiated. The terminal device then uses a specific image processing algorithm to generate approximate subsequent frames based on the received prediction parameters and the currently displayed image. While these images may not be as clear or accurate as the original HD frames, they can ensure video coherence under limited network conditions, allowing users to roughly understand the video content.

[0098] 4. Processing after Network Recovery: When network quality returns, the system retransmits the original HD frame data to replace the predicted transition images. During in-car video playback, if the network condition returns from poor to good, the system immediately stops using the predicted transition images and switches to transmitting the original HD video frames. This allows users to once again watch clear, smooth video content and enjoy high-quality product display services.

[0099] In another preferred embodiment of the present invention, the continuously collecting network quality data specifically includes:

[0100] Once a user establishes a session with the platform, the system immediately creates a dedicated monitoring thread. Throughout the user's use of the platform, it continuously records latency and throughput fluctuations at each node along the data transmission path. For example, when a user browses products on a cross-border e-commerce platform, data originates from a central server, passes through multiple network nodes, and ultimately reaches the user's terminal. The monitoring thread precisely records the time it takes for data to transmit at each node—the latency—as well as changes in data transmission rate over different time periods—the throughput fluctuations. This meticulous recording enables the platform to monitor data transmission performance in real time.

[0101] Based on real-time monitoring data, the system calculates a network quality degradation index. This index is a quantitative value that comprehensively considers multiple network indicators, incorporating factors such as latency, packet loss rate, and throughput to comprehensively reflect changes in network quality. For example, when latency increases, packet loss rate rises, and throughput decreases, the network quality degradation index will increase accordingly. The platform pre-sets a warning threshold. When the calculated index exceeds this threshold, it means that the network quality has deteriorated to a level that may affect the user experience. At this point, the system automatically triggers the compression strategy upgrade process. As described above, more efficient compression processing is performed on product description text, images, videos, and other data to reduce data transmission volume and ensure that users can continue to use the platform services smoothly.

[0102] The platform has established a heartbeat detection mechanism between the edge computing nodes and the central server. Every certain period of time (for example, 10 seconds), the edge computing node will send a command to the central server, and the central server will immediately return a response information after receiving the command. By calculating the success rate of command responses, the platform can assess the congestion level of the backbone network. If the command response success rate is low within a period of time (such as 1 minute), for example, only 70%, this indicates that the backbone network may be congested and data transmission may be affected. Based on this assessment result, the platform will further adjust the data transmission strategy, such as adjusting the data routing path, selecting nodes with better network conditions for data transmission, or caching some non-critical data and waiting for network conditions to improve before transmitting it.

[0103] In addition, the platform regularly generates network quality trend maps. This map visually illustrates how network quality changes over time, including latency fluctuations, throughput trends, and fluctuations in packet loss. For example, the map might show that during a certain period of time, due to the arrival of network peak hours, latency gradually increased, throughput decreased accordingly, and packet loss also increased. These maps not only help platform operators intuitively understand network conditions but also serve as a training dataset for subsequent policy library updates. By analyzing these maps, the platform can identify patterns in network quality changes and optimize the pre-set processing policy library, enabling the platform to better respond to various network conditions.

[0104] In another preferred embodiment of the present invention, a transmission interruption processing module is further included, specifically including:

[0105] When a network connection is interrupted, the edge computing node responds quickly and saves a snapshot of the current session state. This is like taking a "photo" of the current operation scene, recording the user's operating status at the time of the interruption, such as the product page the user was browsing, the items in the shopping cart, and the completed order details. The edge computing node also records the checksum of the transmitted data block. The checksum acts as a "fingerprint" of the data, verifying whether errors or corruption occurred during transmission. Saving this information provides a critical basis for subsequent recovery operations.

[0106] Once a network recovery signal is detected, the transmission interruption handling module prioritizes retransmitting any missing data packets. This ensures user data integrity and enables the platform to maintain synchronization with the central server. For example, if a network interruption occurs during an order submission, some order data may not be successfully transmitted to the central server. Once the network is restored, the system first searches for the lost data packets and resends them to ensure complete submission of the order data. During the retransmission process, the system utilizes efficient transmission algorithms to minimize retransmission time and improve recovery efficiency.

[0107] The system will perform integrity checks on incomplete pages cached by user terminals. Due to network interruptions, the user terminal may have only cached part of the page content, which may be incomplete or erroneous. The system will use a specific verification algorithm to check whether each element of the page is complete and whether the data is correct. If the page is found to be incomplete, the system will re-request the list of identifiers for the missing content blocks. This list is like a "shopping list", telling the system which missing page content needs to be obtained. In this way, the system can obtain the missing content in a targeted manner, quickly repair the incomplete page, and provide the user with a complete page display.

[0108] During the recovery process, the platform uses an incremental update model. This means the system synchronizes only the business data that changed during the outage, rather than retransmitting all data. For example, if a user added a new item to their shopping cart or modified the quantity during a network outage, upon network restoration, the system will only transmit these changes, without retransmitting the unchanged items in the shopping cart. This incremental update model significantly reduces data transmission, speeds up recovery, and improves the user experience, allowing users to quickly resume normal operations.

[0109] The above is a detailed description of an embodiment of the present invention. However, the content described is only a preferred embodiment of the present invention and should not be considered to limit the scope of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A cross-border e-commerce platform based on Internet mining, characterized by: include: A parameter acquisition module is used to obtain real-time network quality data of the geographical area where the user terminal is located, the data including the current uplink bandwidth, average data packet round trip time, and packet loss rate statistics during continuous transmission; A data processing module is used to match a compression configuration scheme in a preset processing strategy library according to network quality data, wherein the compression configuration scheme includes a combination of text simplification rules, image conversion parameters, and video segmentation standards; The edge computing module is used to create an isolated operating environment on the edge computing node to which the user accesses the base station and load the basic data sets and logical judgment rules required for platform business processing; The real-time monitoring module is used to continuously collect network quality data and dynamically update the compression configuration scheme and resource allocation of edge computing nodes.

2. A cross-border e-commerce platform based on Internet mining according to claim 1, characterized in that: The creation of an isolated operating environment specifically includes: When a user connects to the base station for the first time, an environment initialization request is sent to the edge computing node to allocate independent memory storage space and processor operation threads; Synchronize user historical behavior data from regional data centers, including frequently searched keywords, past order product category preferences, and payment method selection records; Preload a collection of basic product information related to the user's current browsing session, including a product name library, price range index, and seller geographic coordinates; Configure a session persistence mechanism to maintain the environment alive during a temporary network outage until the maximum reconnection wait time is reached.

3. The cross-border e-commerce platform based on Internet mining according to claim 2, characterized in that: The platform business processing specifically includes: Receive the search keyword entered by the user, perform multi-field matching in the local pre-loaded product library, and return a result set containing the intersection of product name, price range, and shipping region; Verify the format compliance of the order information submitted by the user, check whether the product quantity meets the inventory limit, and whether the delivery address contains a valid administrative division code; Pre-generate the framework structure of the order confirmation page, pre-load the interface verification module of the user's commonly used payment channels and the list of historical delivery addresses; Compare the local processing results with the central server and only upload the changed data to reduce the network transmission burden.

4. The cross-border e-commerce platform based on Internet mining according to claim 1, characterized in that: The compression configuration scheme specifically includes: Perform semantic analysis on product descriptions, extract product names, model specifications, and price ranges as retained content, and delete adjective-modifying statements and repetitive parameter descriptions; Decompose the product display image into a basic outline layer and a detailed supplementary layer. Convert the basic outline layer into a monochrome vector graphic format, and save the detailed supplementary layer as a compressed bitmap divided into blocks by area. Segment product demonstration videos, identify segments that showcase core features, and generate independent video streams. Replace remaining transitional frames with keyframe screenshots and audio streams. During data processing, a hierarchical identifier is added to each content element to ensure that the terminal device can reorganize the page elements according to the original structure.

5. The cross-border e-commerce platform based on Internet mining according to claim 4, characterized in that: The generation rule of the hierarchical identifier is: Add semantic type tags to each reserved field in the text content, including product name tags, specification parameter tags, and price information tags; Assign a basic display level code to the image base outline layer, and add secondary layer numbers to the detailed supplementary layer according to regional divisions; Mark the timeline position information of the video key frame screenshot and establish a synchronized timestamp with the corresponding audio stream segment; During the page reorganization phase, the original content tree structure is reconstructed based on the hierarchical identifier to ensure the integrity of the business logic.

6. The cross-border e-commerce platform based on Internet mining according to claim 5, characterized in that: The dynamic update of the compression configuration scheme specifically includes: When the uplink bandwidth remains below the threshold for a period of time exceeding the set time, the high-density compression mode is activated to increase the compression rate of the image detail supplement layer to the preset maximum value. When the round-trip time of a data packet exceeds the upper limit of tolerance, forward predictive coding is enabled for the video stream to generate differential patch data packets of subsequent frames based on the content of the transmitted picture; When a sharp increase in packet loss rate is detected, dynamic summary generation is implemented for the text content, retaining only key parameter items and hiding the expandable control for detailed descriptions; After each policy adjustment, a configuration update instruction is sent to the user terminal, and the resource loading order and content parsing logic of the page rendering engine are modified synchronously.

7. The cross-border e-commerce platform based on Internet mining according to claim 6, characterized in that: The forward predictive coding specifically includes: Extract key frame features from the transmitted video stream and record the position coordinates and movement trajectory of the product in the picture; Establish a motion vector prediction model based on the difference analysis of adjacent frames to generate a set of displacement compensation parameters for subsequent frames; The predicted parameters are combined with the current frame image to reconstruct the approximate image content on the terminal device; when the network quality is restored, the original high-definition frame data is retransmitted to replace the predicted transition image.

8. The cross-border e-commerce platform based on Internet mining according to claim 1, characterized in that: The continuous collection of network quality data specifically includes: After a user session is established, a dedicated monitoring thread is created to continuously record the latency and throughput fluctuations of each node in the data transmission path. Calculate the network quality degradation index based on real-time monitoring data, and trigger the compression strategy upgrade process when the index exceeds the warning threshold; Establish a heartbeat detection mechanism between edge computing nodes and central servers, and count the success rate of command responses to assess the degree of backbone network congestion; Regularly generate network quality change trend maps as a source of training data sets for subsequent policy library updates.

9. The cross-border e-commerce platform based on Internet mining according to claim 1, characterized in that: It also includes a transmission interrupt processing module, specifically including: When the network connection is abnormally interrupted, the edge computing node saves a snapshot of the current session state and records the checksum of the transmitted data block; After the monitoring network recovers the signal, it will give priority to retransmitting the difference data packets lost due to the interruption to keep the status synchronized with the central server; Perform integrity checks on the incomplete pages cached by the user terminal and re-request the list of identifiers of the missing content blocks; During the recovery transmission process, the incremental update mode is adopted to synchronize only the business data that has changed during the interruption.

Images