Page caching method and device

Abstract

The embodiment of the invention provides a page caching method and device, and the method comprises the steps: building a double-cache structure according to a current page stack and a historical page pool, building a five-layer data collection matrix, carrying out the user access behavior capture, obtaining a multi-dimensional access behavior feature, inputting the multi-dimensional feature into a three-level connection integration model, and carrying out the page access probability prediction. A predicted access page list is obtained, and predicted access pages in the predicted access page list are sorted according to the probability from high to low; extracting a complete label of a high-probability access page from a historical page pool, extracting a key resource label of a medium-probability access page from the historical page pool, extracting a cache resource label of a low-probability access page from the historical page pool, determining corresponding pre-loaded page data according to the complete label, the key resource label and the cache resource label, and storing the pre-loaded page data in a database; when the user accesses the preloaded page, page display is performed on the user based on the preloaded page data, and the user experience and the use efficiency of the webpage page can be improved.

Description

Technical Field

[0001] The present application relates to the field of data processing, and in particular to a page caching method and device. Background Art

[0002] In current Web application development, page caching technology is an important means to improve user experience and reduce server load. However, traditional page caching technology has many shortcomings, especially in terms of resource consumption and state maintenance.

[0003] On the one hand, native HTML pages require a full HTTP request to the server each time they are accessed. This mechanism not only increases the server's response burden but can also lead to longer page load times, especially over poor network conditions or when the page content is complex. This can cause users to wait longer to see the content they need, severely impacting the user experience.

[0004] On the other hand, traditional technologies cannot effectively maintain user operation records. When a user browses multiple pages in a browser and wants to return to a previous page, they often need to use the browser history or click the menu again to do so. However, once the page is refreshed, the state of the previously viewed page is lost, and the user must perform a series of operations again to restore the previous state. This lack of state maintenance not only increases user operation complexity but also may lead to loss of user information and duplication of work.

[0005] In summary, existing page caching technologies have significant shortcomings in terms of resource consumption and state maintenance, and cannot meet the high efficiency and user experience requirements of modern web applications. Therefore, a new page caching technology is urgently needed to address these issues and improve the user experience and efficiency of web pages. Summary of the Invention

[0006] In response to the problems in the prior art, the present application provides a page caching method and device, which can improve the user experience and usage efficiency of web pages.

[0007] In order to solve at least one of the above problems, the present application provides the following technical solutions:

[0008] In a first aspect, the present application provides a page caching method, comprising:

[0009] A dual cache structure is constructed based on a preset current page stack and a preset historical page pool, wherein the current page stack is used to store information about pages currently being accessed by the user, and the historical page pool is used to store information about all historical pages that the user has visited;

[0010] Construct a five-layer data collection matrix, capture the user's real-time page access behavior data based on the five-layer data collection matrix, determine the corresponding multi-dimensional features, input the multi-dimensional features into a preset Markov engine to construct a state transition matrix, determine the corresponding short-term access prediction probability based on the sliding window technology, input the multi-dimensional features into a preset time series model, determine the corresponding long-term access prediction probability, construct a policy engine based on the multi-dimensional features, determine the corresponding access behavior rules, dynamically adjust the short-term access prediction probability, the long-term access prediction probability and the access behavior rules according to a dynamic fusion algorithm, and determine the corresponding predicted access page list, wherein the predicted access pages in the predicted access page list are sorted from high to low according to probability;

[0011] Extract the complete tags of pages with high probability of access from the historical page pool, extract the key resource tags of pages with medium probability of access from the historical page pool, and extract the cache resource tags of pages with low probability of access from the historical page pool. Determine the corresponding preloaded page data based on the complete tags, the key resource tags, and the cache resource tags. When a user accesses a preloaded page, display the page to the user based on the preloaded page data.

[0012] Furthermore, the data of users' real-time page access behavior is captured according to the five-layer data collection matrix to determine corresponding multi-dimensional features, including:

[0013] Capture real-time user page access behavior data using the performance analysis interface at the page layer to determine the corresponding basic page access characteristics;

[0014] Capture real-time user page access behavior data using FormData technology in the interactive layer to determine the corresponding user micro-behavioral characteristics;

[0015] Capture real-time user page access behavior data based on SessionStorage chain storage at the time layer to determine the corresponding page access sequence characteristics;

[0016] Capture real-time user page access behavior data using the NetworkInformation API of the environment layer to determine the corresponding user terminal characteristics;

[0017] Based on the business layer's tracking SDK technology, the user's real-time page access behavior data is captured to determine the corresponding behavioral logic association characteristics.

[0018] Furthermore, the method of capturing the user's real-time page access behavior data based on the SessionStorage chain storage of the time sequence layer and determining the corresponding page access sequence characteristics includes:

[0019] According to the encrypted sessionID string, all page accesses in a single session are concatenated, and a corresponding page access sequence is determined;

[0020] According to the PrefixSpan algorithm, high-frequency sub-sequences are extracted from the page access sequence, time decay coefficients are added to the high-frequency sub-sequences after the sub-sequence extraction, and a corresponding page access sequence feature is determined.

[0021] Further, the multi-dimensional features are input into a preset Markov engine to construct a state transition matrix, and a corresponding short-term access prediction probability is determined based on a sliding window technique, which includes:

[0022] The page access sequence features are input into a preset Markov engine to construct a state transition matrix, and a corresponding processed transition probability matrix is determined based on the jump frequency statistical data in the sliding window, using Laplace smoothing to handle the zero probability problem.

[0023] According to the highest probability item of the transition probability matrix, a corresponding short-term access prediction probability is determined.

[0024] Further, the multi-dimensional features are input into a preset time series model to determine a corresponding long-term access prediction probability, which includes:

[0025] According to the attention mechanism, the page basic access features, the user micro-behavior features, and the page access sequence features are weighted and fused to determine a corresponding fusion feature.

[0026] The fusion feature is input into a preset time series model for model prediction to determine a corresponding long-term access prediction probability.

[0027] Further, the multi-dimensional features are input into a preset time series model to determine a corresponding long-term access prediction probability, which includes:

[0028] According to the user terminal features and the behavior logic association features, a corresponding user preference is determined.

[0029] According to the user preference, a strategy engine is constructed to determine a corresponding access behavior rule.

[0030] Further, the short-term access prediction probability, the long-term access prediction probability, and the access behavior rule are dynamically adjusted according to the dynamic fusion algorithm to determine a corresponding predicted access page list, which includes:

[0031] If the short-term access prediction probability, the long-term access prediction probability, and the access behavior rule appear prediction conflicts, the access behavior rule is executed preferentially to determine a corresponding predicted access page list.

[0032] If not, the short-term access prediction probability, the long-term access prediction probability and the access behavior rule are probability normalized according to Sigmoid calibration to determine a corresponding predicted access page list.

[0033] In a second aspect, the present application provides a page cache device, comprising:

[0034] a dual-cache architecture determination module, configured to construct a dual-cache structure based on a preset current page stack and a preset historical page pool, wherein the current page stack is used to store information about pages currently being accessed by a user, and the historical page pool is used to store information about all historical pages that the user has visited;

[0035] A predicted access page determination module is used to construct a five-layer data acquisition matrix, capture the user's real-time access page behavior data based on the five-layer data acquisition matrix, determine the corresponding multi-dimensional features, input the multi-dimensional features into a preset Markov engine to construct a state transition matrix, determine the corresponding short-term access prediction probability based on a sliding window technology, input the multi-dimensional features into a preset time series model, determine the corresponding long-term access prediction probability, construct a strategy engine based on the multi-dimensional features, determine the corresponding access behavior rules, dynamically adjust the short-term access prediction probability, the long-term access prediction probability and the access behavior rules according to a dynamic fusion algorithm, and determine the corresponding predicted access page list, wherein the predicted access pages in the predicted access page list are sorted from high to low according to probability;

[0036] The cache page display module is used to extract the complete tags of the high-probability access pages from the historical page pool, extract the key resource tags of the medium-probability access pages from the historical page pool, and extract the cache resource tags of the low-probability access pages from the historical page pool. The corresponding preloaded page data is determined according to the complete tags, the key resource tags and the cache resource tags. When the user accesses the preloaded page, the page is displayed to the user based on the preloaded page data.

[0037] In a third aspect, the present application provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the page caching method when executing the program.

[0038] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of the page caching method when executed by a processor.

[0039] In a fifth aspect, the present application provides a computer program product, comprising a computer program / instruction, which implements the steps of the page caching method when executed by a processor.

[0040] It can be seen from the above technical solution that the present application provides a page caching method and device, which constructs a dual cache structure based on the current page stack and the historical page pool, constructs a five-layer data acquisition matrix, captures user access behavior, obtains multi-dimensional access behavior characteristics, and inputs the multi-dimensional characteristics into a three-cascade integrated model to predict the page access probability, and obtains a predicted access page list, wherein the predicted access pages in the predicted access page list are sorted from high to low according to probability; extracts the complete label of the high-probability access page from the historical page pool, extracts the key resource label of the medium-probability access page from the historical page pool, and extracts the cache resource label of the low-probability access page from the historical page pool; determines the corresponding preloaded page data according to the complete label, key resource label and cache resource label; when the user accesses the preloaded page, the page is displayed to the user based on the preloaded page data, thereby improving the user experience and usage efficiency of the web page. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0042] Figure 1 This is one of the flow charts of the page caching method in the embodiment of the present application;

[0043] Figure 2 This is the second flow chart of the page caching method in the embodiment of the present application;

[0044] Figure 3 This is the third flow chart of the page caching method in the embodiment of the present application;

[0045] Figure 4 This is a fourth flow chart of the page caching method in an embodiment of the present application;

[0046] Figure 5 This is the fifth flow chart of the page caching method in the embodiment of the present application;

[0047] Figure 6 This is the sixth flow chart of the page caching method in the embodiment of the present application;

[0048] Figure 7 FIG7 is a flow chart of the page caching method in an embodiment of the present application;

[0049] Figure 8 is a structural diagram of a page cache device in an embodiment of the present application;

[0050] Figure 9 Schematic diagram of the structure of the electronic device in the embodiment of the present application.

[0051] Reference numerals:

[0052] Electronic device 9600, central processing unit 9100, memory 9140, communication module 9110, input unit 9120, audio processor 9130, display 9160, power supply 9170, buffer memory 9141, application / function storage unit 9142, data storage unit 9143, driver program storage unit 9144, antenna 9111, speaker 9131, microphone 9132. DETAILED DESCRIPTION

[0053] To make the purpose, technical solutions, and advantages of the embodiments of this application more clear, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0054] The acquisition, storage, use, and processing of data in this application's technical solution comply with relevant national laws and regulations.

[0055] Considering that the existing page caching technology has obvious deficiencies in resource consumption and state maintenance, and cannot meet the high requirements of modern Web applications for efficiency and user experience. The present application provides a page caching method and device, which constructs a dual cache structure based on the current page stack and the historical page pool, constructs a five-layer data acquisition matrix, captures user access behavior, obtains multi-dimensional access behavior features, inputs the multi-dimensional features into a three-cascade integrated model to predict page access probability, and obtains a predicted access page list, wherein the predicted access pages in the predicted access page list are sorted from high to low according to probability; extracts the complete tags of high-probability access pages from the historical page pool, extracts the key resource tags of medium-probability access pages from the historical page pool, and extracts the cache resource tags of low-probability access pages from the historical page pool, determines the corresponding preloaded page data according to the complete tags, key resource tags and cache resource tags, and when the user accesses the preloaded page, displays the page to the user based on the preloaded page data, thereby improving the user experience and usage efficiency of the web page.

[0056] In order to improve the user experience and usage efficiency of web pages, this application provides an embodiment of a page caching method, see Figure 1 , the page caching method specifically includes the following contents:

[0057] Step S101: constructing a dual cache structure based on a preset current page stack and a preset historical page pool, wherein the current page stack is used to store information about pages currently being accessed by the user, and the historical page pool is used to store information about all historical pages that the user has visited;

[0058] Optionally, in this embodiment, two arrays are used to manage page cache:

[0059] Current page array: The current page array is used to store information about the page currently being viewed. It stores the N most recently visited pages (N is configurable) in a stack structure. Each page is encapsulated as an independent sandbox containing a DOM snapshot, JS execution status, and resource fingerprints.

[0060] History array: The history array records information about all pages the user has browsed, manages all historical pages using the LRU (least recently used) algorithm, and supports dynamic adjustment of cache priority.

[0061] Optionally, the double buffer structure can achieve the following functions:

[0062] When a user visits a page for the first time, the system extracts the page information of the page and then dynamically renders a <iframe>Tag to load the page and add it to the top history navigation bar to generate a clickable tag. At the same time, update the current page array and history array to record the information of the page. If the user visits a page that has already been visited, the system will first check whether the information of the page exists in the cache. If it exists, the corresponding<iframe> Tags without reloading the page. This can significantly reduce repeated requests and improve browsing efficiency. When the user refreshes the page , the system will first extract the information of the current page from the local cache and dynamically generate the corresponding<iframe> This ensures that when the user refreshes the page, they will not see a white screen, but will directly see the page they were browsing.

[0065] At the same time, the system will also asynchronously rebuild the history navigation bar, generate history record tags based on the information in the history record array, and bind click events. This way, users can easily go back to pages they browsed before. In order to further improve browsing efficiency, a record is generated for each history entry .<iframe> The label is set to hidden state (display:none). When the user clicks the history label, the system simply switches<iframe> The display state of the tag is used to display the corresponding page without re-requesting the server. Based on the dual -buffer structure design of this step, this embodiment implements key functions such as page loading and cache management, refresh optimization and state recovery, and loading historical pages, achieving the goals of reducing repeated requests, supporting rapid historical record retracing, and maintaining browsing history after refreshing. This not only improves the efficiency of web browsing, but also enhances the user experience.

[0068] Step S102: Constructing a five-layer data collection matrix, capturing user real-time access page behavior data according to the five-layer data collection matrix, determining corresponding multi-dimensional features, inputting the multi-dimensional features into a preset Markov engine to construct a state transition matrix, determining corresponding short-term access prediction probability based on a sliding window technique, inputting the multi-dimensional features into a preset time sequence model to determine corresponding long-term access prediction probability, constructing a strategy engine according to the multi-dimensional features to determine corresponding access behavior rules, dynamically adjusting the short-term access prediction probability, the long-term access prediction probability and the access behavior rules according to a dynamic fusion algorithm, and determining a corresponding predicted access page list, wherein the predicted access pages in the predicted access page list are sorted in descending order of probability.

[0069] Optionally, on the basis of the double buffer structure, for the page information stored in the historical page pool, the prediction probability (high / medium / low probability) of the three-level combined model is further introduced, and the page information in the historical page pool is stored and managed in a hierarchical manner.

[0070] Optionally, in the present embodiment, first is the data set collection process, and a five-layer data collection matrix is constructed.

[0071] Specifically, the page layer collects basic access features, and the Performance API is used to collect page loading performance data, including but not limited to:

[0072] page loading time (DNS resolution, TCP connection, DOM rendering and other stage time consumption);

[0073] resource loading status (JS / CSS / image resource loading status);

[0074] URL and page identifier (used to uniquely identify the current page);

[0075] in combination with MutationObserver to monitor DOM changes to determine whether the page content is completely loaded.Specifically , the interaction layer collects user micro-behaviors, uses FormData technology to capture form submission data, and records user interaction behaviors through event monitoring (such as click, scroll, hover): Mouse trajectory ( Heatmap analysis); Click frequency (to determine the user's preference for a certain function); Scroll depth (determines whether the user has read the page completely);

[0080] Combine custom embedding points (such as data-* attributes) to mark key interactive elements. Specifically , the timing layer collects page access sequences, uses SessionStorage chain storage, and uses encrypted sessionIDs to concatenate all page accesses in a single session, recording the user's access path (e.g., A→B→C). The PrefixSpan algorithm extracts high-frequency subsequences, for example, finding that 70% of users follow the path "home→search page→product page→shopping cart." A time decay factor is added to high-frequency subpaths to ensure that recent behavior is given a higher weight. Specifically , the environment layer collects terminal and network status and uses the NetworkInformationAPI to obtain: Network type ( 4G / Wi-Fi); Network latency ( RTT );

[0085] Device memory / CPU information. Specifically , the business layer builds behavioral logic associations and collects business-related data through the embedding SDK: User login status ( VIP / ordinary user);

[0088] Business operations (payment status, browsing process); A / B testing data (behavior differences between different user groups);

[0090] Define business logic associations in conjunction with the rule engine (such as "the user must click B after clicking A").

[0091] Next, after obtaining the above-mentioned multi-dimensional features, a three-cascade model is used to predict the access page. Specifically, the Markov engine is used to extract the short-term real-time behavior probability based on the multi-dimensional features, the timing model is used to extract the long-term behavior probability based on the multi-dimensional features, and the strategy engine is used to construct a "hard conflict" decision based on the rules.

[0092] Specifically, in this step, a Markov engine is introduced to chain-store the high-frequency user access paths (such as A→B→C→A→D) recorded based on the SessionStorage of the time series layer.

[0093] First, convert the sequence into a state set {A, B, C, D}.

[0094] Then, based on a sliding window (such as the last 10 visits), the number of jumps between states is counted, and a small constant (usually 1) is added to all transition frequencies based on Laplace smoothing to avoid zero probability, and then the transition probability matrix is ​​output.

[0095] For the current state in the transition probability matrix (such as the user is visiting A), select the next state with the highest transition probability.

[0096] Example: Among the next states of A, B (0.5) has the highest probability → predict that the user will visit B next.

[0097] It is understood that Laplace smoothing solves the zero probability problem, making new page jumps predictable (e.g., the probability of A→C increases from 0 to 0.25). By limiting the amount of data through the sliding window (the amount of data can be customized), the computational overhead is reduced and the system can quickly adapt to changes in user behavior.

[0098] Specifically, in this step, a time series model is constructed to perform long-term behavior prediction based on basic page access characteristics, user micro-behavioral characteristics, and page access sequence characteristics. First , we calculate the attention score of each feature using a trainable Query-Key-Value (QKV) structure. This score can automatically adjust the weights of different features (page loading speed, user click behavior, and access order), capture the long-term dependencies of user access habits, and predict the pages that are likely to be visited in the future.

[0100] Preferably, the attention mechanism can also select multi-head attention to capture feature relationships in different dimensions.For example , if a user frequently clicks a button in the near future (micro-behavior feature weight is increased), the probability of predicting related pages is higher. If a page loads slowly (basic feature weight is increased), its preload priority is reduced.

[0102] Secondly, the model selected is LSTM (Long Short-Term Memory Network), which is suitable for capturing the long-term pattern of user access sequences. By inputting the fusion features into the pre-trained Long Short-Term Memory Network, the long-term access probability of the next N pages can be obtained.

[0103] Specifically, in this step, a policy engine is constructed. The policy engine is implemented based on static rules and statistical rules.

[0104] First, static rules are business preset rules, such as "all users must load the home page when visiting for the first time."

[0105] Statistical rules are generated by real-time analysis of user preferences, such as "90% of user A's recent visits are concentrated on module B, so B should be cached first." Specifically , user preference analysis generates a device fingerprint based on terminal features extracted from the environment layer, expressed as a 128-bit hash value, and vectorizes the device fingerprint. A spatiotemporal matrix is ​​constructed based on features extracted from the business layer. Association rules from the spatiotemporal matrix are mined using an association rule mining algorithm, and strong associations are output and recorded in chronological order as a "behavior sequence table."

[0107] User preference modeling is performed, and the feature space is combined to splice and normalize the features. The spliced ​​features are clustered and user groups are grouped based on the MiniBatch K-Means clustering algorithm.

[0108] For example, through statistical analysis, it is found that the probability of WiFi users clicking on videos is twice that of 4G users. Therefore, a rule is formulated to preload the high-definition video module for Wi-Fi users and only cache the low-resolution version for 4G users.

[0109] After the statistical rules are generated based on the user preference model, the access behavior rules are output in combination with the set static rules. Among them, the static rules are mandatory rules and are above all other rules. In other words, if all dynamic rules based on model prediction conflict with the static rules, the static rules will be given the highest priority.

[0110] Optional, examples of the above conflict scenarios:

[0111] Short-term prediction (Markov chain): The user is likely to visit page A next (probability 70%).

[0112] Long-term prediction (time series model): The user prefers page B in the long term (probability 60%).

[0113] Behavioral rules: Business mandates preloading of page C (such as a promotional page). At this point , we check whether there is a mandatory rule (such as "all users must load the activity page during the promotion period"). If so, the behavior rule is executed first (covering the predicted probability). The final effect is that page C is loaded first due to the mandatory rule, and A and B are sorted by probability.

[0115] That is, the user may visit the following pages: [C (probability 100%), A (probability 70%), B (probability 60%)] If there is no conflict, Sigmoid calibration is used to map the predictions of the short - term and long-term probabilities and rule weights for the probability of accessing the same page to a uniform range (0-1). For example, the final probability of accessing page A is: SigmoidA = (w1*short-term probability + w2*long-term probability + w3*rule weight)

[0118] Based on the final probability of each page, the final access page list is determined from high to low.

[0119] It can be understood that in this embodiment, short-term prediction, long-term prediction and behavioral rules are combined to predict the user's behavior of accessing pages, and finally a list of pages that the user may visit in the future is generated and sorted by probability (such as: A (70%) > B (30%)).

[0120] In this way, the foundation is laid for subsequently loading the page information stored in the historical page pool in a hierarchical priority manner according to the access prediction probability.

[0121] Step S103: Extract the complete tags of the pages with high probability of access from the historical page pool, extract the key resource tags of the pages with medium probability of access from the historical page pool, and extract the cache resource tags of the pages with low probability of access from the historical page pool. Determine the corresponding preloaded page data based on the complete tags, the key resource tags and the cache resource tags. When the user accesses the preloaded page, display the page to the user based on the preloaded page data.

[0122] Optionally, in this embodiment, based on the prediction results and resource costs, the historical page pool is preloaded in three levels. The granularity of the cache resources is dynamically adjusted according to the page access probability to achieve an optimal balance between resource usage and user experience.Specifically , from the historical page pool (storing all visited pages), pages are divided into three categories according to the predicted access probability, and cache tags of different levels are extracted: High probability page: Use document.cloneNode(true) to deep copy the DOM tree and save the complete HTML, CSS, JS and static resources. Completely clone the DOM and serialize it to ensure instant rendering (similar to the route snapshot of SPA). Example : the entire<iframe> A complete snapshot of the (including interaction state). Medium probability page: Pass<link rel="preload"> Preload core resources and retain only the DOM structure of the visible area (Viewport). Based on Critical Path Rendering (CPR) optimization, prioritize loading content that affects the first screen.

[0127] Example: first screen HTML, key CSS / JS, priority images. Low probability pages: Only store page meta information (URL, title) and lightweight resources (such as favicon, basic style). Dynamically load on demand to reduce memory usage.

[0129] Example: Skeleton Screen or blank placeholder.

[0130] The data extracted with high / medium probability is stored as structured data using IndexedDB (supports large capacity), and the data extracted with low probability is stored in localStorage (lightweight and fast access). When a user accesses a preloaded page, the system responds as follows :

[0132] Matching probability level: Query preloaded data to determine whether the page belongs to the high / medium / low level. On -demand rendering:

[0134] High probability: directly inject the cached complete DOM and restore the interactive state (such as Vue / React's Hydration).

[0135] Medium probability: display key resources first, and load the remaining content asynchronously in the background.

[0136] Low probability: Display the skeleton screen and initiate a real-time request to fill in the content.

[0137] Downgrade strategy: fallback to regular network request if preloaded data is invalid (e.g. version update).

[0138] Based on the hierarchical loading strategy described above, the application realizes the instantaneous loading of high-frequency pages from the historical page pool and the on-demand request of low-frequency pages. At the same time, the DOM and state synchronization cache of high-probability pages solve the SPA refresh white screen problem (better than traditional Service Worker cache). Through the probability-driven hierarchical preloading, a balance is achieved between resource efficiency and user experience, especially suitable for high-frequency interactive web applications.

[0139] This example shows how the embodiment constructs a double cache structure, obtains the user access page probability based on a five-layer data collection matrix and a prediction model, and performs hierarchical preloading on the double cache structure based on the access page probability, thereby improving the user experience and use efficiency of the web page.

[0140] From the above description, it can be known that the page cache method provided by the embodiment can construct a double cache structure according to the current page stack and the historical page pool, construct a five-layer data collection matrix, capture user access behavior, obtain multi-dimensional access behavior features, input the multi-dimensional features into a three-level integrated model for page access probability prediction, obtain a predicted access page list, wherein the predicted access pages in the predicted access page list are sorted from high to low according to probability; extract the complete tags of high-probability access pages from the historical page pool, extract the key resource tags of medium-probability access pages from the historical page pool, and extract the cache resource tags of low-probability access pages from the historical page pool, determine the corresponding preloaded page data according to the complete tags, the key resource tags and the cache resource tags, and when the user accesses the preloaded page, perform page display for the user based on the preloaded page data, thereby improving the user experience and use efficiency of the web page.

[0141] In one embodiment of the page caching method of the present application, referring to FIG. 2 , the method may further specifically include the following contents:

[0142] Step S201: Capture user's real-time page access behavior data based on the performance analysis interface of the page layer to determine the corresponding basic page access characteristics;

[0143] Step S202: Capture the user's real-time page access behavior data using the FormData technology of the interaction layer to determine the corresponding user micro-behavior characteristics;

[0144] Step S203: Capture the user's real-time page access behavior data based on the SessionStorage chain storage of the time sequence layer to determine the corresponding page access sequence characteristics;

[0145] Step S204: Capture the user's real-time page access behavior data based on the Network Information API of the environment layer to determine the corresponding user terminal characteristics;

[0146] Step S205: Capture the user's real-time page access behavior data based on the business layer's embedding SDK technology to determine the corresponding behavior logic association features.

[0147] Optionally, in this embodiment, this step is a data acquisition process, and a five-layer data acquisition matrix is ​​constructed: Specifically , the page layer collects basic access features and uses the Performance API to collect page loading performance data, including but not limited to: Page loading time (DNS resolution, TCP connection, DOM rendering, etc. ) Resource loading status (loading status of resources such as JS / CSS / pictures); URL and page ID (used to uniquely identify the current page);

[0152] Combined with MutationObserver to monitor DOM changes, determine whether the page content is fully loaded.Specifically , the interaction layer collects user micro-behaviors, uses FormData technology to capture form submission data, and records user interaction behaviors through event monitoring (such as click, scroll, hover): Mouse trajectory ( Heatmap analysis); Click frequency (to determine the user's preference for a certain function); Scroll depth (determines whether the user has read the entire page);

[0157] Combine custom embedding points (such as data-* attributes) to mark key interactive elements. Specifically , the timing layer collects page access sequences, uses SessionStorage chain storage, and uses encrypted sessionIDs to concatenate all page accesses in a single session, recording the user's access path (e.g., A→B→C). The PrefixSpan algorithm extracts high-frequency subsequences. For example, it was found that 70% of users follow the path "home→search page→product page→shopping cart." A time decay factor is added to high-frequency subpaths to ensure that recent behavior is given a higher weight. Specifically , the environment layer collects terminal and network status using the NetworkInformation API: Network type ( 4G / Wi-Fi); Network latency ( RTT );

[0162] Device memory / CPU information. Specifically , the business layer builds behavioral logic associations and collects business-related data through the embedding SDK: User login status ( VIP / ordinary user);

[0165] Business operations (payment status, browsing process); A / B testing data (behavior differences between different user groups);

[0167] Define business logic associations in conjunction with the rule engine (such as "the user must click B after clicking A").

[0168] Through step S205, this embodiment constructs a five-layer data collection matrix for collecting multi-dimensional data, laying the foundation for subsequent prediction of the probability of users visiting pages.

[0169] In one embodiment of the page caching method of the present application, referring to FIG3 , the method may further specifically include the following contents:

[0170] Step S301: Concatenate all page accesses of a single session according to the encrypted session ID to determine a corresponding page access sequence;

[0171] Step S302: extract high-frequency subsequences from the page access sequence according to the PrefixSpan algorithm, add a time attenuation coefficient to the high-frequency subsequence after the subsequence extraction, and determine the corresponding page access sequence features.

[0172] Optionally, in this embodiment, the timing layer collects page access sequences, uses SessionStorage chain storage, and uses encrypted sessionID to concatenate all page accesses in a single session, recording the user access path (e.g., A→B→C). The PrefixSpan algorithm extracts high-frequency subsequences. For example, it is found that 70% of users follow the path "home→search page→product page→shopping cart". A time decay factor is added to the high-frequency subpath to ensure that the most recent behavior is given a higher weight.

[0173] Through step S302, this embodiment realizes the collection and extraction of time series data features, laying the foundation for the subsequent construction of the Markov model.

[0174] In one embodiment of the page caching method of the present application, referring to FIG4 , the method may further specifically include the following contents:

[0175] Step S401: Input the page access sequence features into a preset Markov engine to construct a state transition matrix. Based on the jump frequency statistics in the sliding window, Laplace smoothing is used to process the zero probability problem and determine the corresponding processed transition probability matrix.

[0176] Step S402: Determine the corresponding short-term access prediction probability based on the highest probability item of the transition probability matrix.

[0177] Optionally, in this embodiment, a Markov engine is introduced to chain-store the high-frequency user access paths (such as A→B→C→A→D) recorded based on the SessionStorage of the time series layer.

[0178] First, convert the sequence into a state set {A, B, C, D}.

[0179] Then, based on a sliding window (such as the last 10 visits), the number of jumps between states is counted, and a small constant (usually 1) is added to all transition frequencies based on Laplace smoothing to avoid zero probability, and then the transition probability matrix is ​​output.

[0180] For the current state in the transition probability matrix (e.g., the user is visiting A), select the next state with the highest transition probability.

[0181] Example: Among the next states of A, B (0.5) has the highest probability → predict that the user will visit B next.

[0182] It is understood that Laplace smoothing solves the zero probability problem, making new page jumps predictable (e.g., the probability of A→C increases from 0 to 0.25). By limiting the amount of data through the sliding window (the amount of data can be customized), the computational overhead is reduced, and the user behavior can be quickly adapted to changes.

[0183] Through step S402, this embodiment successfully predicts the user's short-term page access probability based on the Markov model.

[0184] In one embodiment of the page caching method of the present application, referring to FIG5 , the method may further specifically include the following contents:

[0185] Step S501: Perform feature weighted fusion on the page basic access features, the user micro-behavior features, and the page access sequence features according to the attention mechanism to determine corresponding fusion features;

[0186] Step S502: Input the fusion features into a preset time series model for model prediction to determine the corresponding long-term access prediction probability.

[0187] Optionally, in this embodiment, a time series model is constructed to perform long-term behavior prediction based on basic page access characteristics, user micro-behavioral characteristics, and page access sequence characteristics. First , the attention score of each feature is calculated using a trainable Query-Key-Value (QKV) structure. The attention score can automatically adjust the weights of different features (page loading speed, user click behavior, access order), capture the long-term dependencies of user access habits, and predict the pages that may be visited in the future.

[0189] Preferably, the attention mechanism can also select multi-head attention (Multi-HeadAttention) to capture feature relationships in different dimensions. For example , if a user frequently clicks a button in the near future (micro-behavior feature weight is increased), the probability of predicting related pages is higher. If a page loads slowly (basic feature weight is increased), its preloading priority is reduced.

[0191] Secondly, the LSTM (Long Short-Term Memory) model is selected to capture the long-term pattern of user access sequences. By inputting the fusion features into the pre-trained LSTM network, the long-term access probability of the next N pages can be obtained.

[0192] Through step S502, this embodiment successfully predicts the user's long-term page access probability based on the time series model.

[0193] In one embodiment of the page caching method of the present application, referring to FIG6 , the method may further specifically include the following contents:

[0194] Step S601: Determine corresponding user preferences based on the user terminal characteristics and the behavior logic association characteristics;

[0195] Step S602: Build a policy engine based on the user preferences and determine corresponding access behavior rules.

[0196] Optionally, in this embodiment, a policy engine is constructed. The policy engine is implemented based on static rules and statistical rules.

[0197] First, static rules are business preset rules, such as "all users must load the home page when visiting for the first time."

[0198] Statistical rules are generated by real-time analysis of user preferences, such as "90% of user A's recent visits are concentrated on module B, so B should be cached first."

[0199] Specifically, user preference analysis generates a device fingerprint based on the terminal features extracted from the environment layer, represented by a 128-bit hash value, and vectorizes the device fingerprint. Based on the features extracted from the business layer, a spatiotemporal matrix is ​​constructed. The association rules of the spatiotemporal matrix are mined using an association rule mining algorithm, and strong associations are output and recorded in chronological order as a "behavior sequence table."

[0200] User preference modeling, feature splicing and normalization are performed in the joint feature space, and the spliced ​​features are clustered and user groups are grouped based on the MiniBatch K-Means clustering algorithm.

[0201] For example, through statistical analysis, it is found that the probability of WiFi users clicking on videos is twice that of 4G users. Therefore, a rule is formulated to preload the high-definition video module for Wi-Fi users and only cache the low-resolution version for 4G users.

[0202] After the statistical rules are generated based on the user preference model, they are combined with the set static rules to output access behavior rules. Among them, static rules are mandatory rules and are above all other rules. In other words, if all dynamic rules based on model prediction conflict with static rules, the static rules will be given the highest priority.

[0203] Through step S602, this embodiment successfully formulates business rules to build a decision engine, laying the foundation for improving the accuracy of page access prediction probability.

[0204] In one embodiment of the page caching method of the present application, referring to FIG. 7 , the method may further specifically include the following contents:

[0205] Step S701: Determine whether there is a prediction conflict between the short-term access prediction probability, the long-term access prediction probability, and the access behavior rule. If so, prioritize the access behavior rule and determine the corresponding predicted access page list.

[0206] Step S702: If not, the short-term access prediction probability, the long-term access prediction probability and the access behavior rule are probability normalized according to Sigmoid calibration to determine the corresponding predicted access page list.

[0207] Optionally, in this embodiment, the conflict scenario example:

[0208] Short-term prediction (Markov chain): The user is likely to visit page A next (probability 70%).

[0209] Long-term prediction (time series model): The user prefers page B in the long term (probability 60%).

[0210] Behavioral rules: Business mandates preloading of page C (such as a promotional page).

[0211] At this point, check whether there is a mandatory rule (such as "all users must load the activity page during the promotion period"). If so, the behavior rule is executed first (covering the predicted probability). The final effect is that page C is loaded first due to the mandatory rule, and A and B are sorted by probability.

[0212] That is, the user may visit the following pages: [C (probability 100%), A (probability 70%), B (probability 60%)]

[0213] If there is no conflict, Sigmoid calibration is used to map the predictions of the short-term, long-term probabilities and rule weights for the probability of accessing the same page to a unified range (0 to 1).For example, the final probability of visiting page A is: SigmoidA = (w1*short-term probability + w2*long-term probability + w3*rule weight)

[0215] Based on the final probability of each page, the final list of visited pages is determined from high to low.

[0216] It can be understood that in this embodiment, short-term prediction, long-term prediction and behavioral rules are combined to predict the user's behavior of accessing pages, and finally a list of pages that the user may visit in the future is generated, sorted by probability (such as: A (70%) > B (30%)).

[0217] Through step S702, this embodiment successfully obtains a list of access pages sorted by probability, laying the foundation for subsequent hierarchical priority loading of page information stored in the historical page pool according to the access prediction probability.

[0218] In order to improve the user experience and usage efficiency of web pages, the present application provides an embodiment of a page caching device for implementing all or part of the content of the page caching method. Referring to FIG8 , the page caching device specifically includes the following content:

[0219] The form feature extraction module 10 is used to build a double buffer structure based on a preset current page stack and a preset historical page pool, wherein the current page stack is used to store the page information currently accessed by the user, and the historical page pool is used to store the historical page information of all pages accessed by the user;

[0220] A model training module 20 is used to construct a five-layer data acquisition matrix, capture the user's real-time page access behavior data according to the five-layer data acquisition matrix, determine the corresponding multi-dimensional features, input the multi-dimensional features into a preset Markov engine to construct a state transition matrix, determine the corresponding short-term access prediction probability based on a sliding window technology, input the multi-dimensional features into a preset time series model, determine the corresponding long-term access prediction probability, construct a strategy engine according to the multi-dimensional features, determine the corresponding access behavior rules, dynamically adjust the short-term access prediction probability, the long-term access prediction probability and the access behavior rules according to a dynamic fusion algorithm, and determine the corresponding predicted access page list, wherein the predicted access pages in the predicted access page list are sorted from high to low according to probability;

[0221] The prediction and maintenance optimization module 30 is used to extract the complete tags of the pages with high probability of access from the historical page pool, extract the key resource tags of the pages with medium probability of access from the historical page pool, and extract the cache resource tags of the pages with low probability of access from the historical page pool, determine the corresponding preloaded page data based on the complete tags, the key resource tags and the cache resource tags, and when the user accesses the preloaded page, display the page to the user based on the preloaded page data.

[0222] From the above description, it can be seen that the page cache device provided in the embodiment of the present application can construct a dual cache structure based on the current page stack and the historical page pool, construct a five-layer data acquisition matrix, capture user access behavior, obtain multi-dimensional access behavior characteristics, and input the multi-dimensional characteristics into the three-cascade integrated model to predict the page access probability, and obtain a predicted access page list, wherein the predicted access pages in the predicted access page list are sorted from high to low according to probability; extract the complete label of the high-probability access page from the historical page pool, extract the key resource label of the medium-probability access page from the historical page pool, and extract the cache resource label of the low-probability access page from the historical page pool; determine the corresponding preloaded page data based on the complete label, key resource label and cache resource label; when the user accesses the preloaded page, the page is displayed to the user based on the preloaded page data, thereby improving the user experience and usage efficiency of the web page.

[0223] From a hardware perspective, in order to improve the user experience and usage efficiency of web pages, the present application provides an embodiment of an electronic device for implementing all or part of the content of the page caching method, and the electronic device specifically includes the following content:

[0224] A processor, a memory, a communication interface, and a bus; wherein the processor, the memory, and the communication interface communicate with each other via the bus; the communication interface is used to implement information transmission between the page caching method and related devices such as the core business system, the user terminal, and the related database; the logic controller can be a desktop computer, a tablet computer, and a mobile terminal, etc., but the present embodiment is not limited thereto. In this embodiment, the logic controller can be implemented with reference to the embodiment of the page caching method in the embodiment, and the embodiment of the page caching method, the contents of which are incorporated herein, and repeated parts are not repeated.

[0225] It is understood that the user terminal may include a smart phone, a tablet electronic device, a network set-top box, a portable computer, a desktop computer, a personal digital assistant (PDA), a vehicle-mounted device, a smart wearable device, etc. Among them, the smart wearable device may include smart glasses, a smart watch, a smart bracelet, etc.

[0226] In actual applications, part of the page caching method can be executed on the electronic device side as described above, or all operations can be completed in the client device. The specific selection can be based on the processing capability of the client device and the limitations of the user's usage scenario. This application does not limit this.If all operations are completed in the client device, the client device may further include a processor.

[0227] The client device may have a communication module (i.e., a communication unit) that can communicate with a remote server to achieve data transmission with the server. The server may include a server on the task scheduling center side, and other implementation scenarios may also include a server on an intermediate platform, such as a server on a third-party server platform that has a communication link with the task scheduling center server. The server may include a single computer device, or a server cluster consisting of multiple servers, or a server structure of a distributed device.

[0228] FIG9 is a schematic block diagram of the system structure of an electronic device 9600 according to an embodiment of the present application. As shown in FIG9 , the electronic device 9600 may include a central processing unit 9100 and a memory 9140; the memory 9140 is coupled to the central processing unit 9100. It should be noted that FIG9 is exemplary; other types of structures may be used to supplement or replace this structure to implement telecommunication functions or other functions.

[0229] In one embodiment, the page cache method function can be integrated into the central processing unit 9100.The central processing unit 9100 may be configured to perform the following control:

[0230] Step S101: constructing a dual cache structure based on a preset current page stack and a preset historical page pool, wherein the current page stack is used to store information about pages currently being accessed by the user, and the historical page pool is used to store information about all historical pages that the user has visited;

[0231] Step S102: Construct a five-layer data acquisition matrix, capture the user's real-time page access behavior data according to the five-layer data acquisition matrix, determine the corresponding multi-dimensional features, input the multi-dimensional features into a preset Markov engine to construct a state transition matrix, determine the corresponding short-term access prediction probability based on the sliding window technology, input the multi-dimensional features into a preset time series model, determine the corresponding long-term access prediction probability, construct a strategy engine according to the multi-dimensional features, determine the corresponding access behavior rules, dynamically adjust the short-term access prediction probability, the long-term access prediction probability and the access behavior rules according to a dynamic fusion algorithm, and determine the corresponding predicted access page list, wherein the predicted access pages in the predicted access page list are sorted from high to low according to probability;

[0232] Step S103: Extract the complete tags of the pages with high probability of access from the historical page pool, extract the key resource tags of the pages with medium probability of access from the historical page pool, and extract the cache resource tags of the pages with low probability of access from the historical page pool. Determine the corresponding preloaded page data based on the complete tags, the key resource tags and the cache resource tags. When the user accesses the preloaded page, display the page to the user based on the preloaded page data.

[0233] From the above description, it can be seen that the electronic device provided by the embodiment of the present application constructs a dual cache structure based on the current page stack and the historical page pool, constructs a five-layer data acquisition matrix, captures user access behavior, obtains multi-dimensional access behavior characteristics, and inputs the multi-dimensional characteristics into the three-cascade integrated model to predict the page access probability, and obtains a predicted access page list, wherein the predicted access pages in the predicted access page list are sorted from high to low according to probability; the complete tags of the high-probability access pages are extracted from the historical page pool, the key resource tags of the medium-probability access pages are extracted from the historical page pool, and the cache resource tags of the low-probability access pages are extracted from the historical page pool; the corresponding preloaded page data is determined according to the complete tags, key resource tags and cache resource tags; when the user accesses the preloaded page, the page is displayed to the user based on the preloaded page data, thereby improving the user experience and usage efficiency of the web page.

[0234] In another embodiment, the page cache method can be configured separately from the central processing unit 9100. For example, the page cache method can be configured as a chip connected to the central processing unit 9100, and the page cache method function is implemented under the control of the central processing unit.

[0235] As shown in FIG9 , the electronic device 9600 may further include: a communication module 9110, an input unit 9120, an audio processor 9130, a display 9160, and a power supply 9170. It is worth noting that the electronic device 9600 does not necessarily include all the components shown in FIG9 ; in addition, the electronic device 9600 may also include components not shown in FIG9 , and reference may be made to the prior art.

[0236] As shown in Figure 9 , the central processing unit 9100 is sometimes also called a controller or operation control, and may include a microprocessor or other processor device and / or logic device. The central processing unit 9100 receives input and controls the operation of various components of the electronic device 9600.

[0237] The memory 9140 may be, for example, one or more of a cache, a flash memory, a hard drive, a removable medium, a volatile memory, a non-volatile memory, or other suitable devices. It may store the aforementioned information related to the failure, and may also store a program for executing the relevant information. The CPU 9100 may execute the program stored in the memory 9140 to implement information storage or processing.

[0238] The input unit 9120 provides input to the central processing unit 9100.The input unit 9120 is, for example, a key or touch input device. The power supply 9170 is used to provide power to the electronic device 9600. The display 9160 is used to display display objects such as images and text. The display can be, for example, an LCD display, but is not limited thereto.

[0239] The memory 9140 may be a solid-state memory, such as a read-only memory (ROM), a random access memory (RAM), a SIM card, or the like. Alternatively, it may be a memory that stores information even when the power is off, can be selectively erased, and is provided with more data, an example of which is sometimes referred to as an EPROM, or the like. The memory 9140 may also be some other type of device. The memory 9140 includes a buffer memory 9141 (sometimes referred to as a buffer). The memory 9140 may include an application / function storage unit 9142 for storing application programs and function programs or processes for executing the operation of the electronic device 9600 by the central processing unit 9100.

[0240] The memory 9140 may further include a data storage unit 9143 for storing data, such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. The driver storage unit 9144 of the memory 9140 may include various drivers for the electronic device for communication functions and / or for executing other functions of the electronic device (such as messaging applications, address book applications, etc.).

[0241] The communication module 9110 is a transmitter / receiver that sends and receives signals via the antenna 9111. The communication module 9110 is coupled to the central processing unit 9100 to provide input signals and receive output signals, which may be the same as the case of a conventional mobile communication terminal.

[0242] Based on different communication technologies, multiple communication modules 9110 may be provided in the same electronic device, such as a cellular network module, a Bluetooth module, and / or a wireless local area network module. The communication module 9110 is also coupled to a speaker 9131 and a microphone 9132 via an audio processor 9130 to provide audio output via the speaker 9131 and receive audio input from the microphone 9132, thereby realizing a common telecommunication function. The audio processor 9130 may include any suitable buffer, decoder, amplifier, etc. In addition, the audio processor 9130 is also coupled to the central processing unit 9100, thereby enabling recording on the machine via the microphone 9132 and enabling playback of sounds stored on the machine via the speaker 9131.

[0243] Embodiments of the present application also provide a computer readable storage medium capable of implementing all steps of the page caching method in the above-mentioned embodiments, wherein the computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement all steps of the page caching method in the above-mentioned embodiments, for example, the processor executes the computer program to implement the following steps:

[0244] Step S101: constructing a double cache structure according to a preset current page stack and a preset historical page pool, wherein the current page stack is used to store page information that a user is accessing, and the historical page pool is used to store historical page information that the user has accessed;

[0245] Step S102: constructing a five-layer data collection matrix, capturing user real-time access page behavior data according to the five-layer data collection matrix, determining corresponding multi-dimensional features, inputting the multi-dimensional features into a preset Markov engine to construct a state transition matrix, determining corresponding short-term access prediction probability based on a sliding window technology, inputting the multi-dimensional features into a preset time sequence model to determine corresponding long-term access prediction probability, constructing a strategy engine according to the multi-dimensional features to determine corresponding access behavior rules, dynamically adjusting the short-term access prediction probability, the long-term access prediction probability and the access behavior rules according to a dynamic fusion algorithm, and determining corresponding predicted access page list, wherein predicted access pages in the predicted access page list are sorted in descending order of probability;

[0246] Step S103: extracting complete tags of high-probability access pages from the historical page pool, extracting key resource tags of medium-probability access pages from the historical page pool, extracting cache resource tags of low-probability access pages from the historical page pool, determining corresponding preloading page data according to the complete tags, the key resource tags and the cache resource tags, and performing page display for a user based on the preloading page data when the user accesses a preloading page.

[0247] As can be known from the above description, the computer readable storage medium provided by the embodiment of the present application constructs a double buffer structure according to the current page stack and the historical page pool, constructs a five-layer data collection matrix, captures user access behavior, obtains multi-dimensional access behavior features, inputs the multi-dimensional features into a three-level integrated model for page access probability prediction, and obtains a predicted access page list, wherein the predicted access pages in the predicted access page list are sorted in descending order of probability; the complete tags of high-probability access pages are extracted from the historical page pool, the key resource tags of medium-probability access pages are extracted from the historical page pool, and the cache resource tags of low-probability access pages are extracted from the historical page pool; the corresponding preloaded page data is determined according to the complete tags, the key resource tags and the cache resource tags; and when the user accesses the preloaded page, the user is displayed based on the preloaded page data, thereby improving the user experience and use efficiency of the webpage.

[0248] The present application also provides a computer program product capable of implementing all steps of the page caching method described in the above embodiment, where the execution subject is a server or a client. When the computer program / instruction is executed by a processor, the computer program / instruction implements the steps of the page caching method. For example, the computer program / instruction implements the following steps:

[0249] Step S101: constructing a double cache structure based on a preset current page stack and a preset historical page pool, wherein the current page stack is used to store information about pages currently being accessed by the user, and the historical page pool is used to store information about all historical pages that the user has visited;

[0250] Step S102: Construct a five-layer data acquisition matrix, capture the user's real-time page access behavior data according to the five-layer data acquisition matrix, determine the corresponding multi-dimensional features, input the multi-dimensional features into a preset Markov engine to construct a state transition matrix, determine the corresponding short-term access prediction probability based on the sliding window technology, input the multi-dimensional features into a preset time series model, determine the corresponding long-term access prediction probability, construct a strategy engine according to the multi-dimensional features, determine the corresponding access behavior rules, dynamically adjust the short-term access prediction probability, the long-term access prediction probability and the access behavior rules according to a dynamic fusion algorithm, and determine the corresponding predicted access page list, wherein the predicted access pages in the predicted access page list are sorted from high to low according to probability;

[0251] Step S103: Extract the complete tags of the pages with high probability of access from the historical page pool, extract the key resource tags of the pages with medium probability of access from the historical page pool, and extract the cache resource tags of the pages with low probability of access from the historical page pool. Determine the corresponding preloaded page data based on the complete tags, the key resource tags and the cache resource tags. When the user accesses the preloaded page, display the page to the user based on the preloaded page data.

[0252] From the above description, it can be seen that the computer program product provided by the embodiment of the present application constructs a dual cache structure based on the current page stack and the historical page pool, constructs a five-layer data acquisition matrix, captures user access behavior, obtains multi-dimensional access behavior characteristics, and inputs the multi-dimensional characteristics into a three-cascade integrated model to predict page access probability, and obtains a predicted access page list, wherein the predicted access pages in the predicted access page list are sorted from high to low according to probability; extracts complete tags of high-probability access pages from the historical page pool, extracts key resource tags of medium-probability access pages from the historical page pool, and extracts cache resource tags of low-probability access pages from the historical page pool; determines corresponding preloaded page data based on the complete tags, key resource tags and cache resource tags; when a user accesses the preloaded page, the page is displayed to the user based on the preloaded page data, thereby improving the user experience and usage efficiency of the web page. Those skilled in the art will appreciate that embodiments of the present invention can be provided as method, device or computer program product.Therefore, the present invention can adopt the form of complete hardware embodiment, complete software embodiment or the embodiment in conjunction with software and hardware aspect.And, the present invention can adopt the form of the computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage etc.) that comprise computer-usable program code. The present invention is described with reference to the flow chart and / or the block diagram of the method, equipment (device) and computer program product according to the embodiment of the invention.It should be understood that the combination of each flow process and / or square frame and the flow process and / or square frame in the flow chart and / or the block diagram can be realized by computer program instructions.These computer program instructions can be provided to the processor of general-purpose computer, special-purpose computer, embedded processor or other programmable data processing equipment to produce a machine, so that the instruction executed by the processor of computer or other programmable data processing equipment produces the device for realizing the function specified in flow chart one flow process or multiple flow processes and / or block diagram one square frame or multiple square frames.

[0255] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device that implements the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram .

[0256] These computer program instructions may also be loaded onto a computer or other programmable data processing device so that a series of operating steps are performed on the computer or other programmable device to produce computer-implemented processing, whereby the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram .

[0257] The present invention uses specific embodiments to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.< / iframe>

Claims

1. A page caching method, characterized in that: The method comprises: A dual cache structure is constructed based on a preset current page stack and a preset historical page pool, wherein the current page stack is used to store information about pages currently being accessed by the user, and the historical page pool is used to store information about all historical pages that the user has visited; Construct a five-layer data collection matrix, capture the user's real-time page access behavior data based on the five-layer data collection matrix, determine the corresponding multi-dimensional features, input the multi-dimensional features into a preset Markov engine to construct a state transition matrix, determine the corresponding short-term access prediction probability based on the sliding window technology, input the multi-dimensional features into a preset time series model, determine the corresponding long-term access prediction probability, construct a policy engine based on the multi-dimensional features, determine the corresponding access behavior rules, dynamically adjust the short-term access prediction probability, the long-term access prediction probability and the access behavior rules according to a dynamic fusion algorithm, and determine the corresponding predicted access page list, wherein the predicted access pages in the predicted access page list are sorted from high to low according to probability; Extract the complete tags of pages with high probability of access from the historical page pool, extract the key resource tags of pages with medium probability of access from the historical page pool, and extract the cache resource tags of pages with low probability of access from the historical page pool. Determine the corresponding preloaded page data based on the complete tags, the key resource tags, and the cache resource tags. When a user accesses a preloaded page, display the page to the user based on the preloaded page data.

2. The page caching method according to claim 1, wherein: The data collection matrix is ​​used to capture the user's real-time page access behavior data and determine the corresponding multi-dimensional features, including: Capture real-time user page access behavior data using the performance analysis interface at the page layer to determine the corresponding basic page access characteristics; Capture real-time user page access behavior data using FormData technology in the interactive layer to determine the corresponding user micro-behavioral characteristics; Capture real-time user page access behavior data based on SessionStorage chain storage at the time layer to determine the corresponding page access sequence characteristics; Capture real-time user page access behavior data using the NetworkInformation API of the environment layer to determine the corresponding user terminal characteristics; Based on the business layer's tracking SDK technology, the user's real-time page access behavior data is captured to determine the corresponding behavioral logic association characteristics.

3. The page caching method according to claim 2, wherein: The method of capturing the user's real-time page access behavior data based on the SessionStorage chain storage of the time sequence layer and determining the corresponding page access sequence characteristics includes: Concatenate all page accesses of a single session based on the encrypted sessionID to determine the corresponding page access sequence; A high-frequency subsequence is extracted from the page access sequence according to the PrefixSpan algorithm, and a time attenuation coefficient is added to the high-frequency subsequence after the subsequence extraction to determine the corresponding page access sequence feature.

4. The page caching method according to claim 2, wherein: The step of inputting the multi-dimensional features into a preset Markov engine to construct a state transition matrix and determining the corresponding short-term access prediction probability based on a sliding window technique includes: Input the page access sequence features into a preset Markov engine to construct a state transition matrix, and based on the jump frequency statistics in the sliding window, use Laplace smoothing to process the zero probability problem and determine the corresponding processed transition probability matrix; The corresponding short-term access prediction probability is determined according to the highest probability item of the transition probability matrix.

5. The page caching method according to claim 2, wherein: Inputting the multi-dimensional features into a preset time series model to determine the corresponding long-term access prediction probability includes: Performing feature weighted fusion on the page basic access features, the user micro-behavior features, and the page access sequence features according to the attention mechanism to determine corresponding fusion features; The fusion features are input into a preset time series model for model prediction to determine the corresponding long-term access prediction probability.

6. The page caching method according to claim 2, wherein: The step of constructing a policy engine based on the multi-dimensional features and determining corresponding access behavior rules includes: Determining corresponding user preferences based on the user terminal characteristics and the behavioral logic association characteristics; A policy engine is constructed based on the user preferences to determine corresponding access behavior rules.

7. The page caching method according to claim 2, wherein: The dynamically adjusting the short-term access prediction probability, the long-term access prediction probability, and the access behavior rule according to the dynamic fusion algorithm to determine a corresponding predicted access page list includes: Determine whether there is a prediction conflict between the short-term access prediction probability, the long-term access prediction probability, and the access behavior rule; if so, prioritize the access behavior rule and determine a corresponding predicted access page list; If not, the short-term access prediction probability, the long-term access prediction probability and the access behavior rule are probability normalized according to Sigmoid calibration to determine a corresponding predicted access page list.

8. A page cache device, characterized in that: The device comprises: a dual-cache architecture determination module, configured to construct a dual-cache structure based on a preset current page stack and a preset historical page pool, wherein the current page stack is used to store information about pages currently being accessed by a user, and the historical page pool is used to store information about all historical pages that the user has visited; A predicted access page determination module is used to construct a five-layer data acquisition matrix, capture the user's real-time access page behavior data based on the five-layer data acquisition matrix, determine the corresponding multi-dimensional features, input the multi-dimensional features into a preset Markov engine to construct a state transition matrix, determine the corresponding short-term access prediction probability based on a sliding window technology, input the multi-dimensional features into a preset time series model, determine the corresponding long-term access prediction probability, construct a strategy engine based on the multi-dimensional features, determine the corresponding access behavior rules, dynamically adjust the short-term access prediction probability, the long-term access prediction probability and the access behavior rules according to a dynamic fusion algorithm, and determine the corresponding predicted access page list, wherein the predicted access pages in the predicted access page list are sorted from high to low according to probability; The cache page display module is used to extract the complete tags of the high-probability access pages from the historical page pool, extract the key resource tags of the medium-probability access pages from the historical page pool, and extract the cache resource tags of the low-probability access pages from the historical page pool. The corresponding preloaded page data is determined according to the complete tags, the key resource tags and the cache resource tags. When the user accesses the preloaded page, the page is displayed to the user based on the preloaded page data.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the page caching method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the page cache method according to any one of claims 1 to 7 are implemented.

Images