NFC chip intelligent data loading method and system

Abstract

The present invention proposes an NFC chip intelligent data loading method and system, the method includes: effectively extracting and modeling historical usage data to predict future usage scenarios; generating a priority list of loaded data based on scenario prediction results; adjusting the loading strategy based on real-time signal strength to ensure the loading reliability of high-priority data; quickly parsing the loaded data to ensure that the data can be used by users in a timely manner; optimizing the scenario prediction model and loading strategy based on the feedback data of the loading process to form a closed-loop optimization. The present invention achieves a comprehensive improvement in data loading efficiency and reliability by introducing intelligent scenario prediction, signal strength perception and hierarchical loading strategy.

Description

Technical Field

[0001] The present invention belongs to the technical field of NFC chip intelligent data loading, and in particular relates to an NFC chip intelligent data loading method and system. Background Art

[0002] With the widespread application of near field communication (NFC) technology, NFC chips have become an indispensable core module in payment, access control, data exchange and the Internet of Things. One of the core functions of NFC chips is to load and transmit data quickly and stably to meet the needs of users to complete identity authentication, transaction confirmation or device linkage in a short time. However, the current NFC data loading method still has many problems that need to be solved, which directly affects its performance and user experience in high-frequency usage scenarios.

[0003] Existing NFC data loading methods usually adopt a static data storage and loading strategy, that is, all possible data is written into the NFC chip memory in advance, and then loaded step by step in a fixed order when the user uses it. However, this method has the following significant defects:

[0004] Inefficiency: Existing technologies cannot dynamically adapt to usage scenarios, causing the chip to load data that is irrelevant to the current scenario in most cases, increasing loading time and energy consumption. For example, in a payment scenario, core data such as transaction keys should be loaded first, but low-priority coupons or ancillary information are loaded at the same time, affecting the response speed.

[0005] Insufficient reliability: The communication distance between NFC devices and readers is usually short and susceptible to signal interference. In complex environments (such as dense crowds or multi-device interference), signal strength fluctuations may cause loading failures or partial data loss, especially when all data is loaded at once, and the retry cost is high.

[0006] Lack of intelligence and flexibility: Existing technologies generally lack intelligent loading optimization mechanisms and cannot flexibly adjust loading strategies based on user behavior, usage frequency, or current environment. This one-size-fits-all loading method leads to a waste of resources and fails to meet users' needs for personalization and efficiency.

[0007] Hardware resource limitations: Due to the limited storage and computing power of NFC chips, loading a large amount of redundant data will occupy limited memory space, further reducing performance.

[0008] Based on the above problems, there is an urgent need for an NFC data loading method and system that can dynamically adapt to scenarios, optimize loading efficiency and improve reliability, so as to solve the bottleneck of existing technologies and meet users' needs for fast response and stable performance in high-frequency scenarios. Summary of the invention

[0009] The purpose of the present invention is to design an NFC chip intelligent data loading method and system, which achieves a comprehensive improvement in data loading efficiency and reliability by introducing intelligent scene prediction, signal strength perception and layered loading strategy.

[0010] In order to achieve the above object, a first aspect of the present invention provides an NFC chip intelligent data loading method, the method comprising:

[0011] S1. Collecting generated NFC interaction data, extracting features from the NFC interaction data and aggregating and constructing feature vectors, and constructing a scenario prediction model based on the feature vectors to predict future usage scenarios; wherein the NFC interaction data includes interaction timestamp, interaction location, device type, loaded content type, and loading result;

[0012] The scenario prediction model The multi-task temporal learning model is used for modeling, and the input is the feature vector and the scene label set. The scene label set includes payment scene and access control scene. The scene prediction model Contains the following sections:

[0013] Backbone model: Based on the LSTM structure, learn the temporal pattern of features and output intermediate states ; Classifier: A linear layer predicts the probability distribution of the next scene :

[0014] in, and are the weights and biases of the classifier;

[0015] According to the NFC scenario, add the following scenario distribution balance regularization items :

[0016] in, is the number of scene categories; For Category The predicted probability of is the balance coefficient, which is used to prevent the model from overfitting to high-frequency scenarios;

[0017] S2. Generate a priority list of loaded data based on the scenario prediction results;

[0018] S3, adjusting the priority list of loaded data according to the real-time signal strength to obtain an adjusted loading set;

[0019] S4, constructing a dynamic decoding tree structure that relies on optimization, using the dynamic decoding tree to parse the loaded data from the adjusted loading set to ensure that the data can be used by users in a timely manner; wherein the dynamic decoding tree structure , constructed as follows:

[0020] First, for each data content loaded Assign weight , defined as:

[0021] in, For data and similarity; For data The load size is used to limit the negative impact of large data on decoding priority;

[0022] According to weight Generate decoding paths from high to low, giving priority to decoding content that is highly correlated with other data and has a low loading cost;

[0023] S5. Optimize the scenario prediction model and loading strategy based on the feedback data of the loading process to form a closed-loop optimization.

[0024] In the above scheme, the NFC interaction data is divided into scene feature data sets , characterize scene features; load result dataset, record loaded performance data; data content dataset , record the characteristics of the loaded content;

[0025] Wherein, the feature vector is expressed as:

[0026] in, Represents the frequency distribution of different interaction locations within a day; Represents the distribution of device types; Represents the statistical characteristics of the loaded content; Indicates the characteristics of the loading results, including the normalized values ​​of the average latency and the loading success rate;

[0027] The optimization goal of the scenario prediction model is:

[0028] in, The loss predicted for the scenario; is the L2 regularization term; is the scene distribution balance regularization term, is the probability distribution of the current scene, is the total number of scenes.

[0029] The above scheme specifically includes:

[0030] According to the predicted probability distribution of the next scene output by the scene prediction model and the loaded content of the NFC chip, a content relevance calculation formula is designed to evaluate the matching degree between each data content and the predicted scene;

[0031] Based on relevance score Sort the data content collection of the loaded content and divide the priority data collection:

[0032] High priority data collection :satisfy ,in is the relevance score threshold;

[0033] Low priority data collection :satisfy , used for alternate loading or delayed loading;

[0034] Design a load weight function based on urgency to optimize high priority data sets Loading order, expressed as

[0035] in, Data Content The final loading weight of To balance the coefficient, the relative importance of relevance score and urgency is controlled;

[0036] is the urgency of the data content, calculated by normalizing the historical access frequency:

[0037] in, For content Historical visit frequency;

[0038] Finally, we get the high priority data set , according to the loading weight After sorting, it is used as the main loading object and the low priority data collection , alternate loading data set.

[0039] In the above scheme, the relevance score The calculation is as follows:

[0040] in, Data Content The relevance score reflects The degree of match with the predicted scenario; For the scene The predicted probability of Data Content With scene The similarity is calculated as follows:

[0041] in, and are the feature vectors of data content and scene respectively; It is a dynamic loading cost correction item designed for NFC scenarios, used to introduce constraints on loading conditions, expressed as:

[0042] in, Data Content The size of the load bytes; Data Content Average loading delay of and is a weight parameter used to balance the impact of loading cost on the relevance score.

[0043] In the above scheme, S3 specifically includes:

[0044] Based on the generated priority data set and its sorted load list , and collect signal strength in real time , mapping it to the signal status category; wherein the signal strength The data is collected through the reader’s near field sensing module, and the signal status is updated in real time. ; Wherein, the signal status categories include strong signal status, weak signal status and failure status;

[0045] According to different signal status categories, formulate specific loading strategies, including:

[0046] For the strong signal state:

[0047] By priority list Load all data contents in order;

[0048] For low priority data , introduce content loading smoothing item , optimize the continuity of loading traffic:

[0049] in, Data Content The loading size; is the traffic smoothing factor, which is used to limit the impact of low-priority content on the total loading time; The value is the largest and the loading weight is the lowest;

[0050] For the weak signal state:

[0051] Load only high priority data ;

[0052] right Optimize the loading order and calculate the cost-effectiveness ratio of data loading :

[0053] in, Data Content The loading size; is the estimated delay during loading; and To balance the parameters, the weights of loading performance and transmission efficiency are dynamically adjusted;

[0054] For the failure status:

[0055] Start the alternate loading mechanism from Filter the most urgent data subset , wherein the urgency Defined as:

[0056] in, Data Content The frequency of access history; is the urgency adjustment coefficient, weighing the weight of historical frequency and loading delay; To load data The estimated delay of

[0057] Define the loading effect feedback function to monitor the performance indicators during the loading process in real time;

[0058] Get the adjusted loading set , and dynamically update policy parameters.

[0059] In the above solution, the signal status categories specifically include:

[0060] Strong signal status: , indicating good communication quality, loading the complete data set;

[0061] Weak signal status: , indicating poor communication stability, and only high-priority data is loaded;

[0062] Failure status: , indicating that the signal is insufficient to support normal loading, and a backup strategy is required to load critical data;

[0063] The loading effect feedback function , expressed as:

[0064] in, Indicates the loading success rate, which is the ratio of successfully loaded data to attempted loaded data; Represents the average delay of the loading process; Indicates the traffic volatility of low-priority data loading, used to evaluate loading smoothness; Indicates the adjustment weight, balancing the importance of success rate, latency, and traffic fluctuation.

[0065] In the above scheme, S4 specifically includes:

[0066] Start the dynamic data loading module to load the content in the data set. During the loading process, a loading status vector is dynamically maintained. , indicating the loading status of each data:

[0067] in, , indicating the data content The loading status, Indicates successful loading. Indicates that loading failed; The adjusted data set The total amount of data in

[0068] After loading is complete, record the successfully loaded data set , remove the data that failed to load;

[0069] Based on the successfully loaded data set , build a dynamic decoding tree structure that relies on optimization , quickly decode the data content;

[0070] During the decoding process, a decoding state matrix is ​​maintained , used to record the decoding dependency and completion time of each data content:

[0071] ,in, Representation data Depends on data The time to complete decoding; the row represents the decoding target data content, and the column represents the basic content that the decoding depends on; when decoding, the time value in the matrix is ​​dynamically updated ,reduce decoding delays caused by dependency conflicts;

[0072] After decoding is completed, based on the successful loading data set and decoding completion time , calculate the decoding efficiency optimization function , evaluate the efficiency of the decoding process:

[0073] in, Data Content Priority; is the longest dependency completion time during decoding; It is the fluctuation value of the decoding time of each data in the decoding process, which is defined as the variance; is a weight parameter used to balance the impact of decoding time fluctuations;

[0074] The final output is the decoded data set And its corresponding decoding efficiency index .

[0075] In the above solution, S5 specifically includes:

[0076] The feedback vector during the acquisition loading process is expressed as:

[0077] in, is the proportion of data successfully loaded, which is defined as the ratio of the number of successfully loaded data to the total number of loaded data; is the average completion time of the loading task, in milliseconds; is the fluctuation value of the loading flow, calculated as the standard deviation of the loading time series;

[0078] Define loading effect evaluation function The multi-dimensional performance of the loading strategy is comprehensively evaluated and expressed as:

[0079] in, To adjust the weight and control the impact of each indicator on the evaluation results; if Below the set optimization threshold , enter the optimization process, Decoding efficiency index:

[0080] like If the value is higher than the threshold, the cost-effectiveness ratio of loading data is increased. The balance parameters in and , limit the loading of large data;

[0081] like Below threshold, lower the signal strength threshold , expand the loading range; if the signal strength is too low, dynamically enable the backup loading mode;

[0082] If the decoding efficiency index If it is lower than the preset value, the weight distribution strategy of the decoding path is re-evaluated to optimize the decoding order of the dependencies.

[0083] In the above solution, the S5 also includes, based on the loading feedback vector , for the scene prediction model Parameters Perform incremental update, where the incremental update formula is defined as:

[0084] in, It is a feedback-driven loss function used to adjust the adaptability of the prediction model to the current scenario; Reason The predicted probability distribution of the next scenario; is the learning rate, which controls the step size of parameter adjustment;

[0085] Among them, the feedback-driven loss function is designed as:

[0086] in, is the mean square error of the feedback vector, is the predicted feedback value; Balance the regularization term for scene distribution to ensure that the model is adaptable to various scenes; is the weight of the control regularization term.

[0087] In a second aspect of the present invention, a NFC chip intelligent data loading system is provided, the system comprising:

[0088] An NFC interaction data collection unit, configured to collect the generated NFC interaction data, extract features from the NFC interaction data, aggregate and construct a feature vector, and construct a scenario prediction model based on the feature vector to predict future usage scenarios; wherein the NFC interaction data includes an interaction timestamp, an interaction location, a device type, a loaded content type, and a loaded result;

[0089] The scenario prediction model The multi-task temporal learning model is used for modeling, and the input is the feature vector and the scene label set. The scene label set includes payment scene and access control scene. The scene prediction model Contains the following sections:

[0090] Backbone model: Based on the LSTM structure, learn the temporal pattern of features and output intermediate states ; Classifier: A linear layer predicts the probability distribution of the next scene :

[0091] in, and are the weights and biases of the classifier;

[0092] According to the NFC scenario, add the following scenario distribution balance regularization items :

[0093] in, is the number of scene categories; For Category The predicted probability of is the balance coefficient, which is used to prevent the model from overfitting to high-frequency scenarios;

[0094] An NFC scene prediction unit, used to generate a priority list of loaded data according to the scene prediction results;

[0095] An NFC data loading unit, used for adjusting a priority list of loaded data according to real-time signal strength to obtain an adjusted loading set;

[0096] The NFC data decoding unit is used to construct a dynamic decoding tree structure that relies on optimization, and use the dynamic decoding tree to parse the loaded data from the adjusted loading set to ensure that the data can be used by the user in a timely manner; wherein the dynamic decoding tree structure , constructed as follows:

[0097] First, for each data content loaded Assign weight , defined as:

[0098] in, For data and similarity; For data The load size is used to limit the negative impact of large data on decoding priority;

[0099] According to weight Generate decoding paths from high to low, giving priority to decoding content that is highly correlated with other data and has a low loading cost;

[0100] The NFC decoding system optimization unit is used to optimize the scenario prediction model and loading strategy based on the feedback data of the loading process to form a closed-loop optimization.

[0101] The beneficial technical effects of the present invention are at least as follows:

[0102] (1) The present invention aims to solve the problem that the existing technology cannot dynamically adapt to the usage scenario. The present invention analyzes the user's usage habits and current environment through a scenario prediction algorithm, determines the high-frequency usage scenarios in advance, and dynamically optimizes the loading content and order. By using time series analysis and scene feature extraction, a list of high-priority and low-priority data is generated, and only the data required for the current scenario is loaded, which significantly improves loading efficiency and reduces energy consumption.

[0103] (2) To address the problems of loading failure and low reliability caused by signal strength fluctuations, the present invention introduces a real-time signal strength sensing module to dynamically adjust the loading process according to the sensing strength. For example, when the signal is weak, high-priority data is loaded first to ensure key functions. A step-by-step loading mechanism is designed to ensure the complete transmission of core data in complex environments, while reducing the time wasted due to retries.

[0104] (3) The loading and decoding of redundant data in the prior art of the present invention will occupy a large amount of hardware resources. The present invention reduces the requirements for chip storage and computing power by introducing a fast decoding algorithm and a layered loading strategy. The efficient decoding algorithm based on the dynamic Huffman tree realizes on-demand loading and step-by-step parsing of data, reduces the internal storage pressure of the chip and improves the loading speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0105] The present invention is further described using the accompanying drawings, but the embodiments in the accompanying drawings do not constitute any limitation to the present invention. A person skilled in the art can obtain other drawings based on the following drawings without creative work.

[0106] Figure 1 This is a flow chart of the NFC chip intelligent data loading method of the present invention.

[0107] Figure 2 This is a framework diagram of the NFC chip intelligent data loading system of the present invention. DETAILED DESCRIPTION

[0108] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0109] In one or more embodiments, Figure 1 As shown, a method for intelligently loading data into an NFC chip is disclosed, and the method comprises the following steps S1-S5:

[0110] S1. Collect the generated NFC interaction data, extract features from the NFC interaction data, aggregate and construct feature vectors, and construct a scenario prediction model based on the feature vectors to predict future usage scenarios; wherein the NFC interaction data includes interaction timestamp, interaction location, device type, loaded content type and loading result.

[0111] Specifically, the real-time generated NFC interaction history data is represented as , each Contains the following fields: Indicates the interaction timestamp; Indicates the location of interaction; Indicates the device type, such as payment terminal, access control device, etc. Indicates the type of content being loaded, such as keys, authentication information, coupons, etc. Indicates the loading result, including loading success rate and latency.

[0112] Furthermore, data is stored in the following three categories:

[0113] Scene feature dataset ,Include , characterizes the scene features.

[0114] Load the result dataset ,Include , record the loaded performance data.

[0115] Data Content Dataset ,Include , records the characteristics of the loaded content.

[0116] Store data in time series format and group them into different time periods (for example, by day).

[0117] Further, Group by time, extract statistical features of each time period, and construct feature vectors :

[0118] in, Represents the frequency distribution of different interaction locations within a day, e.g. Indicates the interaction ratio of the three locations. Indicates the distribution of device types, such as the proportion of payment terminals and access control devices. Indicates the statistical characteristics of the loaded content, such as the entropy value of the content type or the total number of bytes loaded. Indicates the characteristics of the loading result, including the normalized value of the average latency and loading success rate.

[0119] Further, scenario prediction modeling:

[0120] Extracted feature data and scene label sets (Define scenario categories, such as payment scenarios, access control scenarios).

[0121] Using a multi-task temporal learning model Modeling, including the following parts:

[0122] Backbone model: Based on the LSTM structure, learn the temporal pattern of features and output intermediate states ; Classifier: A linear layer predicts the probability distribution of the next scene :

[0123] in, and are the weights and biases of the classifier.

[0124] Regularization: To adapt to NFC scenarios, add the following regularization items:

[0125] in, is the number of scene categories; Indicates category The predicted probability of Represents the balance coefficient, which is used to prevent the model from overfitting to high-frequency scenarios.

[0126] Furthermore, the optimization goal of the scenario prediction model is:

[0127] in, The loss predicted for the scenario; is the L2 regularization term; is the scene distribution balance regularization term, is the probability distribution of the current scene, is the total number of scenes.

[0128] Furthermore, the trained scene prediction model is output , which can be used in the inference phase based on the feature vector , predict the probability distribution of the next scene .Will It is stored as the core module of the system and is used for subsequent priority data generation and loading strategy adjustment.

[0129] Understandably, introducing uncommon regularization terms for NFC scenarios , so that the model can be optimized in terms of predictive ability to balance a small number of usage scenarios. It provides accurate scenario prediction for subsequent priority data generation, ensures the adaptability of loading content and scenarios, and improves the effectiveness of the overall patent solution.

[0130] S2. Generate a priority list of loaded data based on the scenario prediction results.

[0131] Specifically, use the scene prediction model output from step 1 and the current feature vector , calculate the predicted probability distribution of the next scene :

[0132] in, is the predicted probability distribution of the next scene, Representation scene The predicted probability of is the feature vector of the current time period extracted in step 1, including the frequency distribution of interaction locations , Equipment Type Statistics , Loading content statistics And load the results .

[0133] Further, based on and data content collection , design a content relevance calculation formula to evaluate the matching degree between each data content and the prediction scenario:

[0134] in, Indicates data content The relevance score reflects The degree of match with the predicted scenario; Representation scene The predicted probability of Indicates data content With scene The similarity is calculated as follows:

[0135] in and They are the feature vectors of data content and scene respectively, which come from the content statistical features and scene features extracted in step 1. Represents the dynamic loading cost correction item designed for NFC scenarios, which is used to introduce constraints on loading conditions:

[0136] in, Indicates data content The size of the load bytes; Indicates data content Average loading delay of and Represents a weight parameter that balances the impact of loading cost on the relevance score.

[0137] Furthermore, according to the relevance score Data content collection Sort and prioritize data sets:

[0138] High priority data collection :satisfy ,in is the relevance score threshold;

[0139] Low priority data collection :satisfy , used for alternate loading or delayed loading.

[0140] Furthermore, for high priority data sets Further optimize the loading order and design a loading weight function based on urgency:

[0141] in, Indicates data content The final loading weight of represents the balance coefficient, controlling the relative importance of relevance score and urgency; Indicates the urgency of the data content, calculated by normalizing the historical access frequency:

[0142] in, For content The historical visit frequency.

[0143] Final output priority load list : : By loading weight After sorting, it is used as the main loading object; : Alternate loading data collection.

[0144] It can be understood that the scenario prediction model generated in step 1 , the predicted probability of future scenarios As a direct input to the priority data generation, an innovative dynamic loading cost modifier is designed for NFC loading scenarios , effectively balances loading adaptability and actual cost, provides a clear set of loading objects and optimized loading order, and provides support for subsequent loading strategy adjustment and execution.

[0145] S3. A strategy for adjusting the priority list of loaded data according to the real-time signal strength is used to obtain an adjusted loading set.

[0146] Specifically, input the priority data set generated in step 2 and its sorted load list , and collect signal strength in real time , mapping it to a signal status category:

[0147] Strong signal status: , the communication quality is good and the complete data set can be loaded;

[0148] Weak signal status: , communication stability is poor, and only high-priority data is loaded;

[0149] Failure status: , the signal is not sufficient to support normal loading, and a backup strategy is required to load critical data.

[0150] The data is collected through the reader’s near field sensing module, and the signal status is updated in real time. .

[0151] Furthermore, specific loading strategies are formulated for different signal states:

[0152] Strong signal status ( ):

[0153] By priority list Load all data contents in order;

[0154] For low priority data , introduce content loading smoothing item , optimize the continuity of loading traffic:

[0155] in, Indicates data content The loading size; Represents the traffic smoothing factor, which is used to limit the impact of low-priority content on the total loading time; The larger the value, the lower the loading weight.

[0156] Weak signal status ( ):

[0157] Load only high priority data ;

[0158] right Optimize the loading order and calculate the cost-effectiveness ratio of data loading :

[0159] in, represents the relevance score calculated in step 2; Indicates data content The loading size; Indicates the estimated delay during loading; and Represents a balancing parameter that dynamically adjusts the weight of loading performance and transmission efficiency.

[0160] according to Sort from high to low and load them one by one to ensure that the scene adaptation utility brought by the unit loading cost is maximized.

[0161] Failed state ( ):

[0162] Start the alternate loading mechanism from Filter the most urgent data subset , urgency Defined as:

[0163] in, Indicates data content The frequency of access history; Indicates the urgency adjustment coefficient, weighing the weight of historical frequency and loading delay; Indicates loading data The estimated delay.

[0164] according to Sort by high to low and load , real-time check after each load Status, decide whether to adjust the loading strategy.

[0165] Furthermore, real-time monitoring of performance indicators during loading is performed to define loading effect feedback functions. :

[0166] in, Indicates the loading success rate, which is the ratio of successfully loaded data to attempted loaded data; Represents the average delay of the loading process; Indicates the traffic volatility of low-priority data loading, used to evaluate loading smoothness; Indicates the adjustment weight, balancing the importance of success rate, latency, and traffic fluctuation.

[0167] Further, output the adjusted load set , and dynamically update the policy parameters:

[0168] When the loading success rate is low under strong signal conditions, increase Threshold to enhance the robustness of signal classification;

[0169] In weak signal state, if the average delay is higher than the threshold, increase and , deprioritize high-cost content;

[0170] Optimize the urgency calculation when multiple loading fails in the failed state Weight.

[0171] S4. Build a dynamic decoding tree structure that depends on optimization, and use the dynamic decoding tree to parse the loaded data from the adjusted loading set to ensure that the data can be used by users in a timely manner.

[0172] Specifically, input the adjusted load set output in step 3 and signal status , start the dynamic data loading module, and load the contents of the data set one by one. During the loading process, a loading status vector is dynamically maintained , indicating the loading status of each data:

[0173] in, , indicating the data content The loading status, Indicates successful loading. Indicates that loading failed; The adjusted data set The total amount of data in .

[0174] Furthermore, after the loading is completed, the successfully loaded data set is recorded , remove the data that failed to load.

[0175] Further, based on the successfully loaded data set , build a dynamic decoding tree structure that relies on optimization , to achieve fast decoding of data content:

[0176] First, for each data content loaded Assign weight , defined as:

[0177] in, Representation data and The similarity of , based on the correlation score calculated in the previous step and Sure; Representation data The loading size is used to limit the negative impact of large data on decoding priority, according to the weight Generate decoding paths from high to low, prioritizing decoding of content that is highly correlated with other data and has a low loading cost.

[0178] Furthermore, during the decoding process, a decoding state matrix is ​​maintained , used to record the decoding dependency and completion time of each data content:

[0179] in, Representation data Depends on data The time to complete decoding; the rows represent the decoding target data content, and the columns represent the basic content that the decoding depends on;

[0180] When decoding, dynamically update the time value in the matrix ,Reducing decoding delays caused by dependency conflicts.

[0181] Further, after decoding is completed, based on the successful loading data set and decoding completion time , calculate the decoding efficiency optimization function , used to evaluate the efficiency of the decoding process:

[0182] in, Indicates data content The priority comes from the utility ratio of step 3 ; Indicates the longest dependency completion time during decoding; It indicates the fluctuation value of the decoding time of each data in the decoding process, which is defined as the variance; Represents the weight parameter, which is used to balance the impact of decoding time fluctuations. , ensuring that the loaded content is highly adapted to the scene requirements during the decoding process, while reducing the fluctuation of decoding delay.

[0183] Understandably, the final output is a data set that is successfully decoded And its corresponding decoding efficiency index , which is used for loading effect evaluation and feedback optimization in subsequent steps. If the value is lower than the preset threshold, the decoding tree structure is dynamically adjusted. The weight distribution strategy in ensures that the next loading and decoding will be more efficient.

[0184] S5. Optimize the scenario prediction model and loading strategy based on the feedback data of the loading process to form a closed-loop optimization.

[0185] Specifically, input the decoded successful data set output in step 4 , decoding efficiency index , and the feedback vector during loading :

[0186] in, Indicates the proportion of data loaded successfully, which is defined as the ratio of the number of successfully loaded data to the total number of loaded data; Indicates the average completion time of the loading task in milliseconds; represents the fluctuation value of the loading flow, calculated as the standard deviation of the loading time series; It represents the decoding efficiency index calculated in step 4, indicating the compatibility of loading and decoding.

[0187] These feedback data are used to evaluate the performance of the current loading strategy and guide subsequent optimization.

[0188] Furthermore, in order to comprehensively evaluate the multi-dimensional performance of the loading strategy, the loading effect evaluation function is defined :

[0189] in, To adjust the weight and control the impact of each indicator on the evaluation results. , get the overall effect of the current loading and decoding strategy. Below the set optimization threshold , enter the optimization process.

[0190] Furthermore, based on the loading feedback vector , for the scene prediction model Parameters Perform incremental updates. Define the incremental update formula:

[0191] in, represents the newly designed feedback-driven loss function, which is used to adjust the adaptability of the prediction model to the current scenario; Indicated by The predicted probability distribution of the next scenario; Represents the learning rate, which controls the step size of parameter adjustment.

[0192] Furthermore, the feedback-driven loss function is designed as:

[0193] in, represents the mean square error of the feedback vector, is the predicted feedback value; represents the scene distribution balance regularization term proposed in step 1, ensuring that the model is adaptable to various scenes; is the weight of the control regularization term.

[0194] Furthermore, dynamic optimization of loading strategy parameters:

[0195] Evaluation function based on loading effect As a result, the loading strategy parameters are optimized and adjusted:

[0196] if If the load priority is higher than the threshold, the utility ratio is increased. Loading cost weight in and , to limit the loading of large data.

[0197] if Below threshold, lower the signal strength threshold , expand the loading range; if the signal strength is too low, dynamically enable the backup loading mode.

[0198] like If the decoding efficiency is lower than expected, the weight distribution strategy of the decoding path is re-evaluated to optimize the decoding order of the dependencies.

[0199] Understandably, the optimized model and strategy are output:

[0200] Updated scenario prediction model , used for the next round of scene prediction;

[0201] Optimized loading strategy parameters , used for the next round of loading strategy adjustment;

[0202] Updated effect evaluation threshold , dynamically adjust the sensitivity of future optimizations.

[0203] In one or more embodiments, Figure 2 As shown, an NFC chip intelligent data loading system is disclosed, and the system includes:

[0204] The NFC interaction data collection unit 301 is used to collect the generated NFC interaction data, extract features from the NFC interaction data, aggregate and construct feature vectors, and construct a scenario prediction model based on the feature vectors to predict future usage scenarios; wherein the NFC interaction data includes interaction timestamp, interaction location, device type, loaded content type and loading result;

[0205] The scenario prediction model The multi-task temporal learning model is used for modeling, and the input is the feature vector and the scene label set. The scene label set includes payment scene and access control scene. The scene prediction model Contains the following sections:

[0206] Backbone model: Based on the LSTM structure, learn the temporal pattern of features and output intermediate states ; Classifier: A linear layer predicts the probability distribution of the next scene :

[0207] in, and are the weights and biases of the classifier;

[0208] According to the NFC scenario, add the following scenario distribution balance regularization items :

[0209] in, is the number of scene categories; For Category The predicted probability of is the balance coefficient, which is used to prevent the model from overfitting to high-frequency scenarios;

[0210] The NFC scene prediction unit 302 is used to generate a priority list of loaded data according to the scene prediction result;

[0211] The NFC data loading unit 303 is used to adjust the priority list of loaded data according to the real-time signal strength to obtain an adjusted loading set;

[0212] The NFC data decoding unit 304 is used to construct a dynamic decoding tree structure that relies on optimization, and use the dynamic decoding tree to parse the loaded data from the adjusted loading set to ensure that the data can be used by the user in a timely manner; wherein the dynamic decoding tree structure , constructed as follows:

[0213] First, for each data content loaded Assign weight , defined as:

[0214] in, For data and similarity; For data The load size is used to limit the negative impact of large data on decoding priority;

[0215] According to weight Generate decoding paths from high to low, giving priority to decoding content that is highly correlated with other data and has a low loading cost;

[0216] The NFC decoding system optimization unit 305 is used to optimize the scenario prediction model and the loading strategy based on the feedback data of the loading process to form a closed-loop optimization.

[0217] It is worth noting that the specific working process of the NFC chip intelligent data loading system provided in the embodiment of the present invention is the same as the working process of the NFC chip intelligent data loading method described in the above embodiment, and will not be described in detail here.

[0218] Compared with the prior art, the NFC chip intelligent data loading system provided by the embodiment of the present invention collects generated NFC interaction data, extracts and models historical usage data to construct feature vectors, and constructs a scenario prediction model based on the feature vectors to predict future usage scenarios; generates a priority list of loaded data based on the scenario prediction results; adjusts the strategy of the loaded data priority list based on the real-time signal strength to obtain an adjusted loading set; constructs a dynamic decoding tree structure that depends on optimization, and uses the dynamic decoding tree to parse the loaded data from the adjusted loading set to ensure that the data can be used by users in a timely manner; optimizes the scenario prediction model and the loading strategy based on the feedback data of the loading process to form a closed-loop optimization.

[0219] The embodiment of the present invention also provides an NFC chip intelligent data loading device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the steps in the above-mentioned NFC chip intelligent data loading method embodiment are implemented, for example Figure 1 or, the processor implements the functions of each module in the above-mentioned system embodiments when executing the computer program.

[0220] Exemplarily, the computer program may be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules may be a series of computer program instruction segments capable of completing specific functions, which are used to describe the execution process of the computer program in the NFC chip smart data loading device.

[0221] The NFC chip intelligent data loading device can be a computing device such as a desktop computer, a notebook, a PDA, and a cloud server. The NFC chip intelligent data loading device may include, but is not limited to, a processor and a memory. Those skilled in the art will appreciate that the NFC chip intelligent data loading device may also include an input and output device, a network access device, a bus, etc.

[0222] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASAC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the NFC chip intelligent data loading device, and uses various interfaces and lines to connect various parts of the entire NFC chip intelligent data loading device.

[0223] The memory can be used to store the computer program and / or module, and the processor realizes various functions of the NFC chip intelligent data loading device by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created according to the operation of the air-conditioning controller, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (SmartMedaaCard, SMC), a secure digital (SecureDagatal, SD) card, a flash card (FlashCard), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0224] Wherein, if the module integrated in the NFC chip intelligent data loading device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device that can carry the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal and software distribution medium, etc.

[0225] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0226] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. NFC chip intelligent data loading method, characterized in that: The method comprises: S1. Collecting generated NFC interaction data, extracting features from the NFC interaction data and aggregating and constructing feature vectors, and constructing a scenario prediction model based on the feature vectors to predict future usage scenarios; wherein the NFC interaction data includes interaction timestamp, interaction location, device type, loaded content type, and loading result; The scene prediction model M scene A multi-task temporal learning model is used for modeling, with the input being a feature vector and a scene label set {S t }, the scene label set includes payment scene and access control scene, the scene prediction model M scene Contains the following sections: Backbone model: Based on the LSTM structure, learn the temporal pattern of features and output the intermediate state h t ; Classifier: A linear layer predicts the probability distribution p of the next scene t+1 : p t+1 =softmax(W h ·h t +b h ) Among them, W h and b h are the weights and biases of the classifier; According to the NFC scenario, add the following scenario distribution balance regularization items Where k is the number of scene categories; P(S i ) is the predicted probability of category i; α is the balance coefficient, which is used to prevent the model from overfitting to high-frequency scenarios; S2. Generate a priority list of loaded data based on the scenario prediction results; S3, adjusting the priority list of loaded data according to the real-time signal strength to obtain an adjusted loading set; S4, constructing a dynamic decoding tree structure that relies on optimization, using the dynamic decoding tree to parse the loaded data from the adjusted loading set to ensure that the data can be used by users in a timely manner; wherein the dynamic decoding tree structure The build is as follows: First, for each data content c loaded i Assign weight w ij , defined as: Among them, Sim(c i ,c j ) is data c i With c j Similarity; Size(c j ) is data c j The load size is used to limit the negative impact of large data on decoding priority; According to the weight w ij Generate decoding paths from high to low, giving priority to decoding content that is highly correlated with other data and has a low loading cost; S5. Optimize the scenario prediction model and loading strategy based on the feedback data of the loading process to form a closed-loop optimization.

2. The NFC chip intelligent data loading method according to claim 1, characterized in that: The NFC interaction data is divided into a scene feature data set D s , characterizes the scene features; Load the result data set and record the loaded performance data; data content data set D c , record the characteristics of the loaded content; Wherein, the feature vector is expressed as: x t =[f t (L),f t (T),f t (C),f t (r)] Among them, f t (L) represents the frequency distribution of different interaction locations within a day; f t (T) represents the distribution of device types; f t (C) represents the statistical characteristics of the loaded content; f t (r) represents the loading result characteristics, including the normalized values ​​of the average delay and the loading success rate; The optimization goal of the scenario prediction model is: Among them, CrossEntropy is the loss of scene prediction; ∥θ∥ 2 is the L2 regularization term; is the scene distribution balance regularization term, p t is the probability distribution of the current scene, and T is the total number of scenes.

3. The NFC chip intelligent data loading method according to claim 1, characterized in that: The S2 specifically includes: According to the predicted probability distribution of the next scene output by the scene prediction model and the loaded content of the NFC chip, a content relevance calculation formula is designed to evaluate the matching degree between each data content and the predicted scene; According to the correlation score R(c j ) Sort the data content collection of the loaded content and divide the data collection into priority groups: High priority data set D high :Satisfy R(c j )≥δ, where δ is the correlation score threshold; Low priority data set D low :Satisfy R(c j )<δ, for alternate loading or delayed loading; Design a loading weight function based on urgency to optimize the high priority data set D high Loading order, expressed as W(c j )=β·R(c j )+(1-β)·U(c j ) Among them, W(c j ) is the data content c j The final loading weight of the ; β is the balance coefficient, which controls the relative importance of the relevance score and the urgency; U(c j ) is the urgency of the data content, which is calculated by normalizing the historical access frequency: Among them, Freq(c j ) is the content c j Historical visit frequency; Finally, we get the high priority data set D high , according to the loading weight W(c j ) are sorted as the main loading object, and the low priority data set D low , and the low priority data set D low As an alternative, load the data collection.

4. The NFC chip intelligent data loading method according to claim 3 is characterized in that: The relevance score R(c j ) is calculated as follows: Among them, R(c j ) is the data content c j The correlation score reflects c j The degree of match with the predicted scenario; P(S i ) is scene S i The predicted probability of Sim(c j ,S i ) is the data content c j With scene S i The similarity is calculated as follows: Among them, v(c j ) and v(S i ) are the feature vectors of data content and scene respectively; φ(c j ,S i ) is a dynamic loading cost correction term designed for NFC scenarios, which is used to introduce constraints on loading conditions and is expressed as: Among them, Size(c j ) is the data content c j Load byte size; Delay(c j ) is the data content c j The average loading delay; λ and μ are weight parameters used to balance the impact of loading cost on the relevance score.

5. The NFC chip intelligent data loading method according to claim 3, characterized in that: The S3 specifically includes: According to the generated priority data set {D high ,D low } and its sorted load list D final , and collect the signal strength RSSI in real time and map it to the signal status category; wherein the signal strength RSSI is collected through the near field sensing module of the reader, and the signal status S is updated in real time rss ; Wherein, the signal status categories include strong signal status, weak signal status and failure status; According to different signal status categories, formulate specific loading strategies, including: For the strong signal state: By priority list D final Load all data contents in order; For low priority data D low , introduce the content loading smoothing term ψ(c j ), optimize the continuity of loading traffic: Among them, Size(c j ) is the data content c j The loading size; γ is the traffic smoothing coefficient, which is used to limit the impact of low-priority content on the total loading time; ψ(c j ) value is the largest and the loading weight is the lowest; For the weak signal state: Only load high priority data D high ; Right D high Optimize the loading order and calculate the data loading cost-utility ratio E(c j ): Among them, Size(c j ) is the data content c j Loading size; Delay(c j ) is the estimated delay during loading; λ and μ are balancing parameters that dynamically adjust the weights of loading performance and transmission efficiency; For the failure status: Start the backup loading mechanism from D high Filter the data subset D with the highest urgency emg , where the urgency level U(c j ) is defined as: Among them, Freq(c j ) is the data content c j The access history frequency; ρ is the urgency adjustment coefficient, weighing the weight of the historical frequency and loading delay; Delay(c j ) is to load data c j The estimated delay of Define the loading effect feedback function to monitor the performance indicators during the loading process in real time; Get the adjusted loading set D' final , and dynamically update policy parameters.

6. The NFC chip intelligent data loading method according to claim 5, characterized in that: The signal status categories specifically include: Strong signal status: RSSI ≥ T high , indicating good communication quality, loading the complete data set; Weak signal status: T low ≤RSSI <T high , indicating poor communication stability, and only high-priority data is loaded; Failure status: RSSI <T low , indicating that the signal is insufficient to support normal loading, and a backup strategy is required to load critical data; The loading effect feedback function Φ is expressed as: Among them, SuccessRate represents the loading success rate, which is the ratio of successfully loaded data to attempted loaded data; AvgDelay represents the average delay of the loading process; LoadVariation represents the traffic volatility of low-priority data loading, which is used to evaluate loading smoothness; α, β, and γ represent adjustment weights, balancing the importance of success rate, delay, and traffic fluctuation.

7. The NFC chip intelligent data loading method according to claim 5, characterized in that: The S4 specifically includes: Start the dynamic data loading module to load the content in the data set. During the loading process, a loading state vector z is dynamically maintained. t , indicating the loading status of each data: With t =[z1,z2,…,z i ,With n ] Among them, z i ∈{0,1}, indicating data content c i The loading state, z i =1 means loading is successful, z i =0 indicates loading failure; n is the adjusted data set D′ final The total amount of data in After loading is complete, record the successfully loaded data set Eliminate data that failed to load; Based on the successfully loaded data set Constructing a dynamic decoding tree structure that relies on optimization Quickly decode the data content; During the decoding process, a decoding state matrix M is maintained decode , used to record the decoding dependency and completion time of each data content: M decode =[t ij ], where t ij Represents data c i Depends on data j The time to complete decoding; the row represents the decoding target data content, and the column represents the basic content that the decoding depends on; when decoding, the time value t in the matrix is ​​dynamically updated ij ,reduce decoding delays caused by dependency conflicts; After decoding is completed, based on the successful loading data set and decoding completion time t ij , calculate the decoding efficiency index Ψ of the decoding efficiency optimization function and evaluate the efficiency of the decoding process: Among them, Priority(c i ) is the data content c i The priority of max(t ij ) is the longest dependency completion time in the decoding process; DelayVariation is the fluctuation value of the decoding time of each data in the decoding process, defined as the variance; ∈ is the weight parameter, which is used to balance the impact of the decoding time fluctuation; The final output is the decoded data set And its corresponding decoding efficiency index Ψ.

8. The NFC chip intelligent data loading method according to claim 7, characterized in that: The S5 specifically includes: The feedback vector during the acquisition loading process is expressed as: f t =[SuccessRate,AvgDelay,LoadVariation,Ψ] Among them, SuccessRate is the ratio of successfully loaded data, which is defined as the ratio of the number of successfully loaded data to the total number of loaded data; AvgDelay is the average completion time of the loading task, in milliseconds; LoadVariation is the fluctuation value of the loading traffic, calculated as the standard deviation of the loading time series; Define loading effect evaluation function The multi-dimensional performance of the loading strategy is comprehensively evaluated and expressed as: Among them, θ, μ,δ are adjustment weights to control the impact of each indicator on the evaluation results; Below the set optimization threshold Enter the optimization process, Ψ is the decoding efficiency indicator: If AvgDelay is higher than the threshold, then the cost-effectiveness ratio of data loading is increased. j ) to limit the loading of big data; If SuccessRate is lower than the threshold, lower the signal strength threshold T low , expand the loading range; if the signal strength is too low, dynamically enable the backup loading mode; If the decoding efficiency index Ψ is lower than the preset value, the weight allocation strategy of the decoding path is re-evaluated to optimize the decoding order of the dependency.

9. The NFC chip intelligent data loading method according to claim 8, characterized in that: The S5 also includes, based on the loading feedback vector f t , for the scene prediction model M scene The parameter θ is incrementally updated, where the incremental update formula is defined as: in, is a feedback-driven loss function used to adjust the adaptability of the prediction model to the current scenario; t+1 For M scene The predicted probability distribution of the next scene; η is the learning rate, which controls the step size of parameter adjustment; Among them, the feedback-driven loss function is designed as: in, is the mean square error of the feedback vector, is the predicted feedback value; is the regularization term for balancing scene distribution to ensure that the model is adaptable to various scenes; τ is the weight for controlling the regularization term.

10. NFC chip intelligent data loading system, characterized in that: The system comprises: An NFC interaction data collection unit, configured to collect the generated NFC interaction data, extract features from the NFC interaction data, aggregate and construct a feature vector, and construct a scenario prediction model based on the feature vector to predict future usage scenarios; wherein the NFC interaction data includes an interaction timestamp, an interaction location, a device type, a loaded content type, and a loaded result; The scene prediction model M scene A multi-task temporal learning model is used for modeling, with the input being a feature vector and a scene label set {S t }, the scene label set includes payment scene and access control scene, the scene prediction model M scene Contains the following sections: Backbone model: Based on the LSTM structure, learn the temporal pattern of features and output the intermediate state h t ; Classifier: A linear layer predicts the probability distribution p of the next scene t+1 : p t+1 =softmax(W h ·h t +b h ) Among them, W h and b h are the weights and biases of the classifier; According to the NFC scenario, add the following scenario distribution balance regularization items Where k is the number of scene categories; P(S i ) is the predicted probability of category i; α is the balance coefficient, which is used to prevent the model from overfitting to high-frequency scenarios; An NFC scene prediction unit, used to generate a priority list of loaded data according to the scene prediction results; An NFC data loading unit, used for adjusting a priority list of loaded data according to real-time signal strength to obtain an adjusted loading set; The NFC data decoding unit is used to construct a dynamic decoding tree structure that relies on optimization, and use the dynamic decoding tree to parse the loaded data from the adjusted loading set to ensure that the data can be used by the user in a timely manner; wherein the dynamic decoding tree structure The build is as follows: First, for each data content c loaded i Assign weight w ij , defined as: Among them, Sim(c i ,c j ) is data c i With c j Similarity; Size(c j ) is data c j The load size is used to limit the negative impact of large data on decoding priority; According to the weight w ij Generate decoding paths from high to low, giving priority to decoding content that is highly correlated with other data and has a low loading cost; The NFC decoding system optimization unit is used to optimize the scenario prediction model and loading strategy based on the feedback data of the loading process to form a closed-loop optimization.

Images