An online incremental learning method and system based on adaptive B+ tree index

Abstract

The application discloses an online incremental learning method and system based on adaptive B+ tree index, relates to the technical field of artificial intelligence, and comprises the following steps: constructing an extended adaptive B+ tree index structure, inserting a new sample and calculating a track disturbance value; combining the disturbance value and a node state to generate a disturbance score, and adding the new sample and associated node samples to an incremental learning scheduling queue when a condition is met; obtaining an affected node and updating the structure code of the affected node when a node splitting, merging or reconstructing event occurs; monitoring sample label track changes, adjusting a label confidence marker when a threshold is exceeded, and limiting the participation frequency according to a screening mechanism; and performing a forgetting compression operation on a leaf node with low access frequency and small label fluctuation degree. Through the construction of the adaptive B+ tree index, the disturbance-driven scheduling, the structure change response and the forgetting compression strategy, the sample management is refined, the learning update is efficient, and the structure maintenance is controllable, so that the stability and resource utilization efficiency of the online incremental learning are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, in particular to an online incremental learning method and system based on adaptive B+ tree index. BACKGROUND

[0002] With the wide application of artificial intelligence in real-time data processing, intelligent decision-making and edge computing, online incremental learning has become an important research direction to adapt to dynamic environmental changes and continuously improve model performance. Online incremental learning systems usually need to have the following capabilities: real-time acceptance of new samples, dynamic updating of models, management of sample distribution changes, and evaluation and control of sample quality and structural stability.

[0003] In the prior art, most online learning methods are based on global model optimization and cannot respond to micro-local perturbations and structural changes in sample distribution in a timely manner, resulting in model update lag, sample redundancy accumulation or low-quality sample interference in the training process. In addition, the traditional index structure has the following defects in sample insertion and dynamic structure maintenance: lack of explicit modeling mechanism for sample trajectory changes, unable to depict the deviation degree of new sample insertion path and historical path; not supporting real-time evaluation and dynamic intervention of node local state; lack of controllable cleaning strategy for redundant samples in nodes, and storage efficiency and model performance decline in long-term operation. SUMMARY

[0004] Based on the above-mentioned defects of the prior art, the purpose of the present application is to provide an online incremental learning method and system based on adaptive B+ tree index to solve the above technical problems.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical scheme: an online incremental learning method based on adaptive B+ tree index, comprising:

[0006] S1: Construct an extended adaptive B+ tree index structure, wherein the sample fields stored in the leaf nodes include feature vectors, timestamps, labels, label confidence markers and structure encodings, and the state parameters of the leaf nodes include access frequencies, label fluctuation degrees and structure evolution rates;

[0007] S2: Obtain the feature vector of the new sample, insert it into the B+ tree index structure, and calculate the trajectory disturbance value according to the sequence difference between the insertion path of the new sample and the historical sample insertion path;

[0008] S3: When the new sample is inserted into the B+ tree index structure, generate a disturbance score according to the trajectory disturbance value and the state parameters of the leaf node, and when the disturbance score is greater than a first threshold and the label confidence marker is credible or unstable, add the new sample and the sample set of the leaf node to the incremental learning scheduling queue;

[0009] S4: When a structural change event is triggered by node splitting, merging or restructuring operation, obtain the affected node subset, update the structure code of the nodes in the subset;

[0010] S5: Obtain the label trajectory of the sample, update the label confidence marker of the sample whose label trajectory changes greater than the second threshold, update the label confidence marker to unstable or suspicious, limit the participation frequency of the sample according to the preset semi-supervised screening mechanism;

[0011] S6: Perform a forgetting compression operation on the leaf node whose access frequency is less than the third threshold and the label fluctuation degree is less than the fourth threshold, and delete the sample in the node.

[0012] The application further provides that, in the sample field, the feature vector is used to record the multi-dimensional representation of the inserted sample after being mapped by a feature extraction function; the timestamp is used to record the time when the sample is inserted into the index structure; the label is used to record the supervised category information of the sample; the label confidence marker is used to record the credibility state of the label source, including reusable, reliable, unstable and suspicious, and is initially set as reliable; and the structure code is used to represent the local structure distribution state of the leaf node where the sample is located, and is generated based on the feature vectors of all samples in the node through an aggregation mapping function.

[0013] In the state parameters, the access frequency is updated by accumulation every time the node is inserted or queried; the label fluctuation degree is obtained by calculating the category distribution entropy or variation amplitude of the labels of the samples in the node within a set time window; and the structure evolution rate is obtained by performing difference calculation on the structure code sequence generated by the node within a continuous time period.

[0014] The application further provides that, step S2 comprises:

[0015] The feature vector of the new sample is obtained, the feature vector is inserted into the adaptive B+ tree index structure according to a preset index rule, the node sequence recorded in the insertion process is generated, and an insertion path is generated;

[0016] A preset number of historical sample insertion paths are called from the index structure, the historical sample insertion path is composed of the node access sequence recorded in the historical sample insertion process;

[0017] The insertion path and the historical sample insertion path are compared in sequence difference, the trajectory disturbance value is calculated based on the distance between the edit distance, the jump depth and the structure embedding vector of the node sequence, and when there is no historical sample in the index structure, the trajectory disturbance value is set to the maximum value.

[0018] The application further provides that, step S3 comprises:

[0019] When a new sample is inserted into the B+ tree index structure, the trajectory disturbance value, the label fluctuation degree of the leaf node where the new sample is located, and the structure evolution rate are obtained;

[0020] The trajectory disturbance value, the label fluctuation degree, and the structure evolution rate are jointly analyzed to generate a disturbance score;

[0021] When the disturbance score is greater than a preset first threshold value, and the label confidence mark of the sample in the leaf node is credible or unstable, the new sample and the sample set in the leaf node are added to an incremental learning scheduling queue.

[0022] The application further provides that the trajectory disturbance value, the label fluctuation degree, and the structure evolution rate are jointly analyzed to generate a disturbance score, including:

[0023] When the trajectory disturbance value is greater than a first disturbance threshold value, and the label fluctuation degree is greater than a second disturbance threshold value or the structure evolution rate is greater than a third disturbance threshold value, it is determined that the disturbance score is high.

[0024] When the trajectory disturbance value is greater than the first disturbance threshold value, and the label fluctuation degree is less than or equal to the second disturbance threshold value, and the structure evolution rate is less than or equal to the third disturbance threshold value, it is determined that the disturbance score is medium.

[0025] When the trajectory disturbance value is less than or equal to the first disturbance threshold value, it is determined that the disturbance score is low.

[0026] The application further provides that in step S4, when the number of samples in the target node is greater than a preset upper threshold value after the sample is inserted, a node splitting event is triggered; when the node splitting event occurs, the original node before splitting, the new node generated by splitting, and the parent node are set as a subset of affected nodes.

[0027] When the number of samples in the target node is less than a preset lower threshold value after the sample is deleted, and the target node can be merged with an adjacent sibling node, a node merging event is triggered; when the node merging event occurs, the nodes participating in the merging and the parent node are set as a subset of affected nodes.

[0028] When splitting or merging events continuously occur in the subtree area until the height difference between the left and right subtrees is greater than a preset height difference threshold value, a node reconstruction event is triggered; when the node reconstruction event occurs, all nodes participating in the reconstruction are set as a subset of affected nodes.

[0029] For each node in the subset of affected nodes, the hierarchical position, the index serial number, the type mark of whether it is a leaf node, and the current number of child nodes of the node are collected, a structure encoding value is generated through a structure encoding function, and the structure encoding value is written into the structure encoding field of the corresponding node.

[0030] The application further provides that step S5 includes:

[0031] establishing a label trajectory record table with the feature vector as the primary key in the leaf node, when the label of the sample changes, obtaining the current timestamp and label value, and generating a label trajectory sequence;

[0032] According to the label trajectory sequence, the number of occurrences of a pair of continuous different labels is counted, and is normalized to the ratio of the total trajectory length, which is defined as the label change rate;

[0033] According to the label trajectory sequence, the total number of label categories is counted, and the label value range is calculated in the regression task;

[0034] When the label change rate and / or the label value range is greater than the second threshold, the label confidence marker of the sample is updated to an unstable state, and when it is greater than the second threshold for a continuous preset time window, it is updated to a suspicious state;

[0035] Based on the updated label confidence marker, when performing incremental learning scheduling, a preset semi-supervised screening strategy is applied, wherein the unstable state sample is set to have an upper limit of sampling probability, and the suspicious state sample is rejected to enter the scheduling task queue.

[0036] The application further provides that step S6 comprises:

[0037] A forgetting compression strategy triggered at a fixed period is set, the leaf nodes are traversed in the period, the access frequency and the label fluctuation degree of the nodes in a preset sliding time window are calculated, wherein the access frequency is the number of node accesses divided by the window length; and the label fluctuation degree is the average of the label fluctuation degree of the nodes;

[0038] When the access frequency of the leaf node is less than a third threshold and the label fluctuation degree is less than a fourth threshold, a forgetting compression operation is performed, all sample fields stored in the leaf node are deleted, the node state is reset to reusable, and the structure code is retained.

[0039] The application further provides that it further comprises:

[0040] A periodic explainability output mechanism is set, in a preset sample processing period, the running state information of the adaptive B+ tree index structure is collected, and an explainability data file including a node access heat map and a structure change map is generated;

[0041] The node access heat map calculates the standardized heat value based on the access frequency of the leaf node in the sliding window, and generates a structured index map with heat annotation;

[0042] The structure change map records the event type, occurrence time and node evolution relationship by monitoring the node splitting, merging or reconstruction events in the index structure, and generates a time series structure graph.

[0043] The application further provides an online incremental learning system based on adaptive B+ tree index, which is used for realizing the online incremental learning method based on adaptive B+ tree index.

[0044] A structure construction module is configured to construct an extended adaptive B+ tree index structure, wherein a sample field stored in a leaf node comprises a feature vector, a timestamp, a label, a label confidence label and a structure code, and a state parameter of the leaf node comprises a visit frequency, a label fluctuation degree and a structure evolution rate.

[0045] A sample insertion module is configured to obtain a feature vector of a new sample, insert the new sample into the B+ tree index structure, and calculate a trajectory disturbance value according to a sequence difference between an insertion path of the new sample and a historical sample insertion path.

[0046] A score generation module is configured to generate a disturbance score according to the trajectory disturbance value and the state parameter of the leaf node when the new sample is inserted into the B+ tree index structure, and add the new sample and a sample set of the leaf node into an incremental learning scheduling queue when the disturbance score is greater than a first threshold value and the label confidence label is credible or unstable.

[0047] An event update module is configured to obtain a subset of affected nodes when a structure variation event is triggered by a node splitting, merging or reconstructing operation, and update structure codes of the nodes in the subset.

[0048] A label update module is configured to obtain a label trajectory of a sample, update a label confidence label of the sample when a label trajectory change of the sample is greater than a second threshold value, update the label confidence label to be unstable or suspicious, and limit a participation frequency of the sample according to a preset semi-supervised screening mechanism.

[0049] A forgetting compression module is configured to perform a forgetting compression operation on a leaf node with a visit frequency less than a third threshold value and a label fluctuation degree less than a fourth threshold value, and delete samples in the node.

[0050] The application provides an online incremental learning method and system based on adaptive B+ tree index, the method constructs an extended adaptive B+ tree index structure, wherein the sample field stored in the leaf node includes a feature vector, a timestamp, a label, a label confidence mark and a structure code, and the state parameters of the leaf node include the access frequency, the label fluctuation degree and the structure evolution rate; the feature vector of the new sample is obtained, the B+ tree index structure is inserted, and the trajectory disturbance value is calculated according to the sequence difference between the insertion path of the new sample and the insertion path of the historical sample; when the new sample is inserted into the B+ tree index structure, the disturbance score is generated according to the trajectory disturbance value and the state parameters of the leaf node, when the disturbance score is greater than the first threshold value and the label confidence mark is credible or unstable, the new sample and the sample set of the leaf node are added to the incremental learning scheduling queue; when the node splitting, merging or reconstruction operation triggers the structure change event, the affected node subset is obtained, and the structure code of the node in the subset is updated; the label trajectory of the sample is obtained, the label confidence mark of the sample with a label trajectory change greater than the second threshold value is updated, the label confidence mark is updated to unstable or suspicious, and the participation frequency of the sample is limited according to the preset semi-supervised screening mechanism; the forgetful compression operation is performed on the leaf node with an access frequency less than the third threshold value and a label fluctuation degree less than the fourth threshold value, the sample in the node is deleted, and the beneficial effects include:

[0051] 1. Realize sample-level structured management and state awareness: by defining the sample field and state parameters of the leaf node in the extended adaptive B+ tree index structure, the dynamic perception and continuous management of the structure position, evolution trajectory and label credibility of each sample are realized, and the adaptability to the change of streaming data is enhanced;

[0052] 2. Introduce a disturbance-driven incremental learning scheduling mechanism: based on the sequence difference between the insertion path of the new sample and the insertion path of the historical sample, the trajectory disturbance value is calculated, and the disturbance score is generated in combination with the state parameters of the leaf node, so that the model is only updated incrementally at the node with representative changes in structure disturbance or label state, and the risk of redundant training and overfitting is effectively inhibited;

[0053] 3. Support adaptive response and code update of structure change event: for node splitting, merging and reconstruction events, the structure change detection mechanism is triggered, the affected node subset is dynamically identified, and the structure code update is performed, so as to maintain the logical consistency and interpretability of the index structure, and provide support for subsequent visualization and structure analysis;

[0054] 4. Provide a controllable forgetful compression strategy: set the joint judgment mechanism of the access frequency and the label fluctuation degree in the sliding time window, perform the compression and deletion operation on the leaf node with long-time low frequency and stable fluctuation, avoid the expansion of the index structure and the accumulation of samples, and improve the resource utilization rate.

[0055] The above description is only a summary of the technical solutions of the present application. In order to more clearly understand the technical means of the present application, the following specific embodiments of the present application can be implemented according to the content of the description, and in order to make the above and other purposes, characteristics and advantages of the present application more obvious and easy to understand, the following specific embodiments of the present application are described. BRIEF DESCRIPTION OF DRAWINGS

[0056] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without any creative labor. In the drawings:

[0057] Figure 1 A flow chart of an online incremental learning method based on adaptive B+ tree index is shown for an exemplary embodiment of the present application;

[0058] Figure 2 A structural schematic diagram of an online incremental learning system based on adaptive B+ tree index is shown for an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0059] The embodiments of the present application will be described below with reference to the drawings and preferred embodiments, and those skilled in the art can easily understand other advantages and effects of the present application from the content disclosed in the specification. The present application can also be implemented or applied by different specific embodiments, and the details in the specification can be modified or changed based on different views and applications without departing from the spirit of the present application. It should be understood that the preferred embodiments are only for illustrating the present application, but not for limiting the protection scope of the present application.

[0060] It should be noted that the diagrams provided in the following embodiments only illustrate the basic concept of the present application in a schematic manner, and only the components related to the present application are shown in the diagrams, but not the number, shape and size of the components when actually implemented. The actual implementation of each component may be randomly changed in type, number and proportion, and the layout pattern of the components may also be more complex.

[0061] In the following description, a large number of details are discussed to provide a more thorough explanation of the embodiments of the present application, however, it is obvious to those skilled in the art that the embodiments of the present application can be implemented without these specific details, and in other embodiments, the known structures and devices are shown in the form of block diagrams rather than in the form of details, so as not to make the embodiments of the present application difficult to understand.

[0062] Embodiment one:

[0063] An online incremental learning method based on adaptive B+ tree index, as shown in Figure 1 includes:

[0064] S1: Construct an extended adaptive B+ tree index structure, wherein the sample fields stored in the leaf nodes include feature vectors, timestamps, labels, label confidence markers and structure encodings, and the state parameters of the leaf nodes include access frequencies, label fluctuation degrees and structure evolution rates;

[0065] S2: Obtain the feature vector of the new sample, insert it into the B+ tree index structure, and calculate the trajectory disturbance value according to the sequence difference between the insertion path of the new sample and the insertion path of the historical sample;

[0066] S3: When the new sample is inserted into the B+ tree index structure, generate a disturbance score according to the trajectory disturbance value and the state parameters of the leaf nodes, and when the disturbance score is greater than a first threshold value and the label confidence marker is credible or unstable, add the new sample and the sample set of the leaf node to the incremental learning scheduling queue;

[0067] S4: When the node splitting, merging or reconstruction operation triggers a structure change event, obtain the affected node subset, and update the structure encoding of the nodes in the subset;

[0068] S5: Obtain the label trajectory of the sample, update the label confidence marker of the sample whose label trajectory changes by more than a second threshold value, and update the label confidence marker to unstable or suspicious, and limit the participation frequency of the sample according to a pre-set semi-supervised screening mechanism;

[0069] S6: Perform a forgetting compression operation on the leaf node whose access frequency is less than a third threshold value and whose label fluctuation degree is less than a fourth threshold value, and delete the samples in the node.

[0070] B+ tree is a self-balancing tree structure widely used in databases and file systems, mainly used for storing and managing large amounts of data, its characteristics are that all records are stored in leaf nodes, and leaf nodes are linked by linked lists, which is convenient for range query and sequential traversal, it is a multi-way search tree with efficient insertion, deletion and lookup operations, the extended B+ tree is based on the traditional B+ tree, and supports adaptive management of samples, by introducing new fields and state parameters, the B+ tree can not only perform efficient index lookup, but also adapt to dynamic changes of data and self-adjustment. For example, it can dynamically optimize the structure, manage samples according to the access frequency, label fluctuation and node structure change, in step S1, the application is further provided, in the sample field, the feature vector is used to record the multi-dimensional representation of the inserted sample after mapping by the feature extraction function; the timestamp is used to record the time when the sample is inserted into the index structure; the label is used to record the supervised class information of the sample; the label confidence mark is used to record the confidence state of the label source, including reusable, reliable, unstable and suspicious, and is initially set to reliable; the structure code is used to represent the local structure distribution state of the leaf node where the sample is located, which is generated based on the feature vectors of all samples in the node through an aggregation mapping function; specifically, the feature vector is a vector representation in a multi-dimensional space obtained by mapping the original data of the sample through a feature extraction function (such as PCA, CNN, traditional statistical method in machine learning), which is a high-dimensional representation of the sample and is used as input for classification, regression and other tasks, wherein the original sample data includes image, text or sensor data; the original data is mapped into a vector in a multi-dimensional feature space through a series of feature extraction methods (such as convolutional neural network (CNN) for image, TF-IDF for text), which is not described here, the obtained vector is stored in the sample field of the leaf node, so as to facilitate subsequent calculation and index lookup; the timestamp represents the time when the sample is inserted into the B+ tree index structure, which is used to record the historical state of the sample and provide a basis for the time series analysis of the sample, whenever a new sample is inserted into the B+ tree index, the current time (such as Unix timestamp or UTC time) is obtained and stored in the timestamp field of the sample, the timestamp is used for time series analysis, which can track the historical state of the sample and support time-dependent update in incremental learning; the label is the supervised class information of the sample (class label in classification task, numerical value in regression task), which is the target variable of the sample and the basis for model learning, the label is provided by the labeler and stored in the leaf node, so as to be indexed and managed together with the sample feature vector and learned; the label confidence mark is used to represent the confidence state of the label. There are four states of label confidence: reusable, reliable, unstable and suspicious.Initially, the label confidence mark is set to be trusted, which means that the sample label comes from a trusted source and can be used multiple times; trusted means that the label is relatively accurate, and the model or labeling process has been verified; unstable means that the label has uncertainty or large fluctuation, and may need to be recalibrated due to data fluctuation; suspicious means that the label is highly uncertain and cannot be trusted, and needs to be further checked; the label confidence mark is dynamically updated according to the label trajectory, classification result and label fluctuation of the sample. In incremental learning, if the sample label fluctuates frequently or is unstable, the confidence mark of the updated label is unstable or suspicious; the structure code is used to represent the local structure state of the leaf node, which is calculated based on the feature vectors of all samples in the node, and is a summary value of the node, reflecting the overall information of the data distribution in the node; the structure code is obtained by aggregating the feature vectors of all samples in the node (such as taking the average, weighted average, etc.); the structure code can be generated by an aggregation mapping function, such as PCA dimensionality reduction or vector clustering of the feature vector, and then calculating the aggregation result (such as mean or center point) of the feature of all samples in the node, which is a prior art, and will not be described here; the structure code is stored in the leaf node as one of the index values of the node.

[0071] In the state parameters, the access times are updated by accumulation every time the node is inserted or queried; the label fluctuation degree is obtained by calculating the class distribution entropy or variation amplitude of the sample labels in the node within a set time window; the structure evolution rate is obtained by performing difference calculation on the structure code sequence generated by the node in the continuous time period; specifically, the access times represent the number of times the node is accessed, which is used to measure the activity of the node, and the access times of the node are updated by one every time a sample is inserted into the node; each leaf node will maintain an "access times" field, which is accumulated every time the node is accessed, facilitating subsequent analysis of the activity of the node; the label fluctuation degree reflects the change amplitude of the sample labels in the node, which is measured by calculating the entropy or variation amplitude of the label class distribution in the node; the frequency distribution of all sample labels in the node is calculated, and the entropy of the label distribution is calculated based on this: H = -∑ i p i log(p i ), where p i is the probability of each label category; the structure evolution rate measures the speed of the change of the node structure, which represents the structure change frequency of the node in a period of time, and the difference calculation is performed on the structure code generated by the node in the continuous time period to measure the amplitude or frequency of the change. If the structure code changes frequently in a short time, it indicates that the node is in a high evolution rate state, and the structure evolution rate S r The difference between the current code and the code of the previous time period is calculated as follows: Where C current and C previousStructural encoding of the current and previous time period, respectively, and Δt is the time difference.

[0072] The application is further configured to, the step S2 comprises:

[0073] The feature vector of the new sample is obtained, and the feature vector is inserted into the adaptive B+ tree index structure according to a preset index rule, and a node sequence passed in the insertion process is recorded to generate an insertion path. Specifically, the feature vector of the new sample is input into the constructed adaptive B+ tree index structure. In the index process, the node number or unique structure identifier of each intermediate node or leaf node is written into the 'insertion path sequence' every time the node is passed, so as to form the complete insertion track of the sample. Further, the preset index rule is a minimum distance matching rule based on the distance of the cluster center: each sub-node maintains the center point of the sample feature vector in it, and the intermediate node saves the center point set of all sub-nodes. For the feature vector of the new sample, the Euclidean distances between the feature vector and all sub-node center points are calculated in turn, and the path selection follows the minimum distance matching principle: the sub-node with the smallest distance is selected as the next hop, and the sample is inserted into the corresponding leaf node, and if the number of samples exceeds the upper limit, the leaf node is split.

[0074] A preset number of historical sample insertion paths are recalled from the index structure, and the historical sample insertion path is composed of the node access sequence recorded in the historical sample insertion process. Specifically, three insertion paths of historical samples are recalled from the same leaf node in the B+ tree as a reference track set. Each historical path is the node access sequence recorded by the historical sample during insertion, and the historical path needs to meet certain time window or sample selection strategies (such as recently inserted or specific label samples);

[0075] The insertion path and the historical sample insertion path are compared in sequence difference, the trajectory disturbance value is calculated based on the distance between the editing distance of the node sequence, the jump depth and the structure embedding vector, and when there is no historical sample in the index structure, the trajectory disturbance value is set to the maximum value. Specifically, the insertion path of the new sample and each historical sample path are compared in sequence difference: the number of modification steps of the two path sequences at the node level is evaluated, which is recorded as the editing distance; the number of non-continuous jump levels appearing in the path is recorded, which is recorded as the jump depth; the structure embedding vector (structure encoding) of each node is extracted, the average distance between the new sample path and the historical path in the embedding space is measured, which is recorded as the structure embedding vector difference, and the trajectory disturbance value is formed by comprehensively considering the above three indexes as the quantitative result of the structure disturbance degree; further, the trajectory disturbance value = 0.5*editing distance + 0.2*jump depth + 0.3*embedding distance average, wherein the editing distance, the jump depth and the embedding distance average are all normalized values, and when there is no historical sample in the index structure, the trajectory disturbance value is set to 1.

[0076] The application is further configured that the step S3 comprises:

[0077] When the new sample is inserted into the B+ tree index structure, the track disturbance value, the label fluctuation degree of the leaf node where the new sample is located, and the structure evolution rate are obtained; specifically, when the new sample is inserted into the B+ tree index structure, the obtaining logic of the track disturbance value, the label fluctuation degree, and the structure evolution rate has been described in the foregoing, and will not be described here.

[0078] The track disturbance value, the label fluctuation degree, and the structure evolution rate are jointly analyzed to generate a disturbance score; the application is further configured that the track disturbance value, the label fluctuation degree, and the structure evolution rate are jointly analyzed to generate a disturbance score, including: when the track disturbance value is greater than a first disturbance threshold, and the label fluctuation degree is greater than a second disturbance threshold or the structure evolution rate is greater than a third disturbance threshold, it is determined that the disturbance score is high; when the track disturbance value is greater than the first disturbance threshold, and the label fluctuation degree is less than or equal to the second disturbance threshold, and the structure evolution rate is less than or equal to the third disturbance threshold, it is determined that the disturbance score is medium; when the track disturbance value is less than or equal to the first disturbance threshold, it is determined that the disturbance score is low; specifically, after the three types of disturbance-related parameters are extracted, the joint analysis process of the disturbance score is entered. Three threshold parameters are set: the first disturbance threshold is used to determine whether the track disturbance value is significant, the second disturbance threshold is used to determine whether the label fluctuation degree is severe, and the third disturbance threshold is used to determine whether the structure evolution rate is fast; the thresholds are set through historical statistical data; if the track disturbance value is greater than the first disturbance threshold, and either the label fluctuation degree or the structure evolution rate is greater than the corresponding threshold, the disturbance score is “high”; if the track disturbance value is greater than the first disturbance threshold, and both the label fluctuation degree and the structure evolution rate do not exceed the corresponding threshold, the disturbance score is “medium”; if the track disturbance value does not exceed the first disturbance threshold, the disturbance score is “low”. The score result adopts a level identifier (“high, medium, and low”), compared with a continuous numerical form, the level identifier can execute scheduling decisions through simple conditional branching when judging, without the need for fine numerical comparison, thereby reducing model logic complexity, improving system running efficiency and robustness;

[0079] When the disturbance score is greater than a preset first threshold value, and the label confidence mark of the sample in the leaf node is trusted or unstable, the new sample and the sample set in the leaf node are added to the incremental learning scheduling queue; the first threshold value (medium) of the disturbance score triggering scheduling is set. When the disturbance score triggered by the insertion of the current sample exceeds the threshold value, the label confidence mark state of the sample set in the leaf node to which the sample belongs is further checked. If the label confidence mark of the sample set in the leaf node is "trusted" or "unstable", it indicates that the samples in this region are not all artificially labeled, but have a certain reference value. The current sample and all samples in the node are packaged as a learning scheduling unit and pushed to the incremental learning scheduling queue as candidate data for model updating; if the disturbance score does not exceed the threshold value, or the confidence mark is in the "suspicious" state, the scheduling is not triggered to avoid redundant training or the introduction of low-quality samples.

[0080] The application further provides that, in step S4, when the number of samples in the target node is greater than a preset upper threshold value after the insertion of the sample, a node splitting event is triggered; when the node splitting event occurs, the original node before splitting, the new node generated by splitting and the parent node are set as the affected node subset; specifically, after a target leaf node inserts a new sample, the number of samples N_node in the target leaf node exceeds the preset upper threshold value T_upper, and the node splitting event is triggered; sample division is performed on the node, and the samples are divided into two groups in order according to the key; a new leaf node is created, and one group of samples is migrated to the new node; the new node is registered in the parent node; the node before splitting, the new node after splitting and the parent node are added to the "affected node subset";

[0081] When the number of samples in the target node is less than a preset lower threshold value after the deletion of the sample, and the target node can be merged with an adjacent sibling node, a node merging event is triggered; when the node merging event occurs, the nodes participating in the merging and the parent node are set as the affected node subset; specifically, after a target node deletes a sample, the number of samples N_node in the target node is lower than the preset lower threshold value T_lower, and there is a mergeable sibling node, and the node merging event is triggered; the samples of the node and the adjacent sibling node are merged into one node; the indexes of the original two nodes in the parent node are deleted; the merged node and the parent node are added to the "affected node subset";

[0082] When a split or merge event occurs continuously in the subtree area, and the height difference between the left and right subtrees is greater than the preset height difference threshold, a node reconstruction event is triggered; when a node reconstruction event occurs, all nodes participating in the reconstruction are set as the affected node subset; specifically, a node merge / split occurs continuously in a subtree, resulting in a height difference ΔH between the left and right subtrees exceeding a preset height difference threshold T_height; a node reconstruction event is triggered; structure rebalancing (such as AVL-like rotation or B-tree rearrangement) is performed on the entire subtree structure; all structure adjustment nodes involved are recorded; all affected nodes are added to the "affected node subset";

[0083] For each node in the affected node subset, the hierarchical position, index number, type label of whether it is a leaf node, and current number of child nodes of the node are collected, a structure encoding value is generated through a structure encoding function, and the structure encoding value is written into the structure encoding field of the corresponding node; specifically, for each node in the affected node subset, the following operations are performed: collect node structure state parameters, including hierarchical position: record the depth L of the node in the B+ tree structure; index number: record the order position Idx of the node in its parent node; node type label: Boolean variable isLeaf labels whether it is a leaf node; current number of child nodes: record the number of direct child nodes C it has (if it is a leaf node, it can be set to 0); input the above structure state parameter set S={L, Idx, isLeaf, C}; generate a structure encoding value E_struct through a structure encoding function F_enc(S); further, the structure encoding function F_enc(S) is an ordered numerical vector encoding, which directly maps the structure state to a standardized structure vector.

[0084] The application further provides that step S5 comprises:

[0085] In the leaf node, a label trajectory record table is established with the feature vector as the primary key. When the label of the sample changes, the current timestamp and label value are obtained, and a label trajectory sequence is generated; specifically, the label change experienced by each sample in its life cycle is continuously tracked to form a label trajectory; a label trajectory record table is established in each leaf node with the sample feature vector v as the primary key; every time the sample is accessed again, if the label value is different from the last record, the current timestamp t i , the current label value y i ; form a sequence form: T v ={(t1,y1),(t2,y2),…,(t n ,y n )}, if the sample feature vector appears for the first time or there is no record in the table, initialize the label trajectory;

[0086] The ratio of the number of occurrences of consecutive different label pairs in the label trajectory sequence to the total trajectory length is defined as the label variation rate; specifically, from sequence T... v Extract all label change events, i.e., y i ≠y i-1 Let the total number of change events be Δy, and define the label change rate as R. change =Δy / (n-1), the label change rate reflects the degree of instability of the label state;

[0087] Based on the label trajectory sequence, the total number of label categories is counted, and the label value range is calculated in the regression task. When the label change rate and / or the label value range are greater than a second threshold, the label confidence marker of the sample is updated to an unstable state. If it remains greater than the second threshold for a continuous preset time window, it is updated to a suspicious state. Specifically, the number C of different categories in the label trajectory is counted. count Calculate the range R of the labels in the trajectory. range =max(y i )-min(y i The second threshold θ2 is a set of thresholds for judging label confidence updates, with corresponding thresholds set for both label change rate and label value range. t Let R be the system's rolling time window, and k be the continuous window threshold counter. range >θ2 and C count When R > θ2, or when R range >θ2 or C count When the value is greater than θ2, the confidence label of the current sample is set to "unstable". If the above conditions are met within k consecutive time windows, the confidence label is updated to "suspicious", and the label confidence label written to the sample field is updated.

[0088] Based on the updated label confidence markers, a pre-defined semi-supervised screening strategy is applied during incremental learning scheduling. Specifically, an upper limit on the sampling probability is set for unstable state samples, and suspicious state samples are rejected from entering the scheduling task queue. In particular, the impact of unreliable samples on model training is controlled to reduce the risk of label error propagation, and a pre-defined upper limit on the sampling probability P1(0.2)∈(0,1) is set. When the label confidence marker is reusable, it directly enters the scheduling queue. When the label confidence marker is reliable, it directly enters the scheduling queue. When the label confidence marker is unstable, random sampling is performed with probability P1 to determine whether to schedule it. When the label confidence marker is suspicious, it is rejected from entering the scheduling queue.

[0089] The present invention is further configured such that step S6 includes:

[0090] A fixed period triggered forgetting compression strategy is set, leaf nodes are traversed in the period, and the access frequency and label fluctuation degree of the nodes in a preset sliding time window are calculated, wherein the access frequency is the number of node access times divided by the window length; the label fluctuation degree is the mean value of the label fluctuation degree of the node; specifically, the system presets a compression period parameter T_gc; a scheduler is set, and the traversal compression process is triggered once every T_gc period; during execution, all leaf node sets are sequentially evaluated;

[0091] When the access frequency of the leaf node is less than the third threshold value and the label fluctuation degree is less than the fourth threshold value, a forgetting compression operation is performed, all sample fields stored in the leaf node are deleted, the node state is reset to reusable, and the structure code is retained; specifically, the state field maintained by the node: cumulative access times V_total; current system clock t_now; last compression check time t_last_gc; sliding window length W, if t_now-t_last_gc≥T_gc, then the access frequency in the window is defined as: f access =W window / V, W window represents the number of accesses to the node within W time; when the access frequency of the leaf node is less than the third threshold value and the label fluctuation degree is less than the fourth threshold value, the node is marked as a "low activity + stable" node, and a forgetting compression is performed, including: deleting all sample fields (including feature vectors, timestamps, labels, and label confidence markers) in the node; retaining the structure code field (for structure visualization and subsequent reconstruction); setting the state flag as "reusable node" for subsequent sample insertion to preferentially fill; resetting the access count to 0 and emptying the label fluctuation buffer.

[0092] The application further provides that it also includes:

[0093] A periodic interpretability output mechanism is set, running state information of the adaptive B+ tree index structure is collected within a preset sample processing period, and an interpretability data file including a node access heat map and a structure change map is generated;

[0094] The node access heat map calculates a standardized heat value based on the access frequency of the leaf node in the sliding window, and generates a structured index map with heat labels; specifically, for all leaf nodes in the adaptive B+ tree, the access records of the leaf nodes in the sliding time window W are collected; each leaf node stores an access timestamp buffer or an access counter; the number of accesses V i in the sliding window is counted, the access frequency f i =V i / W is calculated, and all access frequencies f i are standardized to heat values h i ∈[0,1], using Min-Max normalization: h i =(fi -f min ) / (f max -f min +∈), where, f max f min Let h be the minimum and maximum access frequencies of all leaf nodes within the sliding window, respectively, where ∈ is a small constant to avoid division by zero; traverse the B+ tree structure and construct a graph structure (node-edge topology) based on the current index topology; assign the heat value h of each leaf node to h. i Map to color / brightness levels; label onto the corresponding node elements to generate a heatmap;

[0095] The structural change graph generates a time series structural graph by monitoring node splitting, merging, or reconstructing events in the index structure, recording event types, occurrence times, and node evolution relationships. Specifically, it aggregates structural change records and constructs a time series graph in chronological order. The graph structure expresses the evolution path of nodes and can generate visual charts (PDF / PNG) or structural graphs (GraphML / JSON), preserving timestamps, structural change chains, and the relationship graphs of the involved nodes.

[0096] Example 2:

[0097] Please see Figure 2 This exemplary online incremental learning system based on an adaptive B+ tree index is used to implement the aforementioned online incremental learning method based on an adaptive B+ tree index, and includes:

[0098] Structure building module: Constructs an extended adaptive B+ tree index structure, in which the sample fields stored in the leaf nodes include feature vector, timestamp, label, label confidence marker and structure encoding, and the state parameters of the leaf nodes include access count, label fluctuation degree and structure evolution rate;

[0099] Sample insertion module: Obtain the feature vector of the new sample, insert it into the B+ tree index structure, and calculate the trajectory perturbation value based on the sequence difference between the insertion path of the new sample and the insertion path of the historical samples.

[0100] The scoring generation module generates a perturbation score based on the trajectory perturbation value and the state parameters of the leaf nodes when a new sample is inserted into the B+ tree index structure. When the score is greater than the first threshold and the label confidence is marked as reliable or unstable, the new sample and the sample set of the leaf nodes are added to the incremental learning scheduling queue.

[0101] Event Update Module: When a node splitting, merging, or reconstructing operation triggers a structural change event, it obtains a subset of affected nodes and updates the structural encoding of the nodes in the subset.

[0102] The label updating module: obtains a label track of the sample, performs label confidence label updating on the sample whose label track changes greater than a second threshold, updates the label confidence label to unstable or suspicious, and limits the participation frequency of the sample according to a preset semi-supervised screening mechanism.

[0103] The forgetting compression module: performs a forgetting compression operation on the leaf node whose access frequency is less than a third threshold and whose label fluctuation degree is less than a fourth threshold, and deletes the sample in the node.

[0104] It should be noted that the online incremental learning system based on adaptive B+ tree index provided in the above embodiment and the online incremental learning method based on adaptive B+ tree index provided in the above embodiment belong to the same concept, wherein the specific operation manner of each module and unit has been described in detail in the method embodiment, which will not be repeated here. The online incremental learning system based on adaptive B+ tree index provided in the above embodiment can be realized by different functional modules to complete the above-described functions according to the needs in actual application, that is, the internal structure of the system is divided into different functional modules to complete all or part of the above-described functions, which is not limited here.

[0105] The above embodiments can be realized all or partially by software, hardware, firmware or other any combination. When realized by software, the above embodiments can be realized all or partially in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network or other programmable devices. The computer instructions can be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another, for example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (such as infrared, wireless, microwave, etc.) mode. The computer readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center and the like containing one or more available medium sets. The available medium can be a magnetic medium (such as a floppy disk, a hard disk, a magnetic tape), an optical medium (such as a DVD) or a semiconductor medium. The semiconductor medium can be a solid state disk.

[0106] It should be understood that the term "and / or" in this document is merely used to describe associated relationship, and it can mean three relationships, for example, A and / or B can mean that A exists alone, A and B exist together, and B exists alone, where A and B can be singular or plural. In addition, the character " / " in this document generally means that the associated objects before and after the " / " are in an "or" relationship, but can also mean an "and / or" relationship, which can be understood according to the context before and after.

[0107] In this application, "at least one" means one or more, and "multiple" means two or more. "At least one of the following" or the like means any combination of the items, including any combination of single or multiple items. For example, at least one of a, b, or c can mean a, b, c, a-b, a-c, b-c, or a-b-c, where a, b, and c can be single or multiple.

[0108] It should be understood that in various embodiments of the present application, the size of the sequence number of the above-mentioned processes does not mean the order of execution, and the execution order of the processes should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0109] Those skilled in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0110] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working process of the above-described system, device and unit can refer to the corresponding process in the foregoing method embodiments, which will not be described here.

[0111] In several embodiments provided in the present application, it should be understood that the disclosed system can be realized by other ways. For example, the above-described device embodiments are only schematic, for example, the division of the units is only a logical function division, and actual implementation can have another division manner, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the shown or discussed objects can be indirect coupling or communication connection through some interfaces, devices or units, and can be electrical, mechanical or other forms.

[0112] The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed on multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0113] In addition, the functional units in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit.

[0114] The functions, if realized in the form of software functional units and sold or used as independent products, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application or the part of the present application that essentially contributes to the prior art or the part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various program code storage media.

[0115] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. An online incremental learning method based on adaptive B+ tree index, characterized in that, Comprise: S1: Construct an extended adaptive B+ tree index structure, wherein the sample fields stored in the leaf nodes include feature vectors, timestamps, labels, label confidence markers, and structure encodings, and the state parameters of the leaf nodes include access frequencies, label fluctuation levels, and structure evolution rates; the feature vectors are vectors in a multi-dimensional space obtained by mapping the original data of the samples through a feature extraction function, are high-dimensional representations of the samples, and are inputs for classification, regression, and other tasks, wherein the original sample data includes images, texts; S2: Obtain the feature vector of a new sample, insert it into the B+ tree index structure, and calculate a trajectory disturbance value according to the sequence difference between the insertion path of the new sample and the insertion paths of historical samples; S3: When the new sample is inserted into the B+ tree index structure, generate a disturbance score according to the trajectory disturbance value and the state parameters of the leaf node, and when the disturbance score is greater than a first threshold value and the label confidence marker of the sample in the leaf node is credible or unstable, add the new sample and the sample set in the leaf node to an incremental learning scheduling queue; S4: When a structure variation event is triggered by a node splitting, merging, or restructuring operation, obtain a subset of affected nodes, and update the structure encodings of the nodes in the subset; S5: Obtain the label trajectory of a sample, update the label confidence marker of the sample when the label trajectory changes by more than a second threshold value, set the label confidence marker to unstable or suspicious, and limit the participation frequency of the sample according to a preset semi-supervised screening mechanism; S6: Perform a forgetting compression operation on a leaf node with an access frequency less than a third threshold value and a label fluctuation level less than a fourth threshold value, and delete the samples in the node; Step S2 comprises: Obtain the feature vector of a new sample, insert it into the B+ tree index structure according to a preset indexing rule, record the node sequence during the insertion process, and generate an insertion path; Retrieve a preset number of historical sample insertion paths from the index structure, wherein the historical sample insertion paths are composed of the node access sequences recorded during the insertion process of historical samples; Compare the insertion path with the historical sample insertion paths in terms of sequence difference, calculate a trajectory disturbance value based on the edit distance of the node sequence, the jump depth, and the distance between the structure embedding vectors, and set the trajectory disturbance value to a maximum value when there is no historical sample in the index structure; Step S3 comprises: When the new sample is inserted into the B+ tree index structure, obtain the trajectory disturbance value and the label fluctuation level and structure evolution rate of the leaf node where the new sample is located; Jointly analyze the trajectory disturbance value, the label fluctuation level, and the structure evolution rate to generate a disturbance score; When the disturbance score is greater than a preset first threshold value and the label confidence marker of the sample in the leaf node is credible or unstable, add the new sample and the sample set in the leaf node to an incremental learning scheduling queue. 2.The online incremental learning method based on adaptive B+ tree index according to claim 1, wherein, The timestamp is used to record the time when the sample is inserted into the index structure; the label is used to record the supervised category information of the sample; the label confidence marker is used to record the credibility state of the label source, including reusable, credible, unstable, and suspicious, and is initially set to credible; the structure encoding is used to represent the local structure distribution state of the leaf node where the sample is located, and is generated based on the feature vectors of all samples in the node through an aggregation mapping function; In the state parameters, the access frequency is updated by accumulation every time the node is inserted or queried; the label fluctuation degree is obtained by counting the category distribution entropy or variation amplitude of the sample label in the node within a set time window; and the structure evolution rate is obtained by performing difference calculation on the structure code sequence generated by the node within a continuous time period. 3.The online incremental learning method based on adaptive B+ tree index according to claim 1, wherein, According to the joint analysis of the trajectory disturbance value, the label fluctuation degree and the structure evolution rate, a disturbance score is generated, including: When the trajectory disturbance value is greater than the first disturbance threshold, and the label fluctuation degree is greater than the second disturbance threshold or the structure evolution rate is greater than the third disturbance threshold, it is determined that the disturbance score is high; When the trajectory disturbance value is greater than the first disturbance threshold, and the label fluctuation degree is less than or equal to the second disturbance threshold, and the structure evolution rate is less than or equal to the third disturbance threshold, it is determined that the disturbance score is medium; When the trajectory disturbance value is less than or equal to the first disturbance threshold, it is determined that the disturbance score is low. 4.The online incremental learning method based on adaptive B+ tree index according to claim 1, wherein, In step S4, when the number of samples in the target node is greater than the preset upper threshold after inserting the sample into the target node, a node splitting event is triggered; when the node splitting event occurs, the original node before splitting, the new node generated by splitting and the parent node are set as the affected node subset; When the number of samples in the target node is less than the preset lower threshold after deleting the sample, and the target node can be merged with the adjacent sibling node, a node merging event is triggered; When the node merging event occurs, the nodes participating in the merging and the parent node are set as the affected node subset; When the splitting or merging events occur continuously in the subtree area until the height difference between the left and right subtrees is greater than the preset height difference threshold, a node reconstruction event is triggered; when the node reconstruction event occurs, all nodes participating in the reconstruction are set as the affected node subset; For each node in the affected node subset, the hierarchical position, index serial number, type label of whether it is a leaf node and current number of child nodes of the node are collected, a structure code value is generated by a structure coding function, and the structure code value is written into the structure code field of the corresponding node.

5. The online incremental learning method based on adaptive B+ tree index according to claim 1, characterized in that, Step S5 includes: In the leaf node, a label trajectory record table is established with the feature vector as the primary key; when the label of the sample changes, the current timestamp and label value are obtained, and a label trajectory sequence is generated; The number of occurrences of consecutive different label pairs is counted according to the label trajectory sequence, and is normalized as the ratio of the total trajectory length, which is defined as the label change rate; The total number of label categories is counted according to the label trajectory sequence, and the label value range is calculated in the regression task; When the label change rate and / or the label value range is greater than the second threshold, the label confidence label of the sample is updated to an unstable state, and when it is greater than the second threshold continuously within a continuous preset time window, it is updated to a suspicious state; Based on the updated label confidence label, a preset semi-supervised screening strategy is applied when performing incremental learning scheduling, wherein the sampling probability upper limit is set for the unstable state sample, and the suspicious state sample is rejected to enter the scheduling task queue.

6. The online incremental learning method based on adaptive B+ tree index according to claim 1, characterized in that, Step S6 includes: A fixed-period triggered forgetting compression strategy is set, and leaf nodes are traversed in the period to calculate the access frequency and label fluctuation degree of the nodes in a preset sliding time window, wherein the access frequency is the number of node accesses divided by the window length, and the label fluctuation degree is the average label fluctuation degree of the nodes; When the access frequency of the leaf node is less than a third threshold value and the label fluctuation degree is less than a fourth threshold value, a forgetting compression operation is performed, all sample fields stored in the leaf node are deleted, the node state is reset to reusable, and the structure code is retained.

7. The online incremental learning method based on adaptive B+ tree index according to claim 6, characterized in that, Further comprising: A periodic interpretability output mechanism is set, and running state information of the adaptive B+ tree index structure is collected in a preset sample processing period to generate an interpretability data file including a node access heat map and a structure change atlas; The node access heat map calculates a standardized heat value based on the access frequency of the leaf node in the sliding window to generate a structured index map with heat labels; The structure change atlas records the event type, occurrence time, and node evolution relationship by monitoring node splitting, merging, or reconstruction events in the index structure to generate a time series structure diagram.

8. An online incremental learning system based on adaptive B+ tree index, for implementing the online incremental learning method based on adaptive B+ tree index according to any one of claims 1-7, characterized in that, Comprising: A structure construction module: an extended adaptive B+ tree index structure is constructed, wherein the sample fields stored in the leaf nodes include feature vectors, timestamps, labels, label confidence markers, and structure codes, and the state parameters of the leaf nodes include access times, label fluctuation degrees, and structure evolution rates; A sample insertion module: the feature vector of a new sample is obtained, and the B+ tree index structure is inserted, and the trajectory disturbance value is calculated according to the sequence difference between the insertion path of the new sample and the historical sample insertion path; A score generation module: when the new sample is inserted into the B+ tree index structure, a disturbance score is generated according to the trajectory disturbance value and the state parameters of the leaf node, and when the disturbance score is greater than a first threshold value and the label confidence marker is credible or unstable, the new sample and the sample set of the leaf node are added to an incremental learning scheduling queue; An event update module: when a node splitting, merging, or reconstruction operation triggers a structure change event, the affected node subset is obtained, and the structure code of the nodes in the subset is updated; A marker update module: the label trajectory of a sample is obtained, and the label confidence marker of a sample with a label trajectory change greater than a second threshold value is updated, and the label confidence marker is updated to unstable or suspicious, and the participation frequency of the sample is limited according to a preset semi-supervised screening mechanism; A forgetting compression module: forgetting compression is performed on the leaf node with an access frequency less than a third threshold value and a label fluctuation degree less than a fourth threshold value, and the samples in the node are deleted.

Images