Imbalanced data regression method, system, storage medium and terminal

Abstract

The present invention discloses a method, system, storage medium, and terminal for unbalanced data regression, belonging to the field of data regression technology. The method includes: constructing a multi-granularity space generation model based on grid division, dividing similar samples into the same granularity space; independently oversampling in each granularity space, wherein the oversampling range is defined by combining spatial importance and dynamic neighborhood radius. In scenarios with large sample sizes, the present invention comprehensively considers data flow and imbalance issues, improving processing efficiency while taking into account the overall distribution of data, reducing the interference of noise samples, and retaining the continuity and intrinsic correlation of samples, providing new ideas for the field of unbalanced data stream regression.

Description

Technical Field

[0001] The present invention relates to the field of data regression technology, and in particular to an unbalanced data regression method, system, storage medium and terminal. Background Art

[0002] Traditional standard machine learning algorithms consider classes with roughly equal numbers of samples, and each class is equally important. However, data in various fields in the real world suffers from significant imbalances. Industrial data collected from mechanical equipment in the industrial sector is particularly imbalanced and complex. Examples include mechanical equipment failure information, abnormal data from disaster monitoring systems, and product quality inspection failure data. In these real-world application scenarios, predicting rare events is crucial, and data imbalance can severely impact the accuracy of rare event predictions. Traditional machine learning algorithms are typically designed to reduce overall error. However, in the case of imbalanced data, they tend to favor predictions for the majority class, neglecting rare events in the minority class. For example, standard logistic regression methods often underestimate the probability of rare events when working with imbalanced datasets, resulting in low recall and insufficient event prediction. The data characteristics of imbalanced regression also complicate model learning, making it difficult for the model to capture the patterns and regularities of rare events, which in turn affects predictive performance. This makes monitoring anomalies in various fields challenging, and mispredictions can lead to significant cost. Therefore, regression methods for imbalanced data are particularly important.

[0003] Most existing methods for solving the imbalanced regression problem divide the samples in the regression task into different categories. This method has two drawbacks:

[0004] 1. Dividing continuous target values ​​into different categories disrupts data continuity and loses important information at the category boundaries. This can easily introduce noise samples during oversampling, reducing the model's prediction accuracy and stability. 2. Data oversampling requires multiple calculations of distances between samples, which can reduce processing efficiency when the number of samples is large. Therefore, an effective approach to addressing imbalanced regression is to reduce the sample size and the search space for sample neighbors while also considering the overall sample distribution, preserving data continuity, and avoiding the interference of noise samples.

[0005] Furthermore, existing imbalanced data processing methods typically deal with static data. That is, once a model is trained and finalized, its parameters and structure are not dynamically adjusted, making it unable to learn the characteristics of newly added data in real time. However, most real-world data is dynamic data that appears in the form of streams, and its data distribution changes over time. Unbalanced stream data is widely used in fields such as network monitoring and fault detection. Online learning is a key technology for processing stream data, but online real-time learning of data streams faces several difficulties: 1. Data distribution changes over time; 2. Intrinsic structural instability leads to concept drift; 3. Data imbalance is unstable; and 4. The model cannot process newly added data in real time. For the classification problem of imbalanced data streams, there are few solutions that comprehensively consider both the data stream and the imbalance problem, and they are not suitable for direct improvement for regression problems.

[0006] To address this, researchers have proposed a variety of algorithms to address issues such as uneven class distribution, concept drift, and a scarcity of minority class samples in dynamic data streams. These methods, ranging from data sampling to classifier optimization, effectively improve learning capabilities for imbalanced data by dynamically adjusting models or data. These algorithms, such as C-SMOTE, CSARF, VFCSMOTE, and ROSE, offer some examples. However, while these imbalanced data regression methods improve processing efficiency, they still suffer from the following drawbacks:

[0007] 1. The idea behind most methods for solving imbalanced regression is to assign a correlation scale to each sample based on the correlation function, dividing the samples into rare and common samples to suit imbalanced classification algorithms. This approach destroys the continuity and inherent correlation of the sample target values, and when the threshold is not selected properly, the resulting sample misclassification will affect the model's predictive performance; 2. Existing clustering-based methods for reducing sample size to improve processing efficiency are highly dependent on clustering parameters and clustering results. When the clustering parameters are not selected properly or the data distribution is irregular, the clustering effect will be poor, affecting the subsequent sampling process; 3. The data distribution in actual application scenarios such as industry and manufacturing is very complex. The traditional method of oversampling boundary samples has difficulty identifying complex decision boundaries and is prone to overlooking some important samples. Summary of the Invention

[0008] The purpose of the present invention is to study the unbalanced data regression method in the scenario of large sample size. In the scenario of large sample size, the existing unbalanced data regression method is difficult to take into account the overall distribution of data while improving the processing efficiency, is easily interfered by noise samples, and destroys the continuity and intrinsic correlation of the samples. An improved granular sphere model (grid self-organizing granular sphere model) is proposed to process the unbalanced regression data with a large sample size; at the same time, the existing unbalanced data regression method is limited to processing static data and cannot process dynamic data that appears in a stream manner. An unbalanced data flow regression method is realized in the regression task scenario that comprehensively considers data flow and imbalance problems.

[0009] The object of the present invention is achieved through the following technical solutions:

[0010] In a first aspect, a method for imbalanced data regression is provided, comprising:

[0011] Construct a multi-granularity space generation model based on grid division to divide similar samples into the same granularity space;

[0012] Oversampling is performed independently in each granularity space, where the oversampling range is defined by combining spatial importance and dynamic neighborhood radius.

[0013] In some embodiments, constructing a multi-granularity spatial generation model based on grid partitioning includes:

[0014] Determine the division area of ​​the granular space according to the spatial area formed by the boundary range of each dimension of the sample;

[0015] Subdivide the partitioned area into several small rectangles;

[0016] According to the number of sparsely distributed samples in each small rectangle, the small rectangles are divided into different types;

[0017] Calculate the distribution degree and minimum distribution degree of the particle size space. If the distribution degree of the particle size space is less than the minimum distribution degree, continue to divide the particle size space until the distribution degree of all particle size spaces is not less than the minimum distribution degree.

[0018] In some embodiments, the types of the small rectangles include a blank rectangle with a sample number of 0, a high-density rectangle with a maximum sample number, and a sparse rectangle between the high-density rectangle and the blank rectangle.

[0019] In some embodiments, if the distribution degree of the granularity space is less than the minimum distribution degree, continuing to divide the granularity space includes:

[0020] For the granularity space that needs to be further divided, its center and the center of the sample in the high-density rectangle are selected as the center points of two new granularity spaces respectively. The distance from each sample to the two center points is calculated, and the sample is assigned to the granularity space to which the nearest center point belongs, thus obtaining two new granularity spaces.

[0021] In some embodiments, the calculation of the spatial importance includes:

[0022] Calculate the sample density of the continuous target value of the sample;

[0023] Use the weighting function to assign weights to the sample density and obtain the density weight of each sample;

[0024] The spatial importance is calculated based on the density weight of each sample.

[0025] In some embodiments, the combining of spatial importance and dynamic neighborhood radius to define the oversampling range includes:

[0026] The amount of oversampling is determined by the following formula:

[0027] ,in, represents the amount of oversampling, represents the neighborhood radius, Indicates the importance of space.

[0028] In some embodiments, the specific process of oversampling includes:

[0029] when When , the samples with higher importance in the granularity space are used as candidate seed samples for oversampling;

[0030] Select a seed sample, select a neighbor sample within its neighborhood radius, and perform linear interpolation to synthesize a new sample;

[0031] The seed samples are cyclically selected to synthesize new samples until the number of synthesized new samples reaches the oversampling number.

[0032] In a second aspect, an imbalanced data regression system is provided, comprising:

[0033] Multi-granularity space partitioning module, used to build a multi-granularity space generation model based on grid partitioning, dividing similar samples into the same granularity space;

[0034] The oversampling module is used to independently perform oversampling in each granularity space, wherein the oversampling range is defined in combination with the spatial importance and the dynamic neighborhood radius.

[0035] In a third aspect, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the imbalanced data regression method described in any one of the first aspects is implemented.

[0036] In a fourth aspect, a terminal is provided, comprising a memory and a processor, wherein the memory stores computer instructions that can be executed on the processor, and when the processor executes the computer instructions, the imbalanced data regression method described in any one of the first aspects is executed.

[0037] It should be further explained that the technical features corresponding to the above embodiments can be combined or replaced with each other to form a new technical solution if there is no conflict.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] 1. In scenarios with larger sample sizes, the present invention constructs a multi-granularity space generation model based on grid partitioning, which groups similar samples into the same granularity space. This improves processing efficiency while also taking into account the overall distribution of the data. The multi-granularity space generation model can iteratively granulate large-scale samples into multiple granularity spaces based on the overall distribution of the data, using the overall space to replace multiple similar samples, ensuring that each resulting partitioned sample is evenly distributed. The subsequent sampling process is performed separately within each space, effectively avoiding sampling overlapping areas between different sample distributions and avoiding the generation of noisy samples.

[0040] 2. The present invention combines spatial importance and dynamic neighborhood radius to define the oversampling range. Subsequent samples are oversampled within the defined oversampling range, which can reduce the possibility of noise samples generated by oversampling within the sphere and avoid destroying the continuity and intrinsic correlation of the samples.

[0041] 3. The multi-granularity space generation model of the present invention can dynamically update the granularity space and sampling strategy as the data flows, and adapt to the dynamic change characteristics of the data flow itself. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A flow chart of an unbalanced data regression method according to an embodiment of the present invention;

[0043] Figure 2 A schematic diagram of sample distribution according to an embodiment of the present invention;

[0044] Figure 3 A schematic diagram of a grid area of ​​a granularity space according to an embodiment of the present invention;

[0045] Figure 4 The blank portion in the rectangular area shown for the embodiment of the present invention where no sample exists;

[0046] Figure 5 This is the result of dividing the grid area of ​​the granularity space using small rectangles shown in an embodiment of the present invention;

[0047] Figure 6 Schematic diagrams of three types of rectangles shown in an embodiment of the present invention;

[0048] Figure 7 The initial granularity space shown in the embodiment of the present invention;

[0049] Figure 8 The embodiment of the present invention is based on Figure 7 The initial granularity space is divided into two new granularity spaces;

[0050] Figure 9 The embodiment of the present invention is based on Figure 8 The two new granularity spaces are divided into four new granularity spaces;

[0051] Figure 10 The embodiment of the present invention is based on Figure 9 The four new granularity spaces are divided into eight new granularity spaces;

[0052] Figure 11 The embodiment of the present invention is based on Figure 10 The eight new granularity spaces are divided into sixteen new granularity spaces;

[0053] Figure 12 The embodiment of the present invention is based on Figure 11 The sixteen new granularity spaces are divided into thirty-two new granularity spaces;

[0054] Figure 13 This is a schematic diagram of oversampling within a sample neighborhood according to an embodiment of the present invention. DETAILED DESCRIPTION

[0055] The technical solutions of the present invention are described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present application generally described and shown in the drawings herein can be arranged and designed in various different configurations. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0056] It should be noted that the defects existing in the solutions in the above-mentioned prior art are the results obtained by the inventor after practice and careful research. Therefore, the discovery process of the above-mentioned problems and the solutions proposed in the embodiments of this application below for the above-mentioned problems should be the contributions made by the inventor to this application in the process of invention and creation, and should not be understood as technical contents known to technical personnel in this field.

[0057] In response to the technical problems pointed out in the background technology, the embodiments provided by the present invention are as follows:

[0058] Reference Figure 1 In an exemplary embodiment, a method for unbalanced data regression is provided, comprising:

[0059] Construct a multi-granularity space generation model based on grid division to divide similar samples into the same granularity space;

[0060] Oversampling is performed independently in each granularity space, where the oversampling range is defined by combining spatial importance and dynamic neighborhood radius.

[0061] Collect sensor data of a mechanical device and give a sample set of granularity space GS , ,in, represents the m-dimensional feature vector of the sample, Represents the category label of the sample, and the center O of GS is calculated by the following formula: .

[0062] Assume that the sample distribution of the granularity space GS is as follows Figure 2 As shown, the space partitioning strategy is defined as follows:

[0063] Definition 1 (division area): The division area of ​​the granular space GS is the spatial area composed of the boundary range of each dimension of the sample. The boundary of the j-th dimension of the sample space is is defined as follows:

[0064] ,in, Represents the value of the sample center O in the jth dimension, It represents the maximum distance from the sample in the divided space to the center of the space O. The calculation formula is as follows:

[0065] .

[0066] According to Definition 1, the range covered by the divided space region in each dimension is , Figure 3 A schematic diagram of the grid area of ​​the granularity space GS is given. represents the th dimension of the sample space.

[0067] Definition 2 (divided space side length): The divided space area is subdivided into several small rectangles. The side length of each small rectangle is defined as follows:

[0068] .

[0069] According to Definition 1 and Definition 2, the number of small grids that can be subdivided into a spatial region is . Figure 4 The gray area in the figure can roughly represent the blank part of the rectangular area where there is no sample. The spatial area division strategy can use small grids to accurately represent the proportion of samples in the grid area. Figure 5 The results of partitioning the grid area of ​​the granularity space GS using small rectangles are shown.

[0070] Definition 3: After the granular space area is divided into grids, the small rectangles are divided into the following categories according to the number of sparsely distributed samples in each small rectangle: 1) the rectangle with 0 samples is called a "blank rectangle"; 2) the grid with the largest number of samples is called a "high-density rectangle"; 3) the remaining grids are called "sparse rectangles". Figure 6 shown.

[0071] Definition 4 (particle size distribution and minimum distribution): After the particle size space is divided into grids, the number of blank rectangular grids obtained is , particle size distribution F and minimum distribution is defined as follows:

[0072] , .

[0073] Based on Definition 4, the granularity space partitioning strategy of the present invention is as follows:

[0074] 1) Given a regression dataset ,in Represents a sample, Indicates the category label of the corresponding sample. Initially, the dataset is used as the sample set of the initial granularity space GS, and the space partitioning strategy is used to divide its area into rectangular grids;

[0075] 2) Calculate the distribution F of the particle size space GS, if , then GS terminates the division and becomes the final granularity space structure; if , then GS needs to be further divided;

[0076] 3) For the granularity space GS that needs to be further divided, select the center O of GS and the center of the sample in the "high-density rectangle" The two centers divide the space into two parts. As two new spaces and The distance between each sample and the two center points is calculated, and the sample is assigned to the particle ball to which the nearest center point belongs, and two new granularity division spaces are obtained. and ;

[0077] 4) respectively and Repeat steps 2) to 4) until all granular spaces are no longer divided.

[0078] According to the rectangular grid division strategy of granularity space, Figure 7-12 The schematic diagram of the granulation process of the new granularity space partitioning strategy on the four-category dataset is shown. Figure 7 represents the initial granularity space, Figure 8 Based on Figure 7 The initial granularity space is divided into two new granularity spaces. Figure 9 Based on Figure 8 The two new granularity spaces are divided into four new granularity spaces. Figure 10 Based on Figure 9 The four new granularity spaces are divided into eight new granularity spaces. Figure 11 Based on Figure 10 The eight new granularity spaces are divided into sixteen new granularity spaces. Figure 12 Based on Figure 11 The sixteen new granularity spaces are divided into thirty-two new granularity spaces.

[0079] In imbalanced regression problems, users pay more attention to rare samples, so measuring the rarity of samples is crucial. We calculate the density function of the continuous target value of each sample and use a weighting function to assign weights to the sample density. This yields a density weight for each sample, which reflects its importance.

[0080] Based on the regression dataset ,The relevant definitions of granularity space importance are as follows.

[0081] Definition 5 (sample density): the density of the sample as follows:

[0082] , where KDE(*) represents the kernel function, which is used to estimate the probability density function of the continuous target value, and h represents the bandwidth. Further, the density of the sample is normalized by min-max, and express The range of the value of , then the normalized sample density as follows:

[0083] .

[0084] Definition 6 (Sample Importance): The importance of a sample as follows:

[0085] ,in, is a hyperparameter that determines the degree of weight increase for important samples. is a small positive constant, which avoids the situation where the weight is 0 or negative.

[0086] Definition 7 (spatial importance): Spatial importance is a weighted function, the importance of the granularity space GS as follows:

[0087] .

[0088] Based on the importance of the granular space, the present invention can quickly find samples with high importance. After representing samples as multiple granular spaces based on the rectangular grid partitioning model, it can be assumed that the samples within the granular space are evenly distributed and highly similar, and therefore the samples also have similar importance. Based on the spatial importance, the present invention can quickly find samples with high importance and perform corresponding oversampling processing on these samples based on the partition unit.

[0089] Definition 8 (Sample Neighborhood Radius): The neighborhood radius R of a sample is as follows:

[0090] ,in, is the side length of the rectangular region in the granularity space defined in 2.

[0091] Furthermore, the number of oversampling in the granularity space GS is given The calculation formula is:

[0092] ,in, represents the amount of oversampling, represents the neighborhood radius, Indicates the importance of space.

[0093] Combining the importance of granular space and the neighborhood radius of the sample, the oversampling strategy of granular space GS is as follows:

[0094] 1) When When , it means that the importance of most samples in GS is higher than the average level. These samples are used as candidate seed samples for oversampling, and the importance of the samples is used as the selection weight, which can reduce the probability of a few samples with lower importance in GS becoming seed samples.

[0095] 2) Select a seed sample, such as Figure 13 The sample at the center of the dotted circle selects a neighbor sample within its neighborhood radius R Perform linear interpolation to synthesize new samples ;

[0096] 3) Repeat step 2) until the number of synthesized new samples reaches .

[0097] This invention targets unbalanced regression datasets with large sample sizes. It groups similar samples into the same rectangle, reducing the sample size. While employing a divide-and-conquer approach to improve processing efficiency, it also avoids the possibility of noise samples being generated between rectangles due to oversampling. It also improves upon existing granular space partitioning models, quantifying the sample distribution based on a rectangular grid of sample sets, ultimately ensuring a uniform distribution of samples within each partition. Furthermore, it uses the probability density function of the sample's continuous target value to calculate the sample's importance. The subsequent oversampling range is based on the dynamically calculated neighborhood radius and sample importance, avoiding disruptions to the sample's continuity and inherent correlation.

[0098] In another exemplary embodiment, based on the same inventive concept as the method embodiment, an unbalanced data regression system is provided, including:

[0099] Multi-granularity space partitioning module, used to build a multi-granularity space generation model based on grid partitioning, dividing similar samples into the same granularity space;

[0100] The oversampling module is used to independently perform oversampling in each granularity space, wherein the oversampling range is defined in combination with the spatial importance and the dynamic neighborhood radius.

[0101] In another exemplary embodiment, based on the same inventive concept as the method embodiment, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, an unbalanced data regression provided by an embodiment of the present invention is implemented. Based on this understanding, the technical solution of this embodiment, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of various embodiments of the present invention. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0102] In another exemplary embodiment, based on the same inventive concept as the method embodiment, a terminal is provided, including a memory and a processor, wherein the memory stores computer instructions that can be executed on the processor, and when the processor executes the computer instructions, an imbalanced data regression method provided by an embodiment of the present invention is executed.

[0103] The processor may be a single-core or multi-core central processing unit or a specific integrated circuit, or one or more integrated circuits configured to implement the present invention.

[0104] Embodiments of the subject matter and functional operations described in this specification may be implemented in: tangibly embodied computer software or firmware, computer hardware including the structures disclosed in this specification and their structural equivalents, or a combination of one or more thereof. Embodiments of the subject matter described in this specification may be implemented as one or more computer programs, i.e., one or more modules of computer program instructions encoded on a tangible, non-transitory program carrier for execution by a data processing apparatus or to control the operation of the data processing apparatus. Alternatively or in addition, the program instructions may be encoded on an artificially generated propagated signal, such as a machine-generated electrical, optical, or electromagnetic signal, that is generated to encode and transmit information to a suitable receiver apparatus for execution by the data processing apparatus.

[0105] The processes and logic flows described in this specification can be performed by one or more programmable computers executing one or more computer programs to perform the corresponding functions by operating on input data and generating output. The processes and logic flows can also be performed by, and apparatus can be implemented as, special purpose logic circuitry, such as an FPGA (field programmable gate array) or an ASIC (application-specific integrated circuit).

[0106] Processors suitable for executing computer programs include, for example, general-purpose and / or special-purpose microprocessors, or any other type of central processing unit. Typically, a central processing unit will receive instructions and data from a read-only memory and / or random access memory. The basic components of a computer include a central processing unit for implementing or executing instructions and one or more memory devices for storing instructions and data. Typically, a computer will also include one or more mass storage devices for storing data, such as magnetic disks, magneto-optical disks, or optical disks, or the computer will be operably coupled to such a mass storage device to receive data from it or to transmit data to it, or both. However, a computer does not necessarily have such a device. In addition, a computer can be embedded in another device, such as a mobile phone, a personal digital assistant (PDA), a mobile audio or video player, a game console, a global positioning system (GPS) receiver, or a portable storage device such as a universal serial bus (USB) flash drive, to name a few.

[0107] It should be understood that each box in the flow chart or block diagram can represent a module, program segment or a part of code, and the part of the module, program segment or code comprises one or more executable instructions for realizing the logical function of the provision. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a sequence different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs the function or action of the provision, or can be implemented with a combination of dedicated hardware and computer instructions.

[0108] The above specific implementation methods are detailed descriptions of the present invention. It cannot be considered that the specific implementation methods of the present invention are limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, they can make several simple deductions and substitutions without departing from the concept of the present invention, which should be regarded as falling within the scope of protection of the present invention.

Claims

1. A method for unbalanced data regression, wherein the unbalanced data includes fault information of mechanical equipment, abnormal data of disaster monitoring system, and unqualified data of product quality inspection, characterized in that: include: Construct a multi-granularity space generation model based on grid division to divide similar samples into the same granularity space; The construction of a multi-granularity space generation model based on grid division includes: Determine the division area of ​​the granular space according to the spatial area formed by the boundary range of each dimension of the sample; Subdivide the partitioned area into several small rectangles; According to the number of sparsely distributed samples in each small rectangle, the small rectangles are divided into different types; Calculate the distribution degree and minimum distribution degree of the particle size space. If the distribution degree of the particle size space is less than the minimum distribution degree, continue to divide the particle size space until the distribution degree of all particle size spaces is not less than the minimum distribution degree. Oversampling is performed independently in each granularity space, wherein the oversampling range is defined by combining the spatial importance and the dynamic neighborhood radius; the oversampling range is defined by combining the spatial importance and the dynamic neighborhood radius, including: The amount of oversampling is determined by the following formula: ,in, represents the amount of oversampling, represents the neighborhood radius, Indicates spatial importance; the calculation formula of the neighborhood radius is as follows: ,in, is the side length of the small rectangle, It represents the importance of samples in the partition space and is calculated by the following formula: ,in, is a hyperparameter, is a positive constant, represents the normalized sample density, represents the density of the sample, express The range of the value, the density of the sample The calculation is as follows: , where KDE(*) represents the kernel function, represents bandwidth, n represents the number of samples, y represents the target label, Indicates the category label of the corresponding sample.

2. The imbalanced data regression method according to claim 1, characterized in that: The types of the small rectangles include a blank rectangle with a sample number of 0, a high-density rectangle with a maximum sample number, and a sparse rectangle between the high-density rectangle and the blank rectangle.

3. The imbalanced data regression method according to claim 2, characterized in that: If the distribution degree of the granularity space is less than the minimum distribution degree, the granularity space is further divided, including: For the granularity space that needs to be further divided, its center and the center of the sample in the high-density rectangle are selected as the center points of two new granularity spaces respectively. The distance from each sample to the two center points is calculated, and the sample is assigned to the granularity space to which the nearest center point belongs, thus obtaining two new granularity spaces.

4. The imbalanced data regression method according to claim 1, wherein: The calculation of the spatial importance includes: Calculate the sample density of the continuous target value of the sample; Use the weighting function to assign weights to the sample density and obtain the density weight of each sample; The spatial importance is calculated based on the density weight of each sample.

5. The imbalanced data regression method according to claim 1, wherein: The specific process of oversampling includes: when When , the samples with higher importance in the granularity space are used as candidate seed samples for oversampling; Select a seed sample, select a neighbor sample within its neighborhood radius, and perform linear interpolation to synthesize a new sample; The seed samples are cyclically selected to synthesize new samples until the number of synthesized new samples reaches the oversampling number.

6. An unbalanced data regression system, used to implement the unbalanced data regression method according to any one of claims 1 to 5, characterized in that: include: Multi-granularity space partitioning module, used to build a multi-granularity space generation model based on grid partitioning, dividing similar samples into the same granularity space; The oversampling module is used to independently perform oversampling in each granularity space, wherein the oversampling range is defined in combination with the spatial importance and the dynamic neighborhood radius.

7. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the unbalanced data regression method according to any one of claims 1 to 5 is implemented.

8. A terminal comprising a memory and a processor, wherein the memory stores computer instructions that can be executed on the processor, characterized in that: When the processor runs the computer instructions, it performs the imbalanced data regression method described in any one of claims 1 to 5.

Images