Learning model generating apparatus, and learning model generating method

By grouping feature data and using a subset of data for training, the method addresses over-detection and memory issues in anomaly detection models, achieving reduced model size and improved computational efficiency.

JP2024165067APending Publication Date: 2024-11-28NEC CORP +1
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
JP2023080904
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-05-16
Publication Date
2024-11-28

AI Technical Summary

Technical Problem

Existing learning models for anomaly detection in wireless communication systems face issues with over-detection and memory requirements due to reliance on unsupervised learning with normal data, leading to large model sizes and computational demands.

Method used

A learning model generation method that groups feature data into multiple subsets, using a subset of data for training and incorporating abnormal data, reducing the overall model size and memory requirements while maintaining detection accuracy.

Benefits of technology

The method effectively reduces the size of the learning model and memory needs, preventing over-detection and improving computational efficiency without compromising detection accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2024165067000001_ABST
    Figure 2024165067000001_ABST
Patent Text Reader

Abstract

To enable reduction of a learning model in size and to make a size of a memory required for learning smaller.SOLUTION: A learning model generating apparatus 10 according to the present invention has data dividing means 11 for dividing a plurality of pieces of feature amount data each indicating a feature amount into groups, and learning model generating means 12 for generating a learning model by using, as learning data, the feature amount data belonging to a first group among a plurality of groups formed by the data dividing means 11 or the feature amount data belonging to the first group and a part of the feature amount data belong to other groups.SELECTED DRAWING: Figure 5
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to a learning model generation device and a learning model generation method for generating a learning model for anomaly detection. [Background technology]

[0002] Wireless communication using radio waves is used in various fields. Accordingly, it is important to detect radio interference and failures in wireless communication systems, that is, anomaly detection. Anomaly detection is sometimes performed using machine learning (for example, see Patent Documents 1 and 2). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2019-159957 A [Patent Document 2] JP 2022-182844 A Summary of the Invention [Problem to be solved by the invention]

[0004] Although it is difficult to collect abnormal data of a target, normal data can be collected relatively easily. Therefore, when using machine learning to detect anomalies, a model (machine learning model, hereafter referred to as the learning model) is often trained by unsupervised machine learning using normal data as training data. Note that normal data is data obtained when there is no radio interference or failure. Abnormal data is data obtained when there is radio interference or failure.

[0005] If learning is performed using only normal data, there is a possibility that overdetection will occur when using a trained learning model, in which data that should be judged as normal is judged as abnormal. There is also a possibility of detection failure.

[0006] To reduce the possibility of false positives and false negatives, a large amount of training data collected in various situations is required, and the size of the training model becomes large.

[0007] Since a large amount of training data is used, the memory size to be prepared is large. Also, since the size of the training model is large, the performance required for the computer to realize the training model is high. In other words, an expensive computer is required.

[0008] An object of the present invention is to provide a learning model generation device and a learning model generation method that can reduce the size of a learning model and the size of memory that needs to be prepared for learning. [Means for solving the problem]

[0009] The learning model generation device according to the present invention includes a data division means for dividing a plurality of feature data, each of which indicates a feature, into groups, and a learning model generation means for generating a learning model using, as learning data, the feature data belonging to a first group among the plurality of groups formed by the data division means, or the feature data belonging to the first group and a portion of the feature data belonging to another group.

[0010] The learning model generation method according to the present invention groups a plurality of feature data, each of which indicates a feature, and generates a learning model using, as learning data, the feature data belonging to a first group among the plurality of groups formed, or the feature data belonging to the first group and a portion of the feature data belonging to another group.

[0011] The learning model generation program according to the present invention causes a computer to execute a process of grouping a plurality of feature data, each of which indicates a feature, and a process of generating a learning model using, as learning data, the feature data belonging to a first group among the plurality of groups formed, or the feature data belonging to the first group and a portion of the feature data belonging to another group. Effect of the Invention

[0012] According to the present invention, the size of the learning model can be reduced, and the size of the memory to be prepared for learning can also be reduced. [Brief description of the drawings]

[0013] [Figure 1] FIG. 1 is a block diagram showing a configuration of an embodiment of a learning model generating device. [Diagram 2] FIG. 2 is an explanatory diagram for explaining the processing of a learning model generation unit. [Diagram 3] 4 is a flowchart showing the operation of the learning model generating device. [Figure 4] 1 is a block diagram showing an example of the configuration of an information processing device capable of realizing the functions of a learning model generating device. [Diagram 5] FIG. 2 is a block diagram showing the main parts of a learning model generating device. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0014] Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

[0015] Fig. 1 is a block diagram showing a configuration of an embodiment of a learning model generation device that generates a learning model for anomaly detection. The learning model generation device 100 shown in Fig. 1 includes a feature detection unit 110, a data storage unit (feature storage unit) 120, a data division unit 130, a divided data set storage unit 140, a learning model generation unit 150 that generates a learning model 200, an anomaly determination unit 160, and an anomaly data storage unit 170. Note that the arrows in Fig. 1 simply indicate the direction of signal (data) flow, but do not exclude bidirectionality. This also applies to other block diagrams.

[0016] The learning model 200 is an estimation model that estimates whether input data is normal data or abnormal data after generation, i.e., after learning. In other words, the learning model 200 after learning can be used as an estimation model for anomaly detection.

[0017] Anomaly detection is the act of judging whether input data is normal or abnormal. Below, we take as an example a case in which it is judged whether radio interference or a fault (for example, a breakdown or malfunction of a system or device) is occurring in the field of wireless communication that uses radio waves. In this case, abnormal data is data acquired when radio interference or a fault is occurring. Moreover, data acquired when radio interference or a fault is not occurring is normal data.

[0018] During anomaly detection operation, i.e., in the estimation phase, the trained learning model 200 is used to output estimation data indicating an estimation result as to whether input data is normal or abnormal. Anomaly detection processing is executed based on the estimation data. The anomaly detection processing is, for example, processing for detecting the occurrence of radio interference or a fault.

[0019] Collected data is input to the feature detection unit 110. When determining whether radio interference, a fault, or the like is occurring, the input data is received data. The received data is, for example, data output from a receiving unit (not shown) that receives radio waves.

[0020] The feature detection unit 110 extracts a feature from each of the input data. When determining whether radio interference or a fault occurs, the feature is a feature for determining whether radio interference is occurring or not, or a feature for determining whether a fault occurs or not. Specifically, the feature is, for example, an amount indicating statistical information including information in the frequency direction and the time direction. As an example, the feature can be an amplitude probability distribution (APD), a cumulative distribution function (CDF), an amplitude histogram, or a frequency spectrum.

[0021] The feature detection unit 110 stores the extracted multiple feature amounts in the data storage unit 120 as multiple feature amount data.

[0022] The data division unit 130 divides the feature amount data group stored in the data storage unit 120 into a plurality of groups. Hereinafter, a collection of data included in a group is referred to as a divided data set. The data division unit 130 stores the divided data set in the divided data set storage unit 140.

[0023] The learning model 200 is generated by repeating machine learning using learning data (training data). Machine learning techniques include random forests, support vector machines, neural networks, and deep neural networks. The learning model generation unit 150 generates the learning model 200 by using the divided datasets stored in the divided dataset storage unit 140 as learning data. When generating the learning model 200, the learning model generation unit 150 may use abnormal data in addition to the divided datasets.

[0024] Hereinafter, the abnormal data used when learning the learning model 200 is referred to as "learning abnormal data", and the feature data when the estimation result by the learning model 200 is abnormal is referred to as "determined abnormal data".

[0025] The abnormality determination unit 160 stores an abnormal data group, which is a collection of determined abnormal data, in the abnormal data storage unit 170. Since the learning model 200 outputs a normal / abnormal determination result, the learning model 200 can also be considered as a part of the abnormality determination unit 160.

[0026] FIG. 2 is an explanatory diagram for explaining the process (learning phase) of the learning model generation unit 150. As shown in FIG.

[0027] Each of x1 to x4 shown in Fig. 2 indicates a divided data set. Note that the number of divided data sets, that is, the number of divisions of the input data, n (n≧2), is arbitrary, but Fig. 2 illustrates the case where n=4.

[0028] In the example shown in Fig. 2, the learning model generation unit 150 first trains the learning model 200 using the feature amount data included in the divided data set x1 as learning data. The learning model 200 after the completion of the learning is set as the learning model y1 (see (1) in Fig. 2). Note that in this embodiment, the learning is assumed to be unsupervised learning, but the learning is not limited to unsupervised learning. In addition, general principal component analysis, cluster analysis, self-organizing map (SOM), etc. can be used as the unsupervised learning.

[0029] The abnormality determination unit 160 performs an abnormality determination on the feature amount data included in the divided data set x2 using the learning model y1. The abnormality determination unit 160 stores the feature amount data group determined to be abnormal, i.e., the determined abnormal data group, in the abnormal data storage unit 170.

[0030] Next, the learning model generation unit 150 trains the learning model y1 using the feature amount data included in the divided data set x1 and the feature amount data included in the abnormal data group stored in the abnormal data storage unit 170 as learning data. The learning model 200 after the learning is completed is referred to as the learning model y 1-2 (See (2) in Figure 2.)

[0031] At the time of training the learning model y1, the abnormal data group stored in the abnormal data storage unit 170 is the feature amount data determined to be abnormal among the feature amount data included in the divided data set x2. Also, in FIG. 2, a collection of the feature amount data included in the divided data set x1 and the feature amount data determined to be abnormal among the feature amount data included in the divided data set x2 is 1-2 As shown in the figure.

[0032] The abnormality determination unit 160 uses the learning model y 1-2 The abnormality determination unit 160 stores the feature amount data group determined to be abnormal, that is, the determined abnormal data group, in the abnormal data storage unit 170.

[0033] Next, the learning model generation unit 150 1-2 The feature amount data included in the abnormal data group stored in the abnormal data storage unit 170 is used as learning data to form a learning model y 1-2 The learning model 200 after the learning is completed is called the learning model y 1-3 (See (3) in Figure 2.)

[0034] In addition, the learning model y 1-2 At the time of learning, the abnormal data group stored in the abnormal data storage unit 170 is the feature amount data determined to be abnormal among the feature amount data included in the split data set x3. Also, in FIG. 2, a collection of the feature amount data included in the split data set x1, the feature amount data determined to be abnormal among the feature amount data included in the split data set x2, and the feature amount data determined to be abnormal among the feature amount data included in the split data set x3 is represented as x 1-3 As shown in the figure.

[0035] The abnormality determination unit 160 uses the learning model y 1-3The abnormality determination unit 160 stores the feature amount data group determined to be abnormal, that is, the determined abnormal data group, in the abnormal data storage unit 170.

[0036] Next, the learning model generation unit 150 1-3 The feature amount data included in the abnormal data group stored in the abnormal data storage unit 170 is used as learning data to form a learning model y 1-3 The learning model 200 after the learning is completed is called the learning model y 1-4 (See (4) in Figure 2.)

[0037] In addition, the learning model y 1-3 At the time of learning, the abnormal data group stored in the abnormal data storage unit 170 is the feature amount data determined to be abnormal among the feature amount data included in the split data set x4. Also, in FIG. 2, a collection of the feature amount data included in the split data set x1, the feature amount data determined to be abnormal among the feature amount data included in the split data set x2, the feature amount data determined to be abnormal among the feature amount data included in the split data set x3, and the feature amount data determined to be abnormal among the feature amount data included in the split data set x4 is called x 1-4 As shown in the figure.

[0038] In the example shown in FIG. 2, the number of divisions is 4. When the number of divisions exceeds 4 (when n>4), the learning model generation unit 150 and the anomaly determination unit 160 may repeatedly execute a process of training the learning model 200 using the feature amount data included in the divided data set and the anomaly data obtained in the processing up to that point as learning data, and a process of detecting anomalies in the feature amount data included in the divided data set that has not yet been used. Finally, the learning model y 1-n is obtained.

[0039] Learning model y 1-nWhen obtaining n, not all of the input data is used as learning data in the learning phase. In other words, not all of the feature data included in the n split data sets is used as learning data. In the example shown in FIG. 2, only the feature data included in the first split data set x1 is used as learning data.

[0040] For the split data sets x2, x3, and x4 other than the split data set x1, only feature data determined to be abnormal data is used as learning data. Therefore, the number of feature data used as learning data is smaller than when all of a large amount of feature data is used as learning data for the purpose of preventing overdetection, etc. Therefore, the size of the learning model can be reduced. Also, the size of the memory to be prepared in the learning phase can be reduced.

[0041] In addition, the feature vector data included in the divided data sets x2, x3, and x4 that are determined to be normal data may be similar to each other. Even if a large number of similar feature vector data are used as learning data, the accuracy of the learning model does not improve. In other words, in this embodiment, when generating a learning model, although the number of learning data is small, it is expected that a learning model (trained learning model) with the same level of accuracy as when all of the large amount of feature vector data is used as learning data can be obtained.

[0042] Next, the operation of the learning model generation device 100 will be described with reference to the flowchart in FIG.

[0043] The feature amount detection unit 110 extracts feature amounts from input data (step S101). The feature amount detection unit 110 stores, in the data storage unit 120, feature amount data indicating the extracted feature amounts.

[0044] The data dividing unit 130 divides the feature amount data group stored in the data storage unit 120 into a plurality of groups (divided data sets) (step S102). The data dividing unit 130 stores the divided data sets in the divided data set storage unit 140.

[0045] The learning model generation unit 150 sets the variable k to 1 (step S103).

[0046] 2, the learning model generation unit 150 provides the k-th (in this case, the first) divided data set and abnormal data (learning abnormal data) stored in the abnormal data storage unit 170 to the learning model 200, and causes the learning model 200 to learn (step S104). Note that when k=1, no abnormal data is stored in the abnormal data storage unit 170. Therefore, when k=1, the learning model 200 performs learning using only the first divided data set (corresponding to x1 in FIG. 2).

[0047] The learning performed when k≧2 can also be considered as relearning.

[0048] The learning model generation unit 150 increments the value of the variable k by 1 (step S105). If the value of the variable k reaches n (the number of divisions), the process ends (step S106).

[0049] When the value of the variable k is less than n, the abnormality determination unit 160 (k-1) Using the above, the next split data set (split data set x k ) is subjected to an abnormality determination (step S107). The abnormality determination unit 160 stores the determined abnormal data group in the abnormal data storage unit 170 as learning abnormal data (step S108). Then, the process returns to step S104.

[0050] 3, a trained learning model 200 is obtained. Specifically, when it is determined in step S106 that the value of the variable k has reached n (the number of divisions), the learning model 200 at that time becomes the final trained learning model.

[0051] In this embodiment, the case of generating a learning model used for detecting anomalies caused mainly by radio interference, failure, etc. is taken as an example. However, the learning model generation device 100 of this embodiment can generate a learning model used for detecting anomalies based on other factors, not limited to anomalies caused by radio interference, failure, etc. Examples of detecting anomalies based on other factors include detection of outliers in IoT (Internet of Things), detection of malware, and quality judgment of products.

[0052] In the estimation phase, anomaly detection is performed using the trained learning model 200. That is, features of input data that is the target of anomaly detection, i.e., feature data, are input to the learning model 200. The learning model 200 outputs an estimation result as to whether the feature data is normal data or abnormal data.

[0053] Fig. 4 is a block diagram showing a configuration example of an information processing device (computer) capable of realizing the functions of the learning model generating device 100 of the above embodiment. The information processing device shown in Fig. 4 includes one or more processors such as CPUs (Central Processing Units), a program memory 1002, and a memory 1003. Fig. 4 shows an example of an information processing device having one processor 1001.

[0054] The program memory 1002 is, for example, a non-transitory computer readable medium. The non-transitory computer readable medium includes various types of tangible storage media. For example, a semiconductor storage medium such as a flash ROM (Read Only Memory) or a magnetic storage medium such as a hard disk can be used as the program memory 1002. The program memory 1002 stores a learning model generation program for realizing the functions of each block (the feature detection unit 110, the data division unit 130, the learning model generation unit 150, and the anomaly determination unit 160) in the learning model generation device 100 of the above embodiment.

[0055] The processor 1001 executes processing according to a learning model generation program stored in the program memory 1002, thereby realizing the functions of the learning model generation device 100. When multiple processors are installed, the functions of the learning model generation device 100 can also be realized by the multiple processors working together.

[0056] For example, a RAM (Random Access Memory) can be used as the memory 1003. The memory 1003 stores temporary data generated when the learning model generation device 100 is executing a process. A configuration in which a learning model generation program is transferred to the memory 1003 and the processor 1001 executes processing based on an image processing program in the memory 1003 can also be assumed. Note that the program memory 1002 and the memory 1003 may be integrated.

[0057] The data storage unit 120, the divided data set storage unit 140, and the abnormal data storage unit 170 can be constructed in the memory 1003. The learning model 200 is constructed, for example, in the memory 1003. The learned learning model 200 can be transplanted to another information processing device. That is, a learning model generated in one computer can be used in another computer.

[0058] Fig. 5 is a block diagram showing the main parts of a learning model generating device. The learning model generating device 10 shown in Fig. 5 includes a data dividing means (data dividing unit) 11 (realized by the data dividing unit 130 in the embodiment) for dividing a plurality of feature data, each of which indicates a feature, into groups, and a learning model generating means (learning model generating unit) 12 (realized by the learning model generating unit 150 in the embodiment) for generating a learning model using, as learning data, feature data belonging to a first group (e.g., divided data set x1) among a plurality of groups (e.g., divided data sets x1 to x4) formed by the data dividing means 11, or feature data belonging to the first group and a part of feature data belonging to another group.

[0059] The learning model generation device 10 may include a feature extraction means (in an embodiment, realized by a feature extraction unit 110) that extracts features from each of the collected data, and the data division means 11 may be configured to group multiple feature data indicating the features extracted by the feature extraction means.

[0060] The learning model generation device 10 may include an abnormality judgment means (realized by the abnormality judgment unit 160 in the embodiment) that judges whether or not feature data belonging to other groups is abnormal data using the learning model generated by the learning model generation means 12, and the learning model generation means 12 may be configured to re-train the learning model by treating the feature data judged to be abnormal data by the abnormality judgment means as part of the feature data belonging to the other group.

[0061] Although a part or all of the above-described embodiments can be described as follows, the present invention is not limited to the following configurations.

[0062] (Appendix 1) A data division means for dividing a plurality of feature data into groups, each of which indicates a feature; a learning model generating means for generating a learning model using, as learning data, feature amount data belonging to a first group out of the plurality of groups formed by the data dividing means, or the feature amount data belonging to the first group and a part of the feature amount data belonging to another group; A learning model generation device comprising:

[0063] (Appendix 2) A feature extraction means for extracting features from each of the collected data, The data division means divides a plurality of feature data items into groups each indicating a feature extracted by the feature extraction means. 2. A learning model generation device as described in appendix 1.

[0064] (Additional Note 3) The method further includes: an abnormality determination unit that determines whether or not the feature amount data belonging to the other group is abnormal data by using the learning model generated by the learning model generation unit; The learning model generating means retrains the learning model using the feature amount data determined to be abnormal data by the abnormality determining means as part of the feature amount data belonging to the other group. 3. A learning model generation device as described in appendix 2.

[0065] (Note 4) The abnormality determination means performs a determination as to whether or not the feature data of each of the other groups is abnormal, The learning model generating means re-learns the learning model for all the other groups. 4. A learning model generation device as described in appendix 3.

[0066] (Note 5) The abnormality determination means stores the feature amount data determined to be abnormal data in an abnormal data storage unit, The learning model generating means uses the abnormal data stored in the abnormal data storage unit as part of the feature amount data belonging to the other group. A learning model generation device according to any one of Supplementary Note 1 to Supplementary Note 4.

[0067] (Appendix 6) Group multiple feature data, each of which indicates a feature, A learning model is generated using, as learning data, feature data belonging to a first group out of the formed groups, or the feature data belonging to the first group and a part of the feature data belonging to another group. Learning model generation method.

[0068] (Appendix 7) Extract features from each of the collected data, Grouping multiple feature data representing extracted features 7. A method for generating a learning model as described in appendix 6.

[0069] (Appendix 8) Using the generated learning model, it is determined whether the feature amount data belonging to the other group is abnormal data; The feature amount data determined to be abnormal data is used as part of the feature amount data belonging to the other group, and the learning model is retrained. 8. The learning model generation method according to claim 7.

[0070] (Appendix 9) To the computer, A process of grouping a plurality of feature data, each of which indicates a feature; generating a learning model using, as learning data, feature data belonging to a first group among the formed groups, or the feature data belonging to the first group and a part of feature data belonging to another group; A learning model generation program for executing the above.

[0071] (Appendix 10) To the computer, A process of extracting features from each of the collected data; A process of grouping multiple feature data indicating the extracted features; The learning model generation program according to claim 9, which executes the above.

[0072] (Appendix 11) To the computer, executing a process of determining whether or not the feature amount data belonging to the other group is abnormal data using the generated learning model; The feature amount data determined to be abnormal data is used as part of the feature amount data belonging to the other group, and the learning model is retrained. The learning model generation program according to claim 10. [Explanation of symbols]

[0073] 10,100 Learning model generation device 11 Data division methods 12. Learning model generation means 110 Feature detection unit 120 Data storage unit 130 Data division section 140 Split Data Set Storage Unit 150 Learning model generation unit 160 Abnormality judgment section 170 Abnormal data storage unit 200 Learning Models 1001 Processor 1002 Program Memory 1003 Memory

Claims

1. A data division means for dividing a plurality of feature data into groups, each of which indicates a feature; a learning model generating means for generating a learning model using, as learning data, feature amount data belonging to a first group out of the plurality of groups formed by the data dividing means, or the feature amount data belonging to the first group and a part of the feature amount data belonging to another group; A learning model generation device comprising:

2. A feature extraction means for extracting a feature from each of the collected data, The data division means divides a plurality of feature data items into groups each indicating a feature extracted by the feature extraction means. The learning model generating device according to claim 1.

3. an abnormality determination means for determining whether or not the feature amount data belonging to the other group is abnormal data by using the learning model generated by the learning model generation means, The learning model generating means retrains the learning model using the feature amount data determined to be abnormal data by the abnormality determining means as part of the feature amount data belonging to the other group. The learning model generating device according to claim 2.

4. the abnormality determination means executes a determination as to whether or not the feature amount data is abnormal for each of the other groups; The learning model generating means re-learns the learning model for all the other groups. The learning model generating device according to claim 3.

5. The abnormality determination means stores the feature amount data determined to be abnormal data in an abnormal data storage unit, The learning model generating means uses the abnormal data stored in the abnormal data storage unit as part of the feature amount data belonging to the other group. The learning model generating device according to any one of claims 1 to 4.

6. Grouping multiple feature data, each of which indicates a feature, A learning model is generated using, as learning data, feature amount data belonging to a first group out of the formed groups, or the feature amount data belonging to the first group and a part of the feature amount data belonging to another group. Learning model generation method.

7. Extract features from each of the collected data, Grouping multiple feature data representing extracted features The learning model generating method according to claim 6.

8. Using the generated learning model, it is determined whether the feature amount data belonging to the other group is abnormal data; The feature amount data determined to be abnormal data is used as part of the feature amount data belonging to the other group, and the learning model is retrained. The learning model generating method according to claim 7.

9. On the computer, A process of grouping a plurality of feature data, each of which indicates a feature; generating a learning model using, as learning data, feature data belonging to a first group among the formed groups, or the feature data belonging to the first group and a part of feature data belonging to another group; A learning model generation program for executing the above.

10. On the computer, A process of extracting features from each of the collected data; A process of grouping multiple feature data indicating the extracted features; 10. The learning model generation program according to claim 9, which executes the following:

Citation Information

Patent Citations

  • Learning assistance device

    JP2019159957A

  • Radio wave abnormality detection system, radio wave abnormality detection method, and radio wave abnormality detection program

    JP2022182844A