Anomaly detection device, anomaly detection method, and anomaly detection program
The anomaly detection device and method address the challenge of identifying causes in time series data by extracting features in both directions and calculating contributions, effectively determining the causes of anomalies in unsupervised scenarios.
Patent Information
- Application Number
- JP2023529217
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2021-06-21
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2041-06-21
AI Technical Summary
Conventional anomaly detection methods using machine learning fail to easily identify the cause of anomalies in time series data due to the lack of consideration for temporal relationships, especially in unsupervised scenarios.
An anomaly detection device and method that includes feature extraction in both the feature and time directions, calculating an anomaly score and contribution based on these features, and identifying causes by considering temporal relationships.
Facilitates easy identification of causes in unsupervised anomaly detection by accounting for time series nature, enabling accurate determination of features and time influences contributing to anomalies.
Smart Images

Figure 0007586322000005 
Figure 0007586322000006 
Figure 0007586322000007
Abstract
Description
[Technical field]
[0001] The present invention relates to an anomaly detection device, an anomaly detection method, and an anomaly detection program. [Background technology]
[0002] In anomaly detection using machine learning technology, when the frequency of anomalies occurring is extremely low, a model is created using normal data through unsupervised learning. An anomaly score that indicates deviation from the normal state is then calculated. A threshold is set for the calculated anomaly score to determine whether it is abnormal or normal. There are two types of machine learning anomaly detection: a method that can be applied regardless of whether it is time series data by treating each sample independently, and a method for time series data that sets a time window and considers the order of samples within that range (hereinafter referred to as "time series anomaly detection").
[0003] Here, the time window set in time series anomaly detection refers to a window that divides time series data into certain intervals. When creating a model, behavior is learned using data within the time window while shifting it in the time direction. In time series anomaly detection, the behavior of time series data under normal conditions is learned and an anomaly score is calculated using the prediction error, which is the difference between the predicted value and the actual measured value. Samples whose behavior is similar to the normal time series data learned during model learning have a small prediction error, while unlearned samples have a large prediction error; this property can be used to detect anomalies in time series data.
[0004] As described above, in anomaly detection using machine learning technology, it is determined that an anomaly has occurred based on whether or not the predicted anomaly score of a sample exceeds a preset anomaly judgment threshold, and the time when the anomaly occurred is identified. However, since the anomaly score only detects when an anomaly occurred and does not identify the feature that caused the anomaly, additional analysis is required, such as checking the behavior of the sample before and after the threshold was exceeded.
[0005] On the other hand, as an existing technology for identifying a feature that caused an anomaly, there is a technology that calculates the contribution degree using a trained model or the like. Here, the contribution degree indicates the degree of influence on the result output by the machine learning model, and it can be determined that the larger the contribution degree, the more likely it is to be the cause of the anomaly. There is also a technology that outputs the contribution degree taking into account the chronological relationship of the input data, and indicates which time of the input data contributed to the classification result (for example, see Non-Patent Document 1). [Prior art documents] [Non-patent literature]
[0006] [Non-Patent Document 1] R. Assaf, et al. "MTEX-CNN: Multivariate Time series EXplanations for Predictions with Convolutional Neural Networks", 2019 IEEE International Conference on Data Mining (ICDM), pp.952-957, [online], [Retrieved June 8, 2021], Internet<https: / / ieeexplore.ieee.org / stamp / stamp.jsp?tp=&arnumber=8970899&tag=1> Summary of the Invention [Problem to be solved by the invention]
[0007] However, in the above-mentioned conventional techniques, it has not been possible to easily identify causes taking time series into consideration in unsupervised anomaly detection, because the above-mentioned conventional techniques have the following problems.
[0008] First, in the technology that outputs the contribution degree using the above-mentioned trained model, each sample is treated independently in the time direction, so the contribution degree is output without considering the context of the data in the time direction. Also, the technology that outputs the contribution degree by considering the context of the input data in the time direction is a method for supervised learning and classification problems, so it cannot be applied as it is to anomaly detection technology without supervised data. [Means for solving the problem]
[0009] In order to solve the above-mentioned problems and achieve the object, the anomaly detection device of the present invention is characterized by comprising an acquisition unit that acquires time series data of a detection target in which an anomaly is detected at a predetermined time point, a first extraction unit that extracts from the time series data features in a feature direction in a time section prior to the predetermined time point, a second extraction unit that extracts features in a time direction in the time section from the features in the feature direction, and a calculation unit that calculates an anomaly score at the predetermined time point based on the features in the feature direction and the features in the time direction, and calculates the contribution of the feature direction and the contribution of the time direction to the anomaly score prior to the predetermined time point.
[0010] Moreover, an anomaly detection method according to the present invention is an anomaly detection method executed by an anomaly detection device, and includes: an acquisition step of acquiring time series data of a detection target in which an anomaly at a predetermined time point is detected; a first extraction step of extracting from the time series data features in a feature direction in a time section prior to the predetermined time point; a second extraction step of extracting from the feature direction features features in a time direction in the time section from the feature direction features; and a calculation step of calculating an anomaly score at the predetermined time point based on the feature direction features and the time direction features, and calculating the contribution of the feature direction and the time direction to the anomaly score prior to the predetermined time point.
[0011] Furthermore, the anomaly detection program of the present invention is characterized in that it causes a computer to execute the following steps: an acquisition step of acquiring time series data of a detection target in which an anomaly at a predetermined time point is detected; a first extraction step of extracting from the time series data features in a feature direction in a time section prior to the predetermined time point; a second extraction step of extracting from the feature direction features features in a time direction in the time section from the feature direction features; and a calculation step of calculating an anomaly score at the predetermined time point based on the feature direction features and the time direction features, and calculating the contribution of the feature direction and the time direction to the anomaly score prior to the predetermined time point. Effect of the Invention
[0012] The present invention facilitates cause identification taking time series into consideration in unsupervised anomaly detection. [Brief description of the drawings]
[0013] [Figure 1] FIG. 1 is a diagram illustrating an example of an anomaly detection system according to the first embodiment. [Diagram 2] FIG. 2 is a block diagram illustrating an example of the configuration of the anomaly detection device according to the first embodiment. [Diagram 3] FIG. 3 is a diagram illustrating an example of the architecture of a learning model according to the first embodiment. [Figure 4] FIG. 4 is a diagram illustrating an example of the feature extraction process according to the first embodiment. [Diagram 5] FIG. 5 is a diagram illustrating an example of learning data according to the first embodiment. [Figure 6] FIG. 6 is a diagram illustrating an example of evaluation data according to the first embodiment. [Figure 7] FIG. 7 is a diagram showing an example of data processing according to the first embodiment. [Figure 8] FIG. 8 is a diagram showing an example of data processing according to the first embodiment. [Figure 9] FIG. 9 is a diagram showing an example of data processing according to the first embodiment. [Figure 10] FIG. 10 is a diagram illustrating an example of the flow of the evaluation process of the anomaly detection accuracy according to the first embodiment. [Figure 11] FIG. 11 is a diagram illustrating an example of the evaluation process of the anomaly detection accuracy according to the first embodiment. [Figure 12] FIG. 12 is a diagram illustrating an example of the evaluation process of the anomaly detection accuracy according to the first embodiment. [Figure 13] FIG. 13 is a diagram illustrating an example of the flow of the contribution degree evaluation process according to the first embodiment. [Figure 14] FIG. 14 is a diagram illustrating an example of a calculation process of the contribution degree of a feature amount direction according to the first embodiment. [Figure 15] FIG. 15 is a diagram illustrating an example of a calculation process of the contribution degree in the time direction according to the first embodiment. [Figure 16] FIG. 16 is a diagram illustrating an example of the contribution degree evaluation process according to the first embodiment. [Figure 17] FIG. 17 is a diagram showing an example of an evaluation result of the effectiveness of the architecture of the learning model according to the first embodiment. [Figure 18] FIG. 18 is a diagram showing an example of an evaluation result of the effectiveness of the architecture of the learning model according to the first embodiment. [Figure 19] FIG. 19 is a diagram showing an example of an evaluation result of the effectiveness of the architecture of the learning model according to the first embodiment. [Figure 20] FIG. 20 is a diagram illustrating the evaluation results of the effectiveness of the architecture of the learning model according to the first embodiment. [Figure 21] FIG. 21 is a diagram showing an example of an evaluation result of the effectiveness of the architecture of the learning model according to the first embodiment. [Figure 22] FIG. 22 is a diagram showing an example of an evaluation result of the effectiveness of the architecture of the learning model according to the first embodiment. [Figure 23] FIG. 23 is a diagram showing an example of an evaluation result of the effectiveness of the architecture of the learning model according to the first embodiment. [Figure 24]FIG. 24 is a diagram showing an example of an evaluation result of the effectiveness of the architecture of the learning model according to the first embodiment. [Diagram 25] FIG. 25 is a diagram showing an example of an evaluation result of the effectiveness of the architecture of the learning model according to the first embodiment. [Figure 26] FIG. 26 is a flowchart showing an example of the overall flow of processing according to the first embodiment. [Figure 27] FIG. 27 is a diagram illustrating a computer that executes a program. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0014] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An anomaly detection device, an anomaly detection method, and an anomaly detection program according to the present invention will be described in detail below with reference to the accompanying drawings. Note that the present invention is not limited to the embodiments described below.
[0015] [First embodiment] Below, we will explain the processing of the anomaly detection system related to the first embodiment (or, as appropriate, the present embodiment), a comparison between the conventional technology and this embodiment, the configuration of the anomaly detection device 10, details of the processing, and the processing flow, and finally, the effects of this embodiment will be explained.
[0016] [Anomaly detection system processing] The processing of an anomaly detection system according to this embodiment (referred to as this system, where appropriate) will be described with reference to Fig. 1. Fig. 1 is a diagram showing an example of an anomaly detection system according to a first embodiment. This system has an anomaly detection device 10. Note that the anomaly detection system shown in Fig. 1 may include multiple anomaly detection devices 10.
[0017] Furthermore, in this system, time-series data 20 is involved as data acquired by the anomaly detection device 10. Here, the time-series data 20 is data that takes into account the order of each sample and is data that includes time-series information.
[0018] In the above-mentioned system, an example of anomaly detection processing based on a Convolutional Neural Network (hereinafter referred to as "CNN") is described below, which can calculate an anomaly score using a single time-series anomaly detection model, and can calculate the contribution to identifying the time and feature quantities that are likely to be the cause of the anomaly, regardless of the level of the anomaly score.
[0019] First, the anomaly detection device 10 acquires time-series data 20. At this time, it is desirable that the anomaly detection device 10 not only detects an anomaly from the anomaly score through processing, but also determines the feature amount and time that contributed to the anomaly score, taking into account the influence from the time before the anomaly score increased (see FIG. 1(1)).
[0020] Next, the anomaly detection device 10 identifies the cause of the anomaly by going back a specific time from the time the anomaly occurred (see FIG. 1(2)). In the example of FIG. 1, the device goes back w hours from the time t when the anomaly occurred, and identifies the cause of the anomaly at the time t when the anomaly occurred in the section from time tw to time t-1.
[0021] Then, the anomaly detection device 10 calculates the contribution degree of the feature direction based on the time-series data 20 (see FIG. 1(3)). In the example of FIG. 1, the calculated contribution degree shows that the feature amounts that influenced the anomaly score at time t are the feature amounts related to the sensors A and E.
[0022] In addition, the anomaly detection device 10 calculates the contribution degree in the time direction based on the time series data 20 (see FIG. 1(4)). In the example of FIG. 1, the calculated contribution degree shows that the time that influenced the anomaly score at time t was the latter half of the period from time tw to t-1.
[0023] Therefore, by processing the anomaly detection device 10, it becomes possible to grasp not only the feature quantity that is the cause of the anomaly, but also the temporal correlation (see FIG. 1 (5)). As described above, in this system, by calculating the contribution degree in the feature quantity direction and the temporal direction for the same model used for anomaly detection, going back a certain time (w hours in FIG. 1) from the time when the anomaly is to be identified, it becomes easy to identify the cause of time-series anomaly detection. In other words, in addition to time-series anomaly detection, this system can perform cause identification that takes time series into consideration.
[0024] [Conventional anomaly detection processing] Here, a technique related to a conventional anomaly detection process that is generally performed will be described as a reference technique.
[0025] Existing technologies for time series anomaly detection using machine learning include methods that use Recurrent Neural Networks (hereafter referred to as "RNN") and Long Short Time Memory (hereafter referred to as "LSTM"). RNNs are neural networks with an autoregressive structure, and enable predictions by incorporating hidden layers that retain information on past times in time series data. However, RNNs have the disadvantage that it is difficult to model long-term dependencies. LSTM improves on this disadvantage, and is a method that makes it possible to learn long-term dependencies by introducing a forget gate into the model.
[0026] In addition, existing technologies for identifying the features that cause an anomaly include an anomaly cause identification technology that uses reconstruction error, and a technology that calculates the contribution level using a trained model using LIME (Reference 1: Ross, AS, Hughes, MC & DoshiVelez, F. Right for the Right Reasons: TrainingDifferentiable Models by Constraining their Explanations. arXiv [cs.LG] (2017)), SHAP (Reference 2: Lundberg, S. & Lee, S.-I. A Unified Approach to Interpreting Model Predictions. NIPS2017 (2017)), Smooth Grad (Reference 3: Smilkov, D., Thorat, N., Kim, B., Viegas, F. & Wattenberg, M. SmoothGrad: removingnoise by adding noise. arXiv [cs.LG] (2017)), etc.
[0027] First, we will explain the technology for identifying the cause of anomalies using reconstruction errors. The reconstruction error is a value calculated for each feature from the difference between the input layer and output layer of a model that has an input layer, intermediate layer, and output layer. The reconstruction error can be calculated using a method that obtains a compressed representation of data in the intermediate layer, such as an autoencoder or principal component analysis. In the case of a sample that behaves similarly to a trained normal sample, the reconstruction error for each feature will be small as it will be correctly restored in the output layer, but in the case of a sample that behaves differently from the normal data, restoration in the output layer will not go well and the reconstruction error will be large. For this reason, visualization and statistical calculations are performed on the reconstruction error, and features with large values are estimated to be the cause of anomalies.
[0028] Next, we will explain the technology for calculating the contribution using a trained model. In the above-mentioned LIME and SHAP, the sample for which the cause is to be estimated is selected, and a new model for estimating the cause is created to output the contribution of each feature. On the other hand, there are also technologies that output the contribution by calculating the gradient of the input sample with respect to the output result, and a representative example is the above-mentioned Smooth Grad. Here, Smooth Grad creates multiple samples by intentionally adding Gaussian noise to the input sample, and by averaging these results, it is possible to output a contribution with less noise. These methods are mainly applied to classification models using supervised learning, but they can also be applied to unsupervised anomaly detection methods by labeling normal / abnormal based on the anomaly score and threshold.
[0029] However, both the above-mentioned technology for identifying the cause of anomalies using reconstruction errors and the technology for outputting the contribution from a trained model handle each sample independently in the time direction, and output a contribution that does not take into account the context of the data in the time direction. Therefore, these technologies are insufficient as a cause estimation technology for time series anomaly detection.
[0030] On the other hand, MTEX-CNN (see, for example, Non-Patent Document 1) is a technology that outputs the contribution degree taking into account the context of the data in the time direction. MTEX-CNN creates a sequence classification model using supervised learning, and outputs the contribution degree using Grad-CAM, which can present the basis for judgment using the value output by the last convolutional layer of CNN. MTEX-CNN can perform time series classification and output of the contribution degree using the same model, and can output the contribution degree in the feature amount direction, which indicates which feature amount of the input data separated by a time window contributed to the classification result, and the contribution degree in the time direction, which indicates which time of the input data contributed to the classification result.
[0031] However, since the above-mentioned MTEX-CNN is a method for supervised learning and classification problems, some ingenuity is required to apply it to unsupervised anomaly detection.
[0032] Below, we will explain the problems that cannot be solved by conventional technologies. Existing technologies for identifying the cause of anomaly detection handle each sample independently in the time direction, so they output a contribution rate that does not take into account the context of the selected sample. However, it is known that the behavior of time series data changes depending on the previous time (Reference 4: Brockwell, PJ, Davis, RA & Fienberg, SE Time Series: Theory and Methods: Theory and Methods. (Springer Science & Business Media, 1991)).
[0033] In the case of anomalies that progress and grow gradually, such as aging deterioration, the anomaly score also rises gradually, so this can be said to be an anomaly tendency specific to time series data. In order to estimate the cause of anomalies that tend to have a time lag between the occurrence of such anomalies and the rise in the anomaly score, it is desirable to present not only which feature is causing the rise in the anomaly score at the predicted time, but also what time the effect is from. This means outputting the contribution to time in addition to the contribution of the feature, which cannot be handled by existing anomaly cause identification technologies. Therefore, a technology is needed that can output the contribution taking into account the relationship in the time direction before the anomaly occurs for unsupervised time series anomaly detection.
[0034] [Configuration of anomaly detection device 10] Next, the configuration of the anomaly detection device 10 according to the present embodiment will be described in detail with reference to Fig. 2. Fig. 2 is a block diagram showing an example of the configuration of the anomaly detection device according to the first embodiment. The anomaly detection device 10 has an input unit 11, an output unit 12, a communication unit 13, a storage unit 14, and a control unit 15.
[0035] (1. Input section 11) The input unit 11 is responsible for inputting various types of information to the anomaly detection device 10. For example, the input unit 11 is realized by a mouse, a keyboard, or the like, and accepts input of setting information and the like to the anomaly detection device 10.
[0036] (2. Output section 12) The output unit 12 is responsible for outputting various types of information from the anomaly detection device 10. For example, the output unit 12 is realized by a display or the like, and outputs setting information stored in the anomaly detection device 10, and the like.
[0037] (3. Communications Department 13) The communication unit 13 is responsible for data communication with other devices. For example, the communication unit 13 performs data communication with each communication device. The communication unit 13 can also perform data communication with an operator's terminal (not shown).
[0038] (4.Storage unit 14) The storage unit 14 stores various information referenced when the control unit 15 operates and various information acquired when the control unit 15 operates. Here, the storage unit 14 can be realized by, for example, a semiconductor memory element such as a random access memory (RAM) or a flash memory, or a storage device such as a hard disk or an optical disk. Note that, in the example of Fig. 2, the storage unit 14 is installed inside the anomaly detection device 10, but it may be installed outside the anomaly detection device 10, or multiple storage units may be installed.
[0039] (5. Control Unit 15) The control unit 15 controls the entire anomaly detection device 10. The control unit 15 includes an acquisition unit 15a, a first extraction unit 15b, a second extraction unit 15c, a calculation unit 15d, and an identification unit 15e. Here, the control unit 15 is, for example, an electronic circuit such as a CPU (Central Processing Unit) or an MPU (Micro Processing Unit), or an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array).
[0040] (5-1. Acquisition part 15a) The acquiring unit 15a acquires time series data of a detection target in which an abnormality is detected at a predetermined time. For example, the acquiring unit 15a acquires data including sensor values transmitted from a plurality of sensors for each time. Meanwhile, the acquiring unit 15a outputs the acquired time series data to the first extracting unit 15b. The acquiring unit 15a may also store the acquired time series data in the storage unit 14.
[0041] (5-2. First extraction part 15b) The first extraction unit 15b extracts features in the feature direction in a time section before a predetermined time point from the time series data. For example, the first extraction unit 15b performs two-dimensional convolution on each feature of the time series data to extract features in the feature direction. In addition, the first extraction unit 15b outputs a first feature map (feature map 1) as features in the feature direction.
[0042] To explain the details of the process, for example, the first extraction unit 15b performs two-dimensional convolution twice on each feature of the time series data of time window w and d dimension, and compresses the feature map to (w / 4)×d dimension. The first extraction unit 15b sets the number of filters for the first time to 64 and the number of filters for the second time to 128, and performs feature extraction in the feature direction by performing convolution. The feature extraction process in the feature direction by the first extraction unit 15b will be described later in [Details of Processing] (2. Feature Extraction Process).
[0043] (5-3.Second extraction part 15c) The second extraction unit 15c extracts time-direction features in a predetermined time section from the features in the feature direction. For example, the second extraction unit 15c performs one-dimensional convolution on each feature of the features in the feature direction to extract features in the time direction. In addition, the second extraction unit 15c outputs a second feature map (feature map 2) as the features in the time direction.
[0044] To explain the details of the process, for example, the second extraction unit 15c performs one-dimensional convolution on the d-dimensional first feature map so as to use all d-dimensional features, thereby extracting features in the time direction of the entire input data. The feature extraction process in the time direction by the second extraction unit 15c will be described later in [Details of the process] (2. Feature extraction process).
[0045] (5-4. Calculation unit 15d) The calculation unit 15d calculates an anomaly score at a predetermined time point based on the feature in the feature direction and the feature in the time direction, and calculates the contribution degree of the feature direction and the contribution degree of the time direction to the anomaly score before the predetermined time point. For example, the calculation unit 15d calculates the anomaly score at a predetermined time point using an unsupervised learning model, and calculates the contribution degree of the feature direction and the contribution degree of the time direction.
[0046] Furthermore, the calculation unit 15d calculates the anomaly score, as well as the contribution degree in the feature direction and the contribution degree in the time direction, using an unsupervised learning model that has been trained using a loss function consisting of a penalty for at least one of the prediction error regarding the anomaly score, the contribution degree in the feature direction, and the contribution degree in the time direction.
[0047] To explain the details of the process, the calculation unit 15d performs backpropagation using the predicted value for the final layer where the convolution in the feature direction has been performed, and calculates weights from the obtained gradient values. Then, the calculation unit 15d outputs the contribution degree in the feature direction by using an activation function for a matrix obtained by multiplying the obtained weights and the first feature map. The calculation unit 15d also performs backpropagation using the predicted value for the final layer where the convolution in the time direction has been performed, and calculates weights from the obtained gradient values. Then, the calculation unit 15d outputs the contribution degree in the time direction by using an activation function for a matrix obtained by multiplying the obtained weights and the second feature map. The contribution degree calculation process by the calculation unit 15d will be described later in [Details of the Process] (3. Contribution Degree Calculation Process).
[0048] (5-5. Specific part 15e) When the identification unit 15e detects an abnormality based on the abnormality score, the identification unit 15e identifies the cause of the abnormality using the contribution degree in the feature amount direction or the contribution degree in the time direction. For example, the identification unit 15e identifies the type of sensor as a feature that influenced the abnormality score at the abnormality occurrence time using the contribution degree in the feature amount direction. Furthermore, the identification unit 15e may store the identified information in the storage unit 14.
[0049] [Processing details] The details of the processing according to this embodiment will be described using Fig. 3 to Fig. 25, formulas, etc. Below, an overview of the architecture of the learning model, feature extraction processing, contribution degree calculation processing, loss function, and learning model evaluation processing will be described in that order.
[0050] (1. Overview of the learning model architecture) An overview of the architecture of the learning model according to this embodiment (hereinafter, referred to as the architecture as appropriate) will be described with reference to Fig. 3. Fig. 3 is a diagram showing an example of the architecture of the learning model according to the first embodiment. In the following, the architecture of the learning model that outputs the anomaly score, the contribution degree in the feature amount direction, and the contribution degree in the time direction from the same model will be described.
[0051] In this architecture, time series anomaly detection is performed by using CNN to create a model that predicts the actual measured value at a certain time point using input data with d-dimensional features and w time window. The certain time point can be k time points before or after. In addition, in this architecture, two-stage feature extraction is performed on the input data using CNN. That is, in this architecture, the first stage is feature extraction in the feature direction (see Figure 3(1)), and the second stage is feature extraction in the time direction (see Figure 3(2)). After that, in this architecture, a fully connected layer is obtained (see Figure 3(3)), the predicted value y^ is output (see Figure 3(4)), and the error (mean square error, etc.) of the actual measured value y is calculated to calculate the anomaly score (see Figure 3(5)).
[0052] (2. Feature Extraction Processing) Using FIG. 4, the details of the feature extraction process will be described. FIG. 4 is a diagram showing an example of the feature extraction process according to the first embodiment. Hereinafter, the feature extraction process in the feature quantity direction and the feature extraction process in the time direction will be described in this order.
[0053] (2-1. Feature extraction process in the feature quantity direction) In the first-stage feature extraction process in the feature quantity direction (first extraction process), first, the anomaly detection device 10 performs two-dimensional convolution a plurality of times for each feature quantity (see FIG. 4(1)). Next, after performing the last two-dimensional convolution in the feature extraction in the feature quantity direction, the anomaly detection device 10 transposes the matrix to obtain a feature quantity map 1 of size c×d (see FIG. 4(2)).
[0054] Note that c must be a value smaller than the time window w. Also, the filter size w' used for convolution needs to be w'×1, and there is a restriction that 1 < w' < w. For example, the anomaly detection device 10 performs two-dimensional convolution twice, sets w' = 4 and c = w / 4, and compresses the feature quantity map to (w / 4)×d dimensions. Also, the number of filters used for convolution can be set to any value. For example, the anomaly detection device 10 sets the number of filters for the first time to 64 and the number of filters for the second time to 128, and performs convolution. Also, the anomaly detection device 10 may use half-padding for the convolution in the feature quantity direction.
[0055] (2-2. Feature extraction process in the time direction) In the second-stage feature extraction process in the time direction (second extraction process), the anomaly detection device 10 performs one-dimensional convolution on the feature quantity map 1 obtained in the first stage so as to utilize all d-dimensional feature quantities, thereby performing feature extraction in the time direction of the entire input data (see FIG. 4(3)) and obtaining a feature quantity map 2 (see FIG. 4(4)).
[0056] Note that the filter size used in this convolution must be c’×d, with the restriction that 1 < c’ < c. Also, the parameter n is a value determined depending on the filter c’, and n = c - c’ + 1. For example, the anomaly detection device 10 sets c’ = 4 and performs convolution. Also, the anomaly detection device 10 may use half-padding for convolution in the time direction.
[0057] Then, after performing the above-described first extraction process and second extraction process, the anomaly detection device 10 obtains a fully connected layer (see FIG. 4(5)) and outputs a predicted value ŷ (see FIG. 4(6)).
[0058] (3. Contribution Degree Calculation Process) As a process following the feature extraction process, the details of the contribution degree calculation process will be described. Below, the outline of the contribution degree calculation process, the contribution degree calculation process in the feature quantity direction, and the contribution degree calculation process in the time direction will be described in order.
[0059] (3-1. Outline of the Contribution Degree Calculation Process) First, the anomaly detection device 10 outputs a gradient value by backpropagating the value output from the learning model to the selected convolutional layer using the output value of the learning model, and calculates the Global Average Pooling of the gradient value to output a weight. Then, the anomaly detection device 10 calculates the contribution degree by converting the matrix obtained by multiplying the feature map obtained from the selected convolutional layer and the obtained weight using an activation function (such as the ReLU function).
[0060] That is, the anomaly detection device 10 performs a contribution degree calculation process using the predicted value ŷ at k points ahead (k is an arbitrary variable) for each of the feature map 1 (see FIG. 4(2)), which is the output of the layer that performs feature extraction in the feature quantity direction, and the feature map 2 (see FIG. 4(4)), which is the output of the layer that performs feature extraction in the time direction, and outputs the contribution degree in the feature quantity direction, the contribution degree in the time direction, and the contribution degree to the predicted value at k points ahead.
[0061] (3-2. Contribution Degree Calculation Process in the Feature Quantity Direction) The anomaly detection device 10 calculates the predicted value y^ for the final layer in which the feature-wise convolution is performed. l Then, the anomaly detection device 10 performs backpropagation using c, and calculates weights by dividing the gradient values obtained here by c. The anomaly detection device 10 then outputs the contributions of the resulting matrix obtained by multiplying the weights obtained by the feature map 1 by an activation function.
[0062] In addition, since the contribution degree to the feature map 1 is c×d dimensional and the dimension does not match the input data, and the output contribution degree cannot be interpreted, the anomaly detection device 10 changes the size to w×d dimensionality, the same as the size of the input data, and outputs the contribution degree in the feature direction that can be interpreted. For example, the anomaly detection device 10 changes the size from (w / 4)×d dimensionality to w×d dimensionality and outputs the contribution degree in the feature direction.
[0063] (3-3. Calculation of contribution in time direction) The anomaly detection device 10 calculates the predicted value y^ for the final layer in which convolution in the time direction is performed. l Then, the anomaly detection device 10 performs backpropagation using and calculates weights by dividing the gradient value obtained here by n. Then, the anomaly detection device 10 outputs the contribution degree by using an activation function for a matrix obtained by multiplying the obtained weights and the feature amount map 2.
[0064] In addition, since the contribution degree to the feature map 2 is n×m dimensional and does not match the size of the time window w of the input data, the anomaly detection device 10 outputs the contribution degree in the time direction by changing the size to w×1 dimensionality.
[0065] (4. Loss Function) The loss function for learning the learning model according to this embodiment will be described in detail below. First, the loss function Loss is expressed as the following formula (1).
[0066]
number
[0067]
number
[0068]
number
[0069]
number
[0070] Here, ||y in the above formula (2) i -y^ l || represents the distance between two vectors, and is calculated using Euclidean distance or mean square error. In addition, A in the above formula (3) represents the feature contribution matrix, and B in the above formula (4) represents the time contribution matrix. In addition, the loss function Loss in the above formula (1) is L, which represents the penalty for the prediction error. ad , which represents the penalty for the contribution of the feature feature , and L which represents the penalty for the time contribution. time It consists of:
[0071] In the loss function Loss in the above formula (1), the penalty for the contribution (L feature , L time ) adds regularization during learning so that the contribution approaches 0. This regularization is expected to have the effect of reducing the contribution to normal samples and increasing the contribution to abnormal samples.
[0072] Note that the loss function Loss in equation (1) does not necessarily need to include a penalty for the contribution rate, but rather the prediction error penalty L ad Only one or both contribution penalties (L feature , L time) may be included. In addition, the above formulas (3) and (4) for the penalty on the contribution degree are not limited to these, as long as they are regularizations that produce similar effects. In the following, a method that uses a loss function of only the prediction error is referred to as a "method without regularization," and a method that performs regularization so that a large contribution degree is output for data that does not exist at the time of learning in addition to the prediction error is referred to as a "method with regularization."
[0073] (5. Evaluation process of learning model) The evaluation process of the learning model according to this embodiment will be described in detail with reference to Figs. 5 to 25. Below, the overview of the evaluation process of the learning model, the evaluation process using a method without regularization, the evaluation process using a method with regularization, and the effectiveness of the learning model will be described in that order. Note that the evaluation process of the learning model according to this embodiment is not limited to the process described below.
[0074] (5-1. Overview of the learning model evaluation process) An overview of the evaluation process of the learning model according to this embodiment will be described with reference to Fig. 5 to Fig. 16. The following describes in order the creation of data to be used in the learning model, data processing, evaluation of the anomaly detection accuracy, and evaluation of the contribution degree.
[0075] (5-1-1. Creating data) Creation of data used in a learning model will be described with reference to Fig. 5 and Fig. 6. Fig. 5 is a diagram showing an example of learning data according to the first embodiment. Fig. 6 is a diagram showing an example of evaluation data according to the first embodiment.
[0076] We create artificial data with five-dimensional features to be used in the learning model. The differences between the learning data and the evaluation data are as follows:
[0077] For the first to fourth dimensions (normal dimensions), both the training data and the evaluation data are generated according to the same rules, and there is no difference. That is, as shown in Figure 5, the training data shows a waveform with no significant fluctuations over the entire range. Also, as shown in Figure 6 (2), the normal dimensions of the evaluation data show the same waveform as the training data.
[0078] For the fifth dimension (anomalous dimension), the training data is generated by combining trigonometric functions and uniform distribution. Meanwhile, the evaluation data is generated according to the same rules as the training data, and an abnormal state is simulated by periodically adding a significantly large value. In other words, the data is generated so that abnormal waveforms appear periodically, as shown in Figure 6(1). Note that in Figure 6(1), all shaded rectangular areas are treated as abnormal sections.
[0079] (5-1-2. Data processing) Processing of data used in a learning model will be described with reference to Fig. 7 to Fig. 9. Fig. 7 to Fig. 9 are diagrams showing an example of data processing according to the first embodiment.
[0080] First, as shown in Fig. 7, the time series data is extracted using a time window, converted into a data format that can be input to the model, and labeled. In Fig. 7, the time window w=20, but is not particularly limited.
[0081] Next, if the fifth dimension of the data extracted using the time window contains an abnormal value at even one time point, an abnormal label is assigned. Here, if no abnormal values are included at even one time point, as in Figure 8(1), a normal label is assigned. If abnormal values are included at several time points, as in Figure 8(2), an abnormal label is assigned. If abnormal values are included at all time points, as in Figure 8(3), this is obviously abnormal, so an abnormal label is assigned.
[0082] In addition, in FIG. 9, 350 abnormal data items are included among 1079 evaluation data items, and it is found that the generated artificial data includes approximately 32% abnormal data items.
[0083] (5-1-3. Evaluation of anomaly detection accuracy) Evaluation of anomaly detection accuracy will be described with reference to Fig. 10 to Fig. 12. Fig. 10 is a diagram showing an example of the flow of evaluation processing of anomaly detection accuracy according to the first embodiment. Fig. 11 and Fig. 12 are diagrams showing an example of evaluation processing of anomaly detection accuracy according to the first embodiment.
[0084] Using FIG. 10, we will explain the flow of calculating and evaluating anomaly detection accuracy by comparing the anomaly labels and normal labels of the evaluation data with the anomaly judgment results. In the learning process, first, normal learning data is input and the learning model is trained (see FIG. 10(1)). Next, the learning model calculates an anomaly score and outputs an anomaly score within the normal range (see FIG. 10(2)). Then, a threshold is determined using the output anomaly score. The determination of the threshold will be described later using FIG. 11.
[0085] Meanwhile, in the evaluation process, first, the evaluation data is input into the learning model and a prediction is made (see FIG. 10(4)). Next, the learning model calculates an anomaly score and outputs the anomaly score (see FIG. 10(5)). Then, the output anomaly score is compared with the determined threshold (see FIG. 10(6)), and a judgment of anomaly or normality is made and a judgment result is output (see FIG. 10(7)). Finally, a true / false judgment is made based on the label of the evaluation data and the judgment result, and the accuracy of anomaly detection is evaluated (see FIG. 10(8)). Note that the evaluation indices used are precision, recall, F1 score, and ROC-AUC (Receiver Operating Characteristic-Area Under the Curve), and the average value of five trials is calculated to calculate the numerical value.
[0086] 11 and 12, the determination of the threshold and the determination of anomaly based on the threshold are described. First, the threshold used for anomaly determination is calculated for all learning data, and the 95%tile value is set as the threshold (see FIG. 11). On the other hand, if the determined threshold is exceeded, the evaluation data is determined to be abnormal (see FIG. 12).
[0087] (5-1-4. Evaluation of contribution) The evaluation of the contribution degree will be described with reference to Fig. 13 to Fig. 16. Fig. 13 is a diagram showing an example of the flow of the evaluation process of the contribution degree according to the first embodiment. Figs. 14 to 16 are diagrams showing an example of the evaluation process of the contribution degree according to the first embodiment.
[0088] The flow of the contribution evaluation will be described with reference to FIG. 13. First, the evaluation data is input to the learning model and a prediction is performed (see FIG. 13(1)). Next, the learning model calculates the contribution and outputs the contribution in the feature direction and the time direction (see FIG. 13(2)). Then, the maximum value of the output contribution is calculated and a histogram is drawn (see FIG. 13(3)). Finally, the contribution is evaluated based on the label of the evaluation data and the drawn histogram (see FIG. 13(4)).
[0089] The process of calculating the contribution degree of the feature direction and then calculating the maximum value thereof will be described with reference to Fig. 14. In Fig. 14, the numerical values calculated as the contribution degree of the feature direction are shown in a table format. In Fig. 14, the maximum value "7.6" is output when drawing a histogram of the contribution degree of the feature direction.
[0090] A process of calculating the contribution degree in the time direction and then calculating the maximum value thereof will be described with reference to Fig. 15. In Fig. 15, the values calculated as the contribution degree in the time direction are shown in a table format. In Fig. 15, the maximum value "6.8" is output when drawing a histogram of the contribution degree in the time direction.
[0091] The process of creating histograms of maximum values for normal and abnormal labels in the feature direction and time direction and comparing their shapes will be described below with reference to Figure 16. Note that although the contribution degree in the time direction will be described below, the contribution degree in the feature direction is processed in the same way.
[0092] First, a histogram of maximum values of anomaly labels is drawn from the contributions in the time direction of multiple anomaly labels (see Fig. 16 (1) and (2)). Here, it is desirable that the maximum value histogram of anomaly labels drawn has a heavy-tailed distribution of contributions. In other words, in the case of an anomaly, a high contribution to the cause of the anomaly should be obtained.
[0093] On the other hand, a maximum value histogram of normal labels is drawn from the contribution degrees in the time direction of multiple signals that have been assigned normal labels (see Fig. 16 (3) and (4)). Here, it is desirable for the contribution degree of the maximum value histogram of normal labels to be drawn to be 0. In other words, in the case of normality, the contribution degree should be low since it is not the cause of an anomaly.
[0094] Then, the maximum value histogram of the abnormal label is compared with the maximum value histogram of the normal label to evaluate whether the contribution is appropriately reflected (see FIG. 16 (2) and (3)).
[0095] (5-2. Evaluation process using the method without regularization) The evaluation process using the method without regularization of the learning model according to the present embodiment will be described with reference to Figs. 17 to 20. Figs. 17 to 19 are diagrams showing an example of the evaluation result of the effectiveness of the architecture of the learning model according to the first embodiment. Fig. 20 is a diagram explaining the evaluation result of the effectiveness of the architecture of the learning model according to the first embodiment. Below, the evaluation result of the anomaly detection accuracy and the evaluation result of the contribution degree will be described in that order.
[0096] (5-2-1. Evaluation results of anomaly detection accuracy) First, we will explain the evaluation results of the effectiveness of the anomaly detection accuracy of the learning model architecture. Below, we will explain the overview of the learning model architecture using the no-regularization method, and then explain the evaluation results of the effectiveness.
[0097] First, the architecture of the learning model without regularization uses the mean squared error (L) as the loss function Loss. ad (see Equation 2) is used. That is, Loss = L ad Anomaly detection processing is performed according to the above. Evaluation of anomaly detection accuracy is performed based on the above (5-1-3. Evaluation of anomaly detection accuracy). At this time, as a standard for effectiveness, if the AUC is 0.8 or more, it is judged to be effective.
[0098] As shown in FIG. 17, in the learning model architecture using the above-mentioned method without regularization, the AUC was 0.885 (average of five trials), indicating that the anomaly detection accuracy is effective and that it can be fully utilized for anomaly detection.
[0099] (5-2-2. Contribution evaluation results) Next, we will explain the evaluation results of the effectiveness of the contribution of the learning model architecture. Below, we will explain the evaluation of the maximum value histogram of the abnormal label and the evaluation of the maximum value histogram of the normal label based on the contribution in the time direction.
[0100] First, the evaluation of the maximum value histogram of anomaly labels will be explained using Fig. 18. First, in the maximum value histogram of anomaly labels, when the maximum value of the contribution degree is 0, it is shown that the cause of the anomaly may not be identified well even with abnormal data (see Fig. 18(1)). On the other hand, when the maximum value of the contribution degree is greater than 0, it is shown that the time and feature quantity that are thought to be abnormal are successfully captured (see Fig. 18(2)). Therefore, in Fig. 18, it cannot be determined that a high contribution degree is being output to the anomaly, and it cannot be said that the cause of the anomaly has been effectively isolated.
[0101] Next, the evaluation of the maximum value histogram of the normal label will be described with reference to FIG. 19. First, in the maximum value histogram of the normal label, when the maximum value of the contribution degree is 0, there is no abnormality cause in the normal data, so it is preferable that the maximum values of the contribution degrees are all 0 (see FIG. 19(1)). That is, as shown in FIG. 20, a histogram in which the maximum values of the contribution degrees are all 0 can be said to be the ideal form. On the other hand, when the maximum value of the contribution degree is greater than 0, it indicates that it is determined that there is a portion that is thought to be the cause of an abnormality even though the data is normal (see FIG. 19(2)). Therefore, in FIG. 19, it cannot be determined that a low contribution degree is being output relative to normal, and it cannot be said that the cause of the abnormality has been effectively isolated.
[0102] (5-3. Evaluation process using regularized method) The evaluation process of the learning model according to the present embodiment using the regularized method will be described with reference to Fig. 21 to Fig. 25. Fig. 21 to Fig. 25 are diagrams showing an example of the evaluation result of the effectiveness of the architecture of the learning model according to the first embodiment. Below, the evaluation result of the anomaly detection accuracy and the evaluation result of the contribution degree will be described in that order.
[0103] (5-3-1. Evaluation results of anomaly detection accuracy) First, we will explain the evaluation results of the effectiveness of the anomaly detection accuracy of the learning model architecture. Below, we will explain the overview of the learning model architecture using the regularization method, and then explain the evaluation results of the effectiveness.
[0104] First, the architecture of the learning model using the regularized method uses L ad (See formula 2) plus L feature (see formula 3) and L time (See Equation 4). That is, Loss = L ad +L feature +L time Anomaly detection processing is performed according to (see equation 1). Furthermore, the evaluation of anomaly detection accuracy is performed based on the above (5-1-3. Evaluation of anomaly detection accuracy). At this time, since regularization is an operation that makes optimization difficult, it is sufficient to confirm that the anomaly detection accuracy does not deteriorate.
[0105] As shown in Figure 21, in the learning model architecture using the above regularized method, the AUC is 0.948 (average of 5 trials), which is higher than the AUC of 0.885 based on the method without regularization, and it is evaluated that there is no adverse effect due to regularization.
[0106] (5-3-2. Contribution evaluation results) Next, we will explain the evaluation results of the effectiveness of the contribution of the learning model architecture. Below, we will explain the evaluation of the maximum value histogram of the normal label and the evaluation of the maximum value histogram of the abnormal label based on the contribution in the time direction.
[0107] First, the evaluation of the maximum value histogram of normal labels will be described with reference to Fig. 22 and Fig. 23. Fig. 22 is a maximum value histogram of normal labels drawn using a method without regularization, in which the maximum value of the contribution degree is a value other than 0. On the other hand, Fig. 23 is a maximum value histogram of normal labels drawn using a method with regularization, in which the maximum value of the contribution degree is almost 0. Therefore, it can be seen that regularization increases the proportion of maximum values of the contribution degree that are 0 in the maximum value histogram of normal labels, making it possible to avoid the output of abnormal causes that would cause confusion.
[0108] Next, the evaluation of the maximum value histogram of the abnormal label will be described with reference to Figs. 24 and 25. Fig. 24 is a maximum value histogram of the abnormal label drawn by the method without regularization, and the maximum value of the contribution degree falls within the range of 0 to 10. On the other hand, Fig. 25 is a maximum value histogram of the abnormal label drawn by the method with regularization, and since it has become more over-reactive to abnormalities, the maximum value of the contribution degree takes a large value in the range of 0 to 100. Therefore, it can be seen that regularization has made the maximum value of the contribution degree larger in the maximum value histogram of the abnormal label compared to the method without regularization, and therefore the cause of the abnormality is emphasized, that is, it is easier to identify the cause of the abnormality.
[0109] (5-4. Effectiveness of the learning model) From the above, it can be determined that the architecture of the learning model according to this embodiment has a performance that can be used for anomaly detection. In addition, by regularizing the loss function used for learning the learning model according to this embodiment, it becomes easier to identify the cause of anomalies.
[0110] [Processing flow] The flow of processing according to this embodiment will be described in detail with reference to Fig. 26. Fig. 26 is a flowchart showing an example of the overall flow of processing according to the first embodiment. Below, the flow of the entire anomaly detection processing will be shown, and an overview of each step will be described.
[0111] (Overall process flow) First, the acquisition unit 15a of the anomaly detection device 10 executes a time-series data acquisition process (step S101). Next, the first extraction unit 15b of the anomaly detection device 10 executes a feature extraction process in the feature amount direction (first extraction process) (step S102). Also, the second extraction unit 15c of the anomaly detection device 10 executes a feature extraction process in the time direction (second extraction process) (step S103). Next, the calculation unit 15d of the anomaly detection device 10 executes a contribution degree calculation process (step S104). Finally, the identification unit 15e of the anomaly detection device 10 executes an anomaly cause identification process (step S105) and ends the process. Note that the above steps S101 to S105 can be executed in a different order. Also, some of the above steps S101 to S105 may be omitted.
[0112] (Flow of each process) First, a description will be given of a time-series data acquisition process performed by the acquisition unit 15a. In this process, the acquisition unit 15a acquires time-series data of a detection target for detecting an abnormality.
[0113] Secondly, the feature extraction process in the feature direction by the first extraction unit 15b will be described. In this process, the first extraction unit 15b first performs two-dimensional convolution for each feature multiple times, and outputs the feature map 1 by transposing the matrix after performing the final two-dimensional convolution in the feature extraction in the feature direction.
[0114] Thirdly, the feature extraction process in the time direction by the second extraction unit 15c will be described. In this process, the second extraction unit 15c performs one-dimensional convolution on the feature amount map 1 output in the process of step S102 so as to use all the feature amounts, thereby extracting features in the time direction of the entire input data, and outputs the feature amount map 2.
[0115] Fourthly, the contribution degree calculation process by the calculation unit 15d will be described. In this process, the calculation unit 15d outputs a gradient value by backpropagating the value output from the learning model to the convolution layer selected using the output value of the learning model, and then outputs the weight. Then, the calculation unit 15d calculates the contribution degree by converting the matrix obtained by multiplying the feature map output in the processes of steps S102 and S103 by the obtained weight using an activation function. At this time, the calculation unit 15d outputs the contribution degree in the feature amount direction and the contribution degree in the time direction, respectively.
[0116] Fifthly, an anomaly cause identification process by the identification unit 15e will be described. In this process, the identification unit 15e identifies a time and a feature that is considered to be the cause of the anomaly, based on the contribution degree in the feature direction and the contribution degree in the time direction output in the process of step S104.
[0117] [Advantages of the first embodiment] First, in the anomaly detection process according to the present embodiment described above, time series data of a detection target in which an anomaly at a predetermined time point is detected is acquired, features in the feature amount direction in a time section before the predetermined time point are extracted from the time series data, features in the time direction are extracted from the features in the feature amount direction, and an anomaly score at the predetermined time point is calculated based on the features in the feature amount direction and the features in the time direction, and the contribution of the feature amount direction and the contribution of the time direction before the predetermined time point to the anomaly score are calculated. Therefore, this process makes it easy to identify the cause taking time series into consideration in unsupervised anomaly detection.
[0118] Secondly, in the anomaly detection process according to the present embodiment described above, an unsupervised learning model is used to calculate an anomaly score at a predetermined time point, and the contribution degree in the feature amount direction and the contribution degree in the time direction are calculated, and when an anomaly is detected based on the anomaly score, the cause of the anomaly is identified using the contribution degree in the feature amount direction or the contribution degree in the time direction. Therefore, in unsupervised anomaly detection, this process makes it easy to identify the cause taking into account time series, and can identify the influence of the causative feature or time.
[0119] Thirdly, in the anomaly detection process according to the present embodiment described above, two-dimensional convolution is performed on each feature of time series data to extract features in the feature direction, and one-dimensional convolution is performed on each feature of the feature direction to extract features in the time direction. Therefore, in this process, in unsupervised anomaly detection, cause identification taking time series into consideration can be easily and efficiently performed.
[0120] Fourth, in the anomaly detection process according to the present embodiment described above, an anomaly score is calculated using an unsupervised learning model that has been trained using a loss function consisting of a penalty for at least one of the prediction error for the anomaly score, the contribution degree in the feature amount direction, and the contribution degree in the time direction, and the contribution degree in the feature amount direction and the contribution degree in the time direction are calculated. Therefore, in this process, in unsupervised anomaly detection, cause identification taking time series into consideration can be performed easily and accurately.
[0121] [System configuration, etc.] Each component of each device shown in the figures according to the above embodiment is a functional concept, and does not necessarily have to be physically configured as shown in the figures. In other words, the specific form of distribution and integration of each device is not limited to that shown in the figures, and all or a part of them can be functionally or physically distributed and integrated in any unit according to various loads, usage conditions, etc. Furthermore, each processing function performed by each device can be realized in whole or in any part by a CPU and a program analyzed and executed by the CPU, or can be realized as hardware using wired logic.
[0122] Furthermore, among the processes described in the above embodiments, all or part of the processes described as being performed automatically can be performed manually, or all or part of the processes described as being performed manually can be performed automatically by a known method. In addition, the information including the processing procedures, control procedures, specific names, various data and parameters shown in the above documents and drawings can be changed arbitrarily unless otherwise specified.
[0123] 〔program〕 It is also possible to create a program in which the processing executed by the anomaly detection device 10 described in the above embodiment is written in a language executable by a computer. In this case, the same effects as those of the above embodiment can be obtained by having the computer execute the program. Furthermore, such a program may be recorded on a computer-readable recording medium, and the program recorded on the recording medium may be read and executed by a computer to realize the same processing as that of the above embodiment.
[0124] Fig. 27 is a diagram showing a computer that executes a program. As shown in Fig. 27, a computer 1000 includes, for example, a memory 1010, a CPU 1020, a hard disk drive interface 1030, a disk drive interface 1040, a serial port interface 1050, a video adapter 1060, and a network interface 1070, and these components are connected by a bus 1080.
[0125] The memory 1010 includes a ROM (Read Only Memory) 1011 and a RAM 1012, as exemplified in FIG. 27. The ROM 1011 stores a boot program such as a BIOS (Basic Input Output System). The hard disk drive interface 1030 is connected to a hard disk drive 1090, as exemplified in FIG. 27. The disk drive interface 1040 is connected to a disk drive 1100, as exemplified in FIG. 27. For example, a removable storage medium such as a magnetic disk or an optical disk is inserted into the disk drive 1100. The serial port interface 1050 is connected to a mouse 1110 and a keyboard 1120, as exemplified in FIG. 27. The video adapter 1060 is connected to a display 1130, as exemplified in FIG. 27.
[0126] 27, the hard disk drive 1090 stores, for example, an OS 1091, an application program 1092, a program module 1093, and program data 1094. That is, the above programs are stored in, for example, the hard disk drive 1090 as program modules in which instructions to be executed by the computer 1000 are written.
[0127] Furthermore, the various data described in the above embodiment are stored as program data, for example, in the memory 1010 or the hard disk drive 1090. Then, the CPU 1020 reads out the program module 1093 and the program data 1094 stored in the memory 1010 or the hard disk drive 1090 into the RAM 1012 as necessary, and executes various processing procedures.
[0128] Note that the program module 1093 and program data 1094 relating to the program are not limited to being stored in the hard disk drive 1090, and may be stored in, for example, a removable storage medium and read by the CPU 1020 via a disk drive or the like. Alternatively, the program module 1093 and program data 1094 relating to the program may be stored in another computer connected via a network (such as a local area network (LAN) or wide area network (WAN)) and read by the CPU 1020 via the network interface 1070.
[0129] The above-described embodiments and their modifications are included in the technology disclosed in this application, as well as in the scope of the invention described in the claims and their equivalents. [Explanation of symbols]
[0130] 10. Anomaly detection device 11 Input section 12 Output section 13. Communications Department 14 Storage section 15 Control section 15a Acquisition part 15b 1st extraction part 15c 2nd extraction part 15d Calculation section 15e Specific part 20 Time Series Data
Claims
1. An acquisition unit that acquires time series data of a detection target in which an abnormality is detected at a predetermined time point; a first extraction unit that extracts, from the time series data, a feature of a feature amount direction in a time section before the predetermined time point; a second extraction unit that extracts a time direction feature in the time section from the feature amount directional feature; a calculation unit that calculates an anomaly score at a predetermined time point based on the feature in the feature amount direction and the feature in the time direction, and calculates a contribution degree of the feature amount direction and a contribution degree of the time direction to the anomaly score before the predetermined time point; An anomaly detection device comprising:
2. The calculation unit calculates an anomaly score at a predetermined time point by an unsupervised learning model, and calculates a contribution degree in the feature amount direction and a contribution degree in the time direction; an identification unit that identifies a cause of the abnormality using a contribution degree in the feature amount direction or a contribution degree in the time direction when the abnormality is detected based on the abnormality score, The anomaly detection device according to claim 1 , further comprising:
3. The first extraction unit performs two-dimensional convolution on each feature amount of the time series data to extract a feature amount direction feature, The second extraction unit performs one-dimensional convolution on each of the feature amounts of the feature amount direction features to extract the feature amount in the time direction.
3. The abnormality detection device according to claim 1 or 2.
4. the calculation unit calculates the anomaly score and calculates the contribution degree of the feature amount direction and the contribution degree of the time direction using an unsupervised learning model that has been trained using a loss function constituted by a penalty for at least one of a prediction error related to the anomaly score, a contribution degree of the feature amount direction, and a contribution degree of the time direction; 4. The abnormality detection device according to claim 1, wherein the abnormality detection device is a detection unit that detects an abnormality.
5. An anomaly detection method executed by an anomaly detection device, comprising: An acquisition step of acquiring time series data of a detection target in which an abnormality is detected at a predetermined time point; a first extraction step of extracting from the time series data a feature of a feature amount direction in a time section before the predetermined time point; a second extraction step of extracting a time direction feature in the time section from the feature amount directional feature; a calculation step of calculating an anomaly score at a predetermined time point based on the feature in the feature amount direction and the feature in the time direction, and calculating a contribution degree of the feature amount direction and a contribution degree of the time direction to the anomaly score before the predetermined time point; An anomaly detection method comprising:
6. An anomaly detection program for causing a computer to function as the anomaly detection device according to any one of claims 1 to 4.
Citation Information
Patent Citations
Abnormality determination device and abnormality determination method
JP2020107248A
Data processing device, data processing method and data processing program
JP2020149601A
Intelligent time-series analytic engine
US20210056430A1
Cited By
Abnormality management device, abnormality management method, and abnormality management system
JP7714828B1
Abnormality management device and abnormality management method
JP7762830B1
Abnormality management device and abnormality management method
JP7762832B1
Abnormality management device and abnormality management method
JP7826559B1
Anomaly detection device and anomaly detection method
JP7844736B1
