KDD 2021: Multivariate Time Series Anomaly Detection and Interpretation using Hierarchical Inter-Metric and Temporal Embedding
MIT License
195
stars
46
forks
source link
Consultation on issues related to 《Multivariate Time Series Anomaly Detection and Interpretation using Hierarchical Inter-Metric and Temporal Embedding》 #20
I am a graduate student currently studying at the National University of Defense Technology. Recently, while conducting literature research, I came across your work titled "Multivariate Time Series Anomaly Detection and Interpretation using Hierarchical Inter-Metric and Temporal Embedding". Firstly, I would like to express my sincere gratitude for the valuable contribution you have made with "Multivariate Time Series Anomaly Detection and Interpretation using Hierarchical Inter-Metric and Temporal Embedding". I am interested in incorporating your work into my research, but I am uncertain whether your methods can address the issues I am encountering. Therefore, I am taking the liberty to reach out to you via email to seek your advice.
Could you please provide your insights on whether your methods are effective in detecting anomalies in the context described? Any guidance or suggestions you could offer would be greatly appreciated.
Problem description: Extract variable groups with a certain linear relationship from the data.(The number of variables with relationships is ≤ 4, and the coefficients before the variables can only be integers)
For example: there are 100 variables (x1 - x100), record the variable values of all variables at 100 timestamps (t1 - t100), which is a 100*100 matrix. And have a matrix that satisfies a certain relationship and a matrix that does not satisfy this relationship. The schematic diagram is as follows:
In the figure above, data 1 satisfies the relationship x3+x98 > 2*x1+x2 from t1 to t100, while data 2 has moments when it does not satisfy this relationship (t5 - t100).
Purpose: Given the above two types of data, we can analyze which variables have a certain relationship in data 1. However, this relationship is not completely satisfied in data 2. The specific relationship does not need to be given, only the variable group is given.And it doesn't need to be very accurate, just provide some candidate variable sets.
Thank you for your time and consideration.. I look forward to hearing from you.
Best regards.
您好，最近在检索相关文献时，注意到了您的工作《Multivariate Time Series Anomaly Detection and Interpretation using Hierarchical Inter-Metric and Temporal Embedding》，首先我想对您为《Multivariate Time Series Anomaly Detection and Interpretation using Hierarchical Inter-Metric and Temporal Embedding》做出的宝贵贡献表示衷心的感谢。我想将您的工作应用在我的研究内容上，但是我并不确定您的方法是否可以解决我所遇到的问题，所以冒昧的写邮件打扰到您，向您咨询。
Hello,
I hope this message finds you well.
I am a graduate student currently studying at the National University of Defense Technology. Recently, while conducting literature research, I came across your work titled "Multivariate Time Series Anomaly Detection and Interpretation using Hierarchical Inter-Metric and Temporal Embedding". Firstly, I would like to express my sincere gratitude for the valuable contribution you have made with "Multivariate Time Series Anomaly Detection and Interpretation using Hierarchical Inter-Metric and Temporal Embedding". I am interested in incorporating your work into my research, but I am uncertain whether your methods can address the issues I am encountering. Therefore, I am taking the liberty to reach out to you via email to seek your advice.
Could you please provide your insights on whether your methods are effective in detecting anomalies in the context described? Any guidance or suggestions you could offer would be greatly appreciated.
Problem description: Extract variable groups with a certain linear relationship from the data.(The number of variables with relationships is ≤ 4, and the coefficients before the variables can only be integers)
For example: there are 100 variables (x1 - x100), record the variable values of all variables at 100 timestamps (t1 - t100), which is a 100*100 matrix. And have a matrix that satisfies a certain relationship and a matrix that does not satisfy this relationship. The schematic diagram is as follows:
In the figure above, data 1 satisfies the relationship x3+x98 > 2*x1+x2 from t1 to t100, while data 2 has moments when it does not satisfy this relationship (t5 - t100).
Purpose: Given the above two types of data, we can analyze which variables have a certain relationship in data 1. However, this relationship is not completely satisfied in data 2. The specific relationship does not need to be given, only the variable group is given.And it doesn't need to be very accurate, just provide some candidate variable sets.
Thank you for your time and consideration.. I look forward to hearing from you.
Best regards.
您好，最近在检索相关文献时，注意到了您的工作《Multivariate Time Series Anomaly Detection and Interpretation using Hierarchical Inter-Metric and Temporal Embedding》，首先我想对您为《Multivariate Time Series Anomaly Detection and Interpretation using Hierarchical Inter-Metric and Temporal Embedding》做出的宝贵贡献表示衷心的感谢。我想将您的工作应用在我的研究内容上，但是我并不确定您的方法是否可以解决我所遇到的问题，所以冒昧的写邮件打扰到您，向您咨询。
您能否提供您的观点，说明您的方法是否能有效地检测所述背景下的异常情况？如能提供任何指导或建议，我们将不胜感激。
问题描述：从数据中提取出具有某种线性关系的变量组（存在关系的变量数量≤4个且变量前的系数只能是整数）。
例如：有100个变量（x1 - x100），记录所有变量在100个时间戳（t1 - t100）的变量值，即为100*100的矩阵。并拥有满足某种关系的矩阵和不满足这种关系的矩阵，示意图如下：
上图中，数据1在 t1 -t100 中均满足 x3+x98 > 2*x1+x2 这一关系，而数据 2 有不满足该关系的时刻（t5 - t100）
目的：给出上述两种类型的数据，能够分析出哪些变量间在数据1存在某种关系，而在数据2中不完全满足该关系，具体是什么关系可以不给出，只需给出变量组。而且可以不需要非常准确，给出一些候选变量集合即可。
再次感谢您，期待您的回复。