利用时间序列异常进行预测性实验 [英] Predictive Experiment Utilizing Time Series Anomaly

问题描述

我正在尝试利用时间序列异常算法来预测被认为不寻常或可能因键入错误而输入的发票.不幸的是，我没有成功地将其发布为一个Web服务应用程序 我可以用来退还可疑发票.我可以显示我提交的发票清单，但不能将我的设计设置为可预测的网络实验来审查所有将来有风险的潜在发票.我相信这是因为没有选择来转换 时间序列异常以进行预测性实验.有什么建议，推荐或实验可以用来识别不良发票吗?谢谢您的考虑.

I am attempting to utilize the Time-Series Anomaly algorithm to predict invoices that are considered unusual or likely to be entered with a keying error. Unfortunately, I have not been successful at publishing this as a web service application that I can utilize to return suspect invoices. I can reveal a list of invoices that I submit but I can't set my design up as a predictive web experiment to review all future potential invoices at risk. I believe this is because there is no option to convert the time series anomaly to a predictive experiment. Is there any advice, recommendation or experiment that I could utilize for achieving the identification of bad invoices? Thank you for considering.

推荐答案

我刚刚使用时间序列异常"检测模块创建了一个示例实验，如下所示，并且能够将其转换为Web服务，甚至可以对其进行测试.

I just created a sample experiment with a 'Time Series Anomaly' detection module as shown below and was able to convert it into a web service and even test it. 

测试Web服务:

Web服务的输出:

此致，
Jaya

Regards,
Jaya

这篇关于利用时间序列异常进行预测性实验的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！