Multivariate Unsupervised Machine Learning for Anomaly Detection in Enterprise Applications

Daniel  Elsner; Pouya Aleatrati Khosroshahi; Alan   MacCormack; Robert  Lagerström

Article | Proceedings of the Hawaii International Conference on System Sciences

Multivariate Unsupervised Machine Learning for Anomaly Detection in Enterprise Applications

by Daniel Elsner, Pouya Aleatrati Khosroshahi, Alan MacCormack and Robert Lagerström

Abstract

Existing application performance management (APM) solutions lack robust anomaly detection capabilities and root cause analysis techniques that do not require manual efforts and domain knowledge. In this paper, we develop a density-based unsupervised machine learning model to detect anomalies within an enterprise application, based upon data from multiple APM systems. The research was conducted in collaboration with a European automotive company, using two months of live application data. We show that our model detects abnormal system behavior more reliably than a commonly used outlier detection technique and provides information for detecting root causes.

Keywords: big data; data science and analytics management; governance and compliance; organizational systems and technology; anomaly detection; application performance management; Machine learning; enterprise architecture;

Format: Electronic

Read Now

Citation:

Elsner, Daniel, Pouya Aleatrati Khosroshahi, Alan MacCormack, and Robert Lagerström. "Multivariate Unsupervised Machine Learning for Anomaly Detection in Enterprise Applications." Proceedings of the Hawaii International Conference on System Sciences 52nd (2019): 5827–5836.

This case describes the operations of a fictitious company that processes Cranberries. The case contains data that allows students to calculate the bottleneck stage in production, and to evaluate alternative investment options for increasing cranberry processing capacity.

Case | HBS Case Collection | October 2018 (Revised February 2020)

Innovation at Uber: The Launch of Express POOL

Chiara Farronato, Alan MacCormack and Sarah Mehta

Set in March 2018, the case follows ride-sharing company Uber as it develops and launches a new product called Express POOL. This product offers a reduced price to riders willing to carpool, walk a short distance to/from their pick-up and drop-off points, and wait a few minutes before being matched to a driver. Two weeks after the launch of Express POOL in six U.S. cities, Uber’s product managers discover that if riders are made to wait five minutes to be matched to a driver—rather than the standard two minutes—rider cancellation rates increase, but Uber’s costs per ride are reduced. Together with data scientists, engineers, and product operations specialists, the product managers must decide whether to keep rider wait times at two minutes or increase wait times to five minutes in the six newly launched cities. The decision is complicated by the fact that Uber’s data science team normally places a five-week moratorium on changes to any new product, to allow robust data to be collected on its performance. This case is paired with a supplementary dataset from Uber (HBS No. 619-702). In advance of the class discussion, students can analyze the data and draw their own conclusions about the trade-offs of maintaining the standard wait times or increasing them.

Keywords: Innovation and Management; Innovation Leadership; Innovation Strategy; Technological Innovation; Technology; Mobile Technology; Software; Technology Platform; Decision Making; Technology Industry; California; San Francisco;

Citation:

Farronato, Chiara, Alan MacCormack, and Sarah Mehta. "Innovation at Uber: The Launch of Express POOL." Harvard Business School Case 619-003, October 2018. (Revised February 2020.) View Details

Cite View Details Educators Purchase

Faculty & Research

Multivariate Unsupervised Machine Learning for Anomaly Detection in Enterprise Applications

Abstract

Citation:

About the Author

More from the Author