State Discovery and Prediction from Multivariate Sensor Data

Olli Pekka Rinta-Koski, Miki Sirola, Le Ngu Nguyen, Jaakko Hollmén*

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference article in proceedingsScientificpeer-review


The advent of cloud computing and autonomous data centers operating fully without human supervision has highlighted the need for fault-tolerant architectures and intelligent software tools for system parameter optimization. Demands on computational throughput have to be balanced with environmental concerns, such as energy consumption and waste heat. Using multivariate time series data collected from an experimental data center, we build a state model using clustering, then estimate the states represented by the clusters using both a hidden Markov model and a long-short term memory neural net. Knowledge of future states of the system can be used to solve tasks such as reduced energy consumption and optimized resource allocation in the data center.

Original languageEnglish
Title of host publicationAdvanced Analytics and Learning on Temporal Data - 6th ECML PKDD Workshop, AALTD 2021, Revised Selected Papers
EditorsVincent Lemaire, Simon Malinowski, Anthony Bagnall, Thomas Guyet, Romain Tavenard, Georgiana Ifrim
Number of pages15
ISBN (Print)9783030914448
Publication statusPublished - 2021
MoE publication typeA4 Conference publication
EventInternational Workshop on Advanced Analytics and Learning on Temporal Data - Virtual, Online
Duration: 13 Sept 202117 Sept 2021
Conference number: 6

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume13114 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349


WorkshopInternational Workshop on Advanced Analytics and Learning on Temporal Data
Abbreviated titleAALTD
CityVirtual, Online


Dive into the research topics of 'State Discovery and Prediction from Multivariate Sensor Data'. Together they form a unique fingerprint.

Cite this