Discovery of activity composites using topic models: An analysis of unsupervised methods. Seiter, J., Amft, O., Rossi, M., & Tröster, G. Pervasive and Mobile Computing, 15:215–-227, December, 2014. doi abstract bibtex In this work we investigate unsupervised activity discovery approaches using three topic model (TM) approaches, based on Latent Dirichlet Allocation (LDA), n-gram TM (NTM), and correlated TM (CTM). While LDA structures activity primitives, NTM adds primitive sequence information, and CTM exploits co-occurring topics. We use an activity composite/primitive abstraction and analyze three public datasets with different properties that affect the discovery, including primitive rate, activity composite specificity, primitive sequence similarity, and composite-instance ratio. We compare the activity composite discovery performance among the TM approaches and against a baseline using k-means clustering. We provide guidelines for method and optimal TM parameter selection, depending on data properties and activity primitive noise. Results indicate that TMs can outperform kk-means clustering up to 17%, when composite specificity is low. LDA-based TMs showed higher robustness against noise compared to other TMs and k-means.
@Article{Seiter2014-J_PervasiveMobComput,
Title = {Discovery of activity composites using topic models: An analysis of unsupervised methods},
Author = {Julia Seiter and Oliver Amft and Mirco Rossi and Gerhard Tr\"{o}ster},
Journal = {Pervasive and Mobile Computing},
Year = {2014},
Month = {December},
Pages = {215–-227},
Volume = {15},
Abstract = {In this work we investigate unsupervised activity discovery approaches using three topic model (TM) approaches, based on Latent Dirichlet Allocation (LDA), n-gram TM (NTM), and correlated TM (CTM). While LDA structures activity primitives, NTM adds primitive sequence information, and CTM exploits co-occurring topics. We use an activity composite/primitive abstraction and analyze three public datasets with different properties that affect the discovery, including primitive rate, activity composite specificity, primitive sequence similarity, and composite-instance ratio. We compare the activity composite discovery performance among the TM approaches and against a baseline using k-means clustering. We provide guidelines for method and optimal TM parameter selection, depending on data properties and activity primitive noise. Results indicate that TMs can outperform kk-means clustering up to 17\%, when composite specificity is low. LDA-based TMs showed higher robustness against noise compared to other TMs and k-means.},
Doi = {10.1016/j.pmcj.2014.05.007},
File = {Seiter2014-J_PervasiveMobComput.pdf:Seiter2014-J_PervasiveMobComput.pdf:PDF},
Keywords = {Activity routines; Daily routines; Topic modeling; Hierarchical activity recognition; Activity discovery},
Owner = {oamft},
Timestamp = {2013/09/16}
}
Downloads: 0
{"_id":"Xwbbya3yZn6fQEkTN","bibbaseid":"seiter-amft-rossi-trster-discoveryofactivitycompositesusingtopicmodelsananalysisofunsupervisedmethods-2014","downloads":0,"creationDate":"2016-03-10T09:24:25.728Z","title":"Discovery of activity composites using topic models: An analysis of unsupervised methods","author_short":["Seiter, J.","Amft, O.","Rossi, M.","Tröster, G."],"year":2014,"bibtype":"article","biburl":"http://cl94754.tmweb.ru/fish/files/neuesFile.bib","bibdata":{"bibtype":"article","type":"article","title":"Discovery of activity composites using topic models: An analysis of unsupervised methods","author":[{"firstnames":["Julia"],"propositions":[],"lastnames":["Seiter"],"suffixes":[]},{"firstnames":["Oliver"],"propositions":[],"lastnames":["Amft"],"suffixes":[]},{"firstnames":["Mirco"],"propositions":[],"lastnames":["Rossi"],"suffixes":[]},{"firstnames":["Gerhard"],"propositions":[],"lastnames":["Tröster"],"suffixes":[]}],"journal":"Pervasive and Mobile Computing","year":"2014","month":"December","pages":"215–-227","volume":"15","abstract":"In this work we investigate unsupervised activity discovery approaches using three topic model (TM) approaches, based on Latent Dirichlet Allocation (LDA), n-gram TM (NTM), and correlated TM (CTM). While LDA structures activity primitives, NTM adds primitive sequence information, and CTM exploits co-occurring topics. We use an activity composite/primitive abstraction and analyze three public datasets with different properties that affect the discovery, including primitive rate, activity composite specificity, primitive sequence similarity, and composite-instance ratio. We compare the activity composite discovery performance among the TM approaches and against a baseline using k-means clustering. We provide guidelines for method and optimal TM parameter selection, depending on data properties and activity primitive noise. Results indicate that TMs can outperform kk-means clustering up to 17%, when composite specificity is low. LDA-based TMs showed higher robustness against noise compared to other TMs and k-means.","doi":"10.1016/j.pmcj.2014.05.007","file":"Seiter2014-J_PervasiveMobComput.pdf:Seiter2014-J_PervasiveMobComput.pdf:PDF","keywords":"Activity routines; Daily routines; Topic modeling; Hierarchical activity recognition; Activity discovery","owner":"oamft","timestamp":"2013/09/16","bibtex":"@Article{Seiter2014-J_PervasiveMobComput,\n Title = {Discovery of activity composites using topic models: An analysis of unsupervised methods},\n Author = {Julia Seiter and Oliver Amft and Mirco Rossi and Gerhard Tr\\\"{o}ster},\n Journal = {Pervasive and Mobile Computing},\n Year = {2014},\n\n Month = {December},\n Pages = {215–-227},\n Volume = {15},\n\n Abstract = {In this work we investigate unsupervised activity discovery approaches using three topic model (TM) approaches, based on Latent Dirichlet Allocation (LDA), n-gram TM (NTM), and correlated TM (CTM). While LDA structures activity primitives, NTM adds primitive sequence information, and CTM exploits co-occurring topics. We use an activity composite/primitive abstraction and analyze three public datasets with different properties that affect the discovery, including primitive rate, activity composite specificity, primitive sequence similarity, and composite-instance ratio. We compare the activity composite discovery performance among the TM approaches and against a baseline using k-means clustering. We provide guidelines for method and optimal TM parameter selection, depending on data properties and activity primitive noise. Results indicate that TMs can outperform kk-means clustering up to 17\\%, when composite specificity is low. LDA-based TMs showed higher robustness against noise compared to other TMs and k-means.},\n Doi = {10.1016/j.pmcj.2014.05.007},\n File = {Seiter2014-J_PervasiveMobComput.pdf:Seiter2014-J_PervasiveMobComput.pdf:PDF},\n Keywords = {Activity routines; Daily routines; Topic modeling; Hierarchical activity recognition; Activity discovery},\n Owner = {oamft},\n Timestamp = {2013/09/16}\n}\n\n","author_short":["Seiter, J.","Amft, O.","Rossi, M.","Tröster, G."],"key":"Seiter2014-J_PervasiveMobComput","id":"Seiter2014-J_PervasiveMobComput","bibbaseid":"seiter-amft-rossi-trster-discoveryofactivitycompositesusingtopicmodelsananalysisofunsupervisedmethods-2014","role":"author","urls":{},"keyword":["Activity routines; Daily routines; Topic modeling; Hierarchical activity recognition; Activity discovery"],"downloads":0,"html":""},"search_terms":["discovery","activity","composites","using","topic","models","analysis","unsupervised","methods","seiter","amft","rossi","tröster"],"keywords":["activity routines; daily routines; topic modeling; hierarchical activity recognition; activity discovery"],"authorIDs":["56e13d4802df45160b0000aa"],"dataSources":["LsXrfomMC874zbKcK"]}