DSpace Repository

Predicting Customer Behavior using Indexed Data from Video Analytics: Machine Learning Application with Stochastic Gradient-boosted Trees

Show simple item record

dc.contributor.advisor Dailey, Matthew
dc.contributor.author Alfian, Alfian
dc.contributor.other Ekpanyapong, Mongkol
dc.contributor.other Esichaikul, Vatcharaporn
dc.date.accessioned 2018-05-09T06:35:21Z
dc.date.available 2018-05-09T06:35:21Z
dc.date.issued 2018-05
dc.identifier.uri http://www.cs.ait.ac.th/xmlui/handle/123456789/895
dc.description.abstract Retail stores strive to optimize sales. However, optimizing sales is difficult for retailers who do not understand their customers' behavior and how that behavior is correlated with sales. Stores that do not understand such behavior will miss the opportunity to maximize sales. To solve this problem, tools for finding patterns in customer behavior would be useful. Towards an understanding of how such tools could be built, I conducted CCTV camera observation in a coffee shop and ran video analytics to index customer movements according to candidate customer behavior variables such as sitting down and taking coffee to go, then I analyzed the data. As one example of a class of regression algorithms able to handle large time series data sets and nonlinear relationships between independent and dependent variables, I apply stochastic gradient-boosted trees (SGBT) to the problem of predicting hourly customer counts. My experiment generates accurate prediction results for sitting customer counts, but inaccurate counts for takeaway customers. I believe the inaccuracy for takeaway counts is due to the sparsity of the data. Accordingly, I find SGBT to be useful for predicting customer behavior time series variables for which the data are dense. The predictive model developed in this study could give retailers a way to plan to anticipate sales and stock sufficient products predicted to be in demand at particular hours. The SGBT analysis can thus help optimize sales. en_US
dc.description.sponsorship Thailand (HM King’s) en_US
dc.language.iso en_US en_US
dc.subject Retail, Stochastic gradient-boosted trees, Video analytics, Predictive analytics, Predictive model, Large data set, Nonlinear relationships, Dense and sparse data, Time series, Optimizing sales, Customer behavior en_US
dc.title Predicting Customer Behavior using Indexed Data from Video Analytics: Machine Learning Application with Stochastic Gradient-boosted Trees en_US
dc.type Research report en_US


Files in this item

This item appears in the following Collection(s)

Show simple item record

Search DSpace


Advanced Search

Browse

My Account