Premium
Toward efficient business behavior prediction using location‐based social networks
Author(s) -
Al Sonosy Ola A.,
Rady Sherine,
Badr Nagwa L.,
Hashem Mohammed
Publication year - 2018
Publication title -
wiley interdisciplinary reviews: data mining and knowledge discovery
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.506
H-Index - 47
eISSN - 1942-4795
pISSN - 1942-4787
DOI - 10.1002/widm.1263
Subject(s) - computer science , interpolation (computer graphics) , field (mathematics) , process (computing) , data mining , similarity (geometry) , filter (signal processing) , big data , machine learning , artificial intelligence , image (mathematics) , computer vision , mathematics , pure mathematics , operating system
Understanding business behavior in a city requires acquiring huge amount of data coming from diverse field studies. The growing use of mobile devices in social media provides massive data transactions that can replace such data acquired by field studies. Location‐based social networks' (LBSNs') data can be exploited in urban analysis for economic reasons. In this research, the spatial correlation of business turnouts for venues registered in LBSNs is studied for business behavior predictions. A novel similarity embedded (SE)‐spatial interpolation technique is proposed for business turnouts' predictions. The proposed technique utilizes diverse features provided by LBSNs in the prediction process to improve prediction performance. Moreover, a local filter is introduced to avoid local extreme involvements in the prediction process issuing better prediction results. To test the proposed techniques, experimental case study is implemented for predicting business behavior of venues registered in Foursquare in Texas. The proposed SE‐spatial interpolation has shown better prediction accuracy than classical spatial interpolation predictions. The additional integration of the local filter shows further alleviated prediction errors. Furthermore, this study extends the work for the efficient application of the proposed prediction technique in big datasets. An iterative nearest neighbors first search method is designed for accelerating the execution time of the prediction technique implementation regardless the dataset size. The proposed method was tested over several size datasets. The test results show accelerated execution time for the proposed method when compared with the classical implementation execution time. This article is categorized under: Algorithmic Development > Spatial and Temporal Data Mining Fundamental Concepts of Data and Knowledge > Big Data Mining Technologies > Prediction