Classifying Business Types on Twitter Based on User Influential Analysis

Chanattha Thongsuk; Choochart Haruechaiyasak; Somkid Saelee

doi:10.4028/www.scientific.net/AMR.403-408.3719

Paper Titles

Improve Abstract Data with Feature Selection for Classification Techniques
p.3699

Web Based Application Maintenance Cost Estimation Modeling Using Bayesian SEM
p.3704

Determining the Optimal Parameter for Education Surveillance System
p.3709

AGRIX: An Ontology Based Agricultural Expertise Retrieval Framework
p.3714

Classifying Business Types on Twitter Based on User Influential Analysis
p.3719

Quality Assessment in Vocational Education Based on a Hierarchical Classification Technique
p.3724

Health Index Evaluation on Acoustic Emission Signal for Concrete Structure by Intensity Analysis Method
p.3729

Studies on Capacitive Switch Model Based on W-EFG
p.3734

Physical Characteristic Analysis of Solenoid-Based Coil Structure
p.3739

HomeAdvanced Materials ResearchAdvanced Materials Research Vols. 403-408Classifying Business Types on Twitter Based on...

Classifying Business Types on Twitter Based on User Influential Analysis

Abstract:

In this paper, we study the correlation between incoming link of users on Twitter, a micro-blogging website online social. Finding the influential user can apply to recommend users to follow their interest’s businesses domain. To analyze and find characteristic of the influential users for applying to improve the performance of recommender system. We use user’s Twitter posts from any solution into predefined business types. In this paper, we propose solution to applied user selection by comparing among three parameters: (1) the number of relevant posts (NumRP) (2) the number of incoming link from business follower (NumUFI) (3) the number of incoming link from every follower (NumTI). Each parameter is ranked and incremental organized into three groups of each parameter: (1) Top-100 (2) Top-200 and (3) Top-300. After that, we applied posts of selected users to build classification model. Comparison between among three user selection parameters and three user groups. From the experimental results, the performance of NumRP yielded the F-measure higher than NumUFI and NumTI respectively. In addition, users who organized into Top-100 user group of each user selection method are influential users.