ANALYSIS OF ATTRIBUTES TO IDENTIFY OPINION LEADERSHIP IN COMMUNITIES USING DECISION TREE TECHNIQUE FOR IMBALANCED DATA

Authors

  • Sudarat Sangkeaw Faculty of Business Administration, Maejo University, Sansai, Chiangmai, 50290
  • Piyawan Siripraseotsin Faculty of Business Administration, Maejo University, Sansai, Chiangmai, 50290
  • Marut Buranarach Language and Semantic Technology Laboratory, National Electronics and Computer Technology Center (NECTEC), Thailand

Keywords:

Opinion Leader, Decision Tree, Imbalanced dataset, SMOTE, HDDT.

Abstract

Opinion Leader is a key person to the Word-of-Mouth strategy in the marketing field.  It represents an individual who has an ability to change the other’s opinion or behavior by using an interpersonal influence. Although opinion leaders have been explored in various contexts, research on tools to identify opinion leaders for marketers are still limited. This research thus proposes to study on the attribute that can help to identify opinion leaders in community for developing a recommender system. A self-assessment questionnaire was designed and tested for validity and reliability using Exploratory Factor Analysis (EFA) and Confirmatory Factor Analysis (CFA) before used to collect data in a cyclist community. The result showed features for predicting opinion leaders are Knowledge, Extrovert and Self-confidence. Moreover, the balanced dataset with SMOTE (Synthetic Minority Over-sampling Technique) and Spread Subsampling is more effective in improving the performance of prediction results over an imbalanced dataset compared to other techniques.

Downloads

Published

2018-08-04

Issue

Section

บทความอื่นๆ (Other Article)