Classify Sina Weibo users into High or Low happiness Groups Using Linguistic and Behavior Features

TL;DR

This study develops a decision tree model to classify Sina Weibo users into high or low happiness groups based on linguistic and behavioral features, demonstrating the feasibility of online happiness assessment.

Contribution

It introduces a novel approach using linguistic and behavior features from social media data to classify user happiness levels with a decision tree model.

Findings

24 features significantly differentiate happiness groups

Decision Tree achieves 67.7% precision in classification

Online happiness classification is feasible using social media data

Abstract

It's of great importance to measure happiness of social network users, but the existing method based on questionnaires suffers from high costs and low efficiency. This paper aims at identifying social network users' happiness level based on their Web behavior. We recruited 548 participants to fill in the Oxford Happiness Inventory (OHI) and divided them into two groups with high/low OHI score. We downloaded each Weibo user's data by calling API, and extracted 103 linguistic and behavior features. 24 features are identified with significant difference between high and low happiness groups. We trained a Decision Tree on these 24 features to make the prediction of high/low happiness group. The decision tree can be used to identify happiness level of any new social network user based on linguistic and behavior features. The Decision Tree can achieve 67.7% on precision. Although the capability of our Decision Tree is not ideal, classifying happiness via linguistic and behavior features on the Internet is proved to be feasible.