When a Teen's Stress Level Comes to the Top/Bottom: A Fuzzy Candlestick Line Based Approach on Micro-Blog

摘要

Recent researches on micro-blog based adolescent stress level prediction prove the feasibility of forecasting a teen's future stress level through the stress time series detected from tweets. The previous work focuses on predicting the stress level or stress level change at the next time point, and doesn't consider the problem of predicting the future stress trend in a period of time. In this paper, we employ a fuzzy candlestick line based model to address this problem on micro-blog, i.e., when a teen's stress level comes to the top/bottom in the future. The candlestick line technique is a widely used stock trend analysis method in the financial domain. Experienced analysts usually use linguistic variables to describe the candlestick lines, such as long, short, and small. Thus we use the fuzzy set theory to represent the stress candlestick line in this paper. We define the stress patterns as a set of neighboring candlestick lines represented with fuzzy linguistic variables. Based on these fuzzy stress patterns, we make predictions using the fuzzy decision tree model. Experiments show the effectiveness of our prediction method.