Application of Negative and Positive Association Rules in Mental Health Analysis of College Students

Abstract

College students are suffering from many mental health problems including somatization, obsessive, interpersonal sensitivity, depression, anxiety, hostility, fear, paranoia and psychosis, which can bring a lot of negative effects to them. Many association rules mining algorithms have been used to analyze the relationships between those problems from mental health datasets. However, they only focus on positive association rules (PARs) and don't consider negative association rules (NARs), which can provide much more informative information than the positive ones. So this paper aims to mine both positive and negative association rules (PNARs) from mental health datasets of college students. The form of NARs like a1→a2⇒1r→b2 we mined contains both positive and negative items in each side, but in traditional forms of NARs, it contains all positive items or all negative items in each side like ala2⇒(b1b2) or →(a1a2)⇒b1b2. We first use e-NFIS algorithm to mine positive frequent itemsets (PFIS) and negative frequent itemsets (NFIS) and then use the support-confidence framework to generate the PARs and NARs based on the obtained PFIS and NFIS respectively. The real dataset is collected form 2275 freshmen's symptom self-rating scale (SCL _90) test from one Chinese college. Experimental results verify our approach can easily find PNARs between the various mental problems, and the obtained rules have practical guiding significance for predicting and preventing the mental health of college students.