TY - JOUR
T1 - Discovering diverse human behavior from two-dimensional preferences
AU - Yang, Pu Tai
AU - Huang, Cheng Kui
AU - Chu, Huan Lin
AU - Chuang, Yung-Ting
N1 - Publisher Copyright:
© 2018 Elsevier B.V.
PY - 2018/7/15
Y1 - 2018/7/15
N2 - Among the many types of ambiguous and diverse human behaviors, preference ranking and preferred temporal relationships are two human subjective perceptions, and both can be expressed as sequences. For example, information management researchers prefer journal A over B, and this preference can be represented as the ranking sequence: (A > B). When the order of submission is first A and then B, it can be expressed as the temporal sequence: (A → B). In practice, these two preference sequences may be applicable to people with regard to the same items at the same time, which is called “two-dimensional preference” in this study. Based on these concepts, this study defines a novel model and an associated algorithm for mining two-dimensional patterns by combining preference ranking and temporal sequences. The discovered two-dimensional patterns can be categorized into eight types, including consensus, ranking-compromise, temporal-compromise, and conflict patterns. Two experiments in two application areas, namely journal submission and stock purchase, were designed to collect authentic datasets, and demonstrate their managerial meaning. The experimental results show that in most cases, the ordering of the temporal sequences follows the preference ranking sequence, except for several two-dimensional patterns involving high-risk items.
AB - Among the many types of ambiguous and diverse human behaviors, preference ranking and preferred temporal relationships are two human subjective perceptions, and both can be expressed as sequences. For example, information management researchers prefer journal A over B, and this preference can be represented as the ranking sequence: (A > B). When the order of submission is first A and then B, it can be expressed as the temporal sequence: (A → B). In practice, these two preference sequences may be applicable to people with regard to the same items at the same time, which is called “two-dimensional preference” in this study. Based on these concepts, this study defines a novel model and an associated algorithm for mining two-dimensional patterns by combining preference ranking and temporal sequences. The discovered two-dimensional patterns can be categorized into eight types, including consensus, ranking-compromise, temporal-compromise, and conflict patterns. Two experiments in two application areas, namely journal submission and stock purchase, were designed to collect authentic datasets, and demonstrate their managerial meaning. The experimental results show that in most cases, the ordering of the temporal sequences follows the preference ranking sequence, except for several two-dimensional patterns involving high-risk items.
KW - Consensus patterns
KW - Data mining
KW - Heterogeneous databases
KW - Ranking sequences
KW - Temporal sequences
UR - http://www.scopus.com/inward/record.url?scp=85045035164&partnerID=8YFLogxK
U2 - 10.1016/j.knosys.2018.04.003
DO - 10.1016/j.knosys.2018.04.003
M3 - Article
AN - SCOPUS:85045035164
SN - 0950-7051
VL - 152
SP - 11
EP - 25
JO - Knowledge-Based Systems
JF - Knowledge-Based Systems
ER -