Frequent Pattern Mining in Multiple Trajectories of Football Players

• Published in: Lecture notes in engineering and computer science Vol. 2233/2234; p. 314
• Main Authors: Suzuki, Yuto, Ozaki, Tomonobu
• Format: Journal Article
• Language: English
• Published: Hong Kong International Association of Engineers 14.03.2018
• Subjects: Algorithms; Athletes; Data mining; Enumeration; Football; Game theory; Pattern analysis; Players; Professional football; Quantitative analysis; Tactics; Trajectories
• ISSN: 2078-0958; 2078-0966

Recently, it has been regarded as important in many sports fields to evaluate the tactics and athletes using actual play records. In this paper, as a first step towards a quantitative evaluation of the strategies in football games, we propose an algorithm for discovering frequent patterns on simultaneous trajectories of multiple football players. In the algorithm, given trajectories are firstly converted into a set of labeled sub-trajectories corresponding to the interval-based events. A pattern enumeration algorithm is then applied to the obtained interval-based events with a consideration of the order of events, the time difference and the spatial spread of subtrajectories. We introduce variables for subjects of events (subtrajectories) in the pattern. By using variables, we can recognize which events were played by the same player and which events were played differently. In addition, it is possible to extract a pattern which absorb the difference of concrete players. To evaluate the proposed algorithm, we conduct experiments using real trajectory datasets on nine matches in Japanese professional football league. The results on the computation time and the number of extracted patterns show the feasibility and effectiveness of the algorithm. In addition, we succeeded in extracting meaningful patterns representing certain offensive and defensive strategies formed by multiple football players