Temporal Patterns of Hotspot Sequences for Early Detection of Peatland Fire in Riau Province

Abstract

Indonesia’s peatland condition is getting worse because of peatland fire. Peatland fire causes many negative impacts, so early detection is needed. Data mining is one of approach that can be used for finding sequential pattern from hotspot data as one of indicators for peatland fire. This study aims to find sequential patterns on hotspot data in Riau province Indonesia. The Douglas-Peucker algorithm and substring tree structure concept were used for finding the patterns. The experiment results three types of sequential patterns, namely sequences of date, day, and location of hotspot data in 2014. The most interesting frequent pattern of hotspot occurrence is 11 March 2014 -1 13 March 2014 meaning that the hotspot occurrences on 11 March 2014 was followed by the occurrences in the same location on 13 March 2014. This pattern was found in 9 of 12 districts in Riau Province. Another interesting frequent pattern based on day of occurrence is Friday -1 Saturday -1 Sunday meaning that there was hotspot in Friday, Saturday, and Sunday in the same location. The experiment results show that about 22.77% hotspots in 2014 are considered as strong indicator for peatland fires because it occurred in sequence patterns.