http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
정현용,이대경,John Losey 한국HCI학회 2022 한국HCI학회 학술대회 Vol.2022 No.2
It is difficult to evaluate the reduction rate of a species that has a high number of habitats and populations. In addition to the amount of insufficient data present to cover the entire habitat, the inconsistency of study areas and investigation efforts between monitoring programs is a major obstacle to quantifying long-term population decline. To overcome the bias from (1) data insufficiency and (2) inconsistency derived from compiling various investigative sources, including citizen science projects, this study proposes a procedure to estimate the reduction rate using machine learning. After predicting the possibility of occupancy every year through a fitted model in all habitats where target species have been recorded, a trend line is formulated and used for approximating reduction. By the newly described oversampling approach, including indirect occupancy data points, we were able to secure about three times the amount of information on the direct occupancy data points. Compared to the model that learned only the direct occupancy data set, the performance matrix of the model utilizing the new oversampling approach was about 10% higher overall. As a result, the estimated reduction rate of Coccinella novemnotata was 31.2% over the past decade, which corresponds to the VU status in the IUCN Red List. In addition, the machine learning method showed a lower degree of fluctuation in approximations (within 13%) than the relative abundance method (within 42%) and the segmented linear extrapolation method (within 55%), when we integrated 10% of fake information into the data set.