Harvest–release decisions in recreational fisheries. Kaemingk, M., A., Hurley, K., L., Chizinski, C., J., & Pope, K., L. Canadian Journal of Fisheries and Aquatic Sciences, 77(1):194-201, 2020.
abstract   bibtex   
Most fishery regulations aim to control angler harvest. Yet, we lack a basic understanding of what actually determines the angler’s decision to harvest or release fish caught. We used XGBoost, a machine learning algorithm, to develop a predictive angler harvest–release model by taking advantage of an extensive recreational fishery data set (24 water bodies, 9 years, and 193 523 fish). We were able to successfully predict the harvest–release outcome for 99% of fish caught in the training data set and 96% of fish caught in the test data set. Unsuccessful predictions were mostly attributed to predicting harvest of fish that were released. Fish length was the most essential feature examined for predicting angler harvest. Other important predictive harvest– release features included the number of individuals of the same species caught, geographic location of an angler’s residence, distance traveled, and time spent fishing. The XGBoost algorithm was able to effectively predict the harvest–release decision and revealed hidden and intricate relationships that are often unaccounted for with classical analysis techniques. Exposing and accounting for these angler–fish intricacies is critical for fisheries conservation and management.
@article{
 title = {Harvest–release decisions in recreational fisheries},
 type = {article},
 year = {2020},
 identifiers = {[object Object]},
 pages = {194-201},
 volume = {77},
 id = {36db48d4-8aa1-34b5-b4c1-36546d1bd497},
 created = {2020-05-26T15:50:05.782Z},
 file_attached = {false},
 profile_id = {a9d8b7dd-4b33-3a9e-9b6e-5dc5cb559585},
 group_id = {c38dcf34-fa27-380f-93ef-2c3ab1ee926b},
 last_modified = {2020-05-26T15:50:05.782Z},
 read = {false},
 starred = {false},
 authored = {false},
 confirmed = {true},
 hidden = {false},
 private_publication = {false},
 abstract = {Most fishery regulations aim to control angler harvest. Yet, we lack a basic understanding of what actually determines the angler’s decision to harvest or release fish caught. We used XGBoost, a machine learning algorithm, to develop a predictive angler harvest–release model by taking advantage of an extensive recreational fishery data set (24 water bodies, 9 years, and 193 523 fish). We were able to successfully predict the harvest–release outcome for 99% of fish caught in the training data set and 96% of fish caught in the test data set. Unsuccessful predictions were mostly attributed to predicting harvest of fish that were released. Fish length was the most essential feature examined for predicting angler harvest. Other important predictive harvest– release features included the number of individuals of the same species caught, geographic location of an angler’s residence, distance traveled, and time spent fishing. The XGBoost algorithm was able to effectively predict the harvest–release decision and revealed hidden and intricate relationships that are often unaccounted for with classical analysis techniques. Exposing and accounting for these angler–fish intricacies is critical for fisheries conservation and management.},
 bibtype = {article},
 author = {Kaemingk, Mark A. and Hurley, Keith L. and Chizinski, Christopher J. and Pope, Kevin L.},
 journal = {Canadian Journal of Fisheries and Aquatic Sciences},
 number = {1}
}

Downloads: 0