Predicting the Spread of Asian Giant Hornets in North America and Identifying Positive Sightings using a Convolutional Neural Network

Loading...
Thumbnail Image

Authors

Crooks, Sabrina
Davidson, Shirley
Norton, Madeline

Date of Issue

2021-02

Type

Research Paper

Language

en_US

Subject Keywords

Research Projects

Organizational Units

Journal Issue

Other Titles

Abstract

Soil temperatures, traveling distances, clutch sizes, and mortality rates among other factors must be analyzed to determine the potential spread of Asian Giant Hornets in North America. Our goals were: to create models to calculate this population growth and spread, to create a convolutional neural network which could correctly distinguish between positive and negative reported sightings, and to format a website which would allow for the spread to be tracked by both the Washington State Department of Agriculture and concerned citizens. We created three models to represent the population of Asian Giant Hornets. Model one was the simplest and did not include any limitations to population growth. Model two brings into account the mortality rate of queens that freeze in winter soil. Model three expands upon the second model by adding in the carrying capacity of a region based on the minimum necessary distance between two hives of Asian Giant Hornets. To best handle the influx of misidentified hornet images, we created a model using a convolutional neural network (CNN) to classify the images. Using the image data provided, the CNN was trained using a large amount of the data to identify what was and what was not an Asian Giant Hornet. This CNN was able to identify a Asian Giant Hornet with 90% accuracy. The convoluted neural network and the population predictive models were combined into a website. This website will allow pictures of reported sightings to be uploaded. Once uploaded, the trained CNN will analyze the photo and determine its verification ID, positive or negative.

Description

Competition: The Interdisciplinary Contest in Modeling (ICM)

Citation

Publisher

License

Journal

Volume

Issue

PubMed ID

DOI

ISSN

EISSN