A Research of Vespa Mandarinia through Visualization Technology and Convolution Neural Network

Abstract

After being discovered in September 2019, Vespa mandarinia, a fierce alien invader, quickly got public’s attention. The inhabitants of Washington State flooded to report the suspected cases around them. However, most of them were proved to be false information. How to interpret the provided data and determine the priority of dealing with the problem is of vital importance. Based on this situation, this paper develops a model that robustly and accurately predicts the spread of Vespa mandarinia and the facticity of provided reports. By adopting the idea of mining common behavior patterns, this paper took full use of other wasps’ data to make up for our lacking data on Vespa mandarinia. Though few samples of Vespa mandarinia been provided, this paper took complete utilizing of other data. To evaluate our external model, this paper gave 2019 data of the Vespa mandarinia to propagate then the prediction of 2020 turned to be perfectly accurate. After this, this paper combined the image, location and time information together to predict the sample. This paper got 87.6% accuracy on positive samples and 98.7% accuracy on negative samples. The calculated renewal frequency was proved to be fit for Vespa madarinia’s actual living habits. And the MSE of our enhanced Logistic function was 10–4, which can simulate the growth of Vespa mandarinia well and also taken human eradication into consideration.