Hybrid Machine Learning Approach using Wingbeat Frequencies for Mosquito Species Classification

Many of life-threatening diseases such as malaria, dengue and Zika virus are vectors through mosquitoes. Three or four effective mosquito species must be identified to monitor and control their spread. In this study, we propose a hybrid model for classifying the mosquitoes based on their wingbeat frequencies using advanced deep learning techniques for monitoring and controlling their spread. This study proposes a hybrid approach for mosquito species classification based on their wingbeat frequencies, leveraging advanced deep learning techniques. We employ the Xception, a state of art deep learning architecture for feature extraction with classification using XGBoost, a strong ensemble learning algorithm. An accuracy of 95.41% is achieved on a comprehensive dataset of mosquito wingbeat recordings when the proposed model is evaluated. This hybrid approach achieves the feature extraction power of Xception and the predictive abilities of XGBoost and outperforms traditional models and single algorithm approaches. Such models are shown to be feasible in the deployment for real world vector monitoring and disease prevention initiatives with the high accuracy.