Prediction of Mosquito Prevalence in a Warm Semi-Arid Climate using Artificial Neural Network (ANN)

Abstract

Mosquito is a disease-causing organism that causes harm to humans and animals alike. Over the years, vector control measures such as the use of insecticides, treated mosquito net, and the prediction of mosquito prevalence using statistical tools and Artificial Neural Network models in different weather terrain have not completely eradicated the problems associated with prevalence of mosquito. More so, there is no research available in literature to predict the prevalence of mosquito using artificial neural network in a warm semi-arid climate such as that of Yola, Northeastern Nigeria. This research endeavored to achieve this aim. This study built a prototype artificial neural network model that is capable of predicting mosquito prevalence. The model is a feed forward multi-layer perceptron that was implemented using the supervised learning method and optimized using the back propagation algorithm. The model has four (4) input features, which are weather data (maximum temperature, minimum temperature, relative humidity and rainfall) which were adopted for the research. After compilation, the new model was trained and validated using sourced data by the researcher. To train the model, 80% of the data was used while 20% was used for the validation. The proposed model is a keras sequential classification model that was built in anaconda using the python programming language. The optimal model has three hidden layers of 40 30 and 20 neurons with Sigmoid and ReLu activation function respectively. The simulation of the prototype model recorded 96.67% accuracy with good fit. This research shows that the artificial neural network model is an effective tool in predicting mosquito prevalence in a warm semi-arid climatic region and thus recommends the use of more data and training epochs to increase accuracy and subsequent implementation of the model in real life for prediction of mosquito prevalence.