Machine Learning Predictive Models for Improved Acoustic Disdrometer

The trend of technology nowadays requires massive machine-to-machine communications and this can be done only through the application of artificial intelligence, deep learning, and machine learning to different devices through wireless sensor networks. One of the applications is through the development of acoustic disdrometer. Acoustic disdrometer is a tool that measures the amount of rainfall through the sound produced as the raindrops hit the piezoelectric sensors. With this, the main purpose of this study is to develop predictive models through the application of machine learning algorithms that can be used to categorize the intensity of the amount of rainfall from ambient noise. In the study, there were three machine learning algorithms that were used, namely: support vector machine (SVM), k-nearest neighbors and Naive-Bayes classifier. All models obtain confusion matrix (CM) accuracies of 99.14%, 99.14% and 89.27%, respectively. These predictive models were successfully implemented and validated through cross validation (CV) and out-of-sample accuracies.