Efficient event prediction in an IOT environment based on LDA model and support vector machine
Sina Dami, Mahtab Yahaghizadeh · 2018
The internet of things (IOT) environments are constantly changing, so that in such environments cannot be guaranteed exactly analyze and predict the events occurrence. Prediction of such events in IOT environments is a challenging task. In the meantime, support vector machine (SVM) efficiently performs nonlinear data analysis using kernel trick, implicitly mapping their inputs into high-dimensional feature spaces. When dealing with the high-dimensional data, it is commonly to ignore the semantic relation between features with the traditional feature selection method. To solve the problem, we propose a method for IOT-based event prediction that uses latent Dirichlet allocation (LDA) model to increase similarity of same events. Using this method in combination with SVM, we designed a faster and much accurate predictor in an IOT environment. Experimental results on historical event data generated by the IOT cloud showed the proposed method improves performance of outbound flight delay event prediction in terms of precision and recall.