Prediction of the disease controllability in a complex network using machine learning algorithms

Abstract

The application of machine learning (ML) techniques spans a vast spectrum of applications ranging from speech, face and character recognition to medical diagnosis to anomaly detection in data and the general classification, prediction and regression problems. In the present work, we demonstrate the application of regression-based state-of-art machine learning techniques to a prediction of disease controllability on complex networks. The complex network models determine the space for distribution of a population of individuals and their interactions with each other. There is numerous epidemic spreading models such as SI (Susceptible-Infected), SIR (Susceptible-Infected-Recovered), SEIR (Susceptible-Exposed-Infected-Recovered), etc., that govern the disease spreading dynamics over time and the stationary state of the disease. We simulate the disease spreading dynamics on a large number of complex networks examples of standard model networks, and determine the basic reproduction number (R 0 ) for each case. R_0 is a metric that determines whether the disease-free epidemic or an endemic state is asymptotically stable. In other words, it determines whether an infectious disease can spread in the population or will die out in the long run and hence indicates the disease controllability on a population. We aim to predict this quantity (R 0 ), based on the importance of complex networks structural properties using the regression techniques of ML, irrespective of the network type. The prediction is possible because of two facts (a) The structure of complex networks plays an essential role in the spreading processes on networks. (b) Availability of non-linear regression techniques with excellent accuracy for prediction of a quantity even for a data which is highly non-linear.