Using Machine Learning Regression Model to Predict the Optimum Election Algorithm for Parallel and Distributed Computing Systems

Parallel and distributed computing has become a necessity when it comes to high-performance computing, artificial intelligence, big data analytics, machine learning, deep learning, signal processing and bioengineering. System engineers connect distributed systems, with large number of nodes, that then compute the solution for a problem concurrently. The distributed system has a leader that manages the overall coordination of these connected nodes. This research provides a novel approach to predict the most efficient and time-feasible election algorithm, that must be ran on a distributed system, using linear regression machine learning approach. This research implements the prediction process for three election algorithms and successfully predicts the most efficient and least time-consuming algorithm with 94.98% accuracy. Initially, the dataset for the regression model training was created by executing the three election algorithms parallelly using MPI running 16 parallel processors (cores). The research also presents three regression line equations, trained and modelled on this dataset, that can be used to predict the execution time for the three algorithms for a system with any number of nodes. The results are duly visualized, and regression lines are plotted for comparison.