A MACHINE LEARNING APPROACH FOR PREDICTIVE SELECTION OF EFFICIENT CPU SCHEDULING ALGORITHMS

Abstract

Basic for the efficient operation of multi-tasking operating systems is good CPU scheduling .To accomplish this, we have designed and developed a tool called CPU Scheduling Simulator which can implement and compare a variety of scheduling algorithms including the First Come First Serve (FCFS), Shortest Job First (SJF), and Round Robin (RR) among others. Its evaluations of these algorithms are based on key performance indices such as Turnaround Time (TAT), Operations Duration (OPD) and Response Time (Response), offering details down to a level finer than you might think into just how well each one works for any given set of applications or circumstances. Optimizing the performance of operating systems is crucial, but with modern computational needs this has to be done efficiently. Traditional scheduling algorithms (FCFS, SJF, Priority, Round Robin) can produce inconsistent results in varying system environments. According to this research, a machine learning based framework has been presented to dynamically predict which CPU Scheduling algorithm is best suited for a given process context. By using models such as Support Vector Machine (SVM), Logistic Regression (LR) and Stochastic Gradient Descent (SGD), the system analyzes process attributes such as arrival time, burst time, priority and quantum, and selects the most effective scheduling strategy. Of all the models tested, SVM achieved the highest prediction accuracy: 94.56%. It is better at catching up on slack sites and in queuing networks. Our results highlight the potential for inserting smart forecasting systems into operating systems to lessen turnaround and save resources. Thus it is in keeping with both sustainable and flexible computer environments.