Academic Performance Prediction Model For Deaf Students Using An Enhanced Pso Algorithm And Pattern Discovery Approaches

Abstract

From census 2011, it is found out that literacy rate of deaf students is very less in newlineIndia and only 26% of deaf student acquire a graduate degree. Prediction is an newlineadvanced form of data analysis. Prediction is forecasting the future, with historic newlinedata in hand. Machine Learning is an emerging field of computer science which newlinespecializes in building advanced prediction model. Recently, many researchers newlineall around the world have developed many customized prediction model, fitting newlinetheir research problem in major domains like health care, finance, newlinetelecommunication, marketing. In educational domain, there are prediction newlinemodels, which predict whether a student will pass the exam or not. Some models, newlinepredicts the final exam marks of students .In this research, an academic newlineprediction model for deaf students is developed. Many machine learning newlinetechniques are used to build prediction models. One among them is the Neural newlineNetwork. Neural Network are biologically inspired computing network, which newlinemimics the activities of the human brain. In Neural Network, inputs are matched newlinewith output and models are built. Weights are adjusted in neural network to newlineproduce more accurate results. Though the Neural Networks are the most newlinepreferred technique for prediction. It suffers problems like high error rate and newlineslow convergence. One of the reason for high error rate and slow convergence is newlinerandom initialization of weights in the network. As the weights are initialized newlinerandomly, neural network takes more time to converge. Many optimization newlinetechniques are used to solve the problems of the neural network. One among newlinethem is Particle Swarm Optimization(PSO). In particle swarm optimization newlinealgorithms, particles are used to initialize the weights of the network. But the newlineproblem with PSO algorithm is that if the particle are initialized with incorrect newlinevalues, it will take more time to converge. In this research, a regression equation newlineis derived to initialize the particles in the PSO algorithm. The developed newlineRegression Based PSO algorithm (RBPSO)