Prediction of Insufficient Accuracy for Human Activity Recognition using Convolutional Neural Network in Compared with Support Vector Machine

Abstract

This work seeks to anticipate human action detection and assess its health in machine learning algorithms in order to increase accuracy and detect human activity. The framework categories Support Vector Machines (SVM) and Novel Convolutional Neural Networks (CNN) to achieve these goals (SVM). A wholly unique Convolutional Neural Network and a Support Vector Machine were used in this analytical investigation to finish the processes and give the best accurate human action recognition for neighboring video frames. The analysis study collected forty seven samples with 2 teams of calculation with an eighty p.c G-power and their activities were no heritable from varied on-line sources with the foremost recent study findings and 0.05 p.c threshold, confidence interval ninety five p.c mean and variance. To forecast the activity analysis, as a result of the novel Convolutional Neural Network formula discovered 93.83 p.c accuracy and also the native invariant approaches, this study needs bigger accuracy for activity prediction with the Support Vector Machine formula in machine learning. This study discovered 90.12 p.c accuracy for human action analysis utilizing the Novel Convolutional Neural Network technique, with a major price of 2 caudate tests of 0.013 (p0.05) and a ninety five p.c confidence interval. This study shows that the Convolutional Neural Network technique performs far better on human action analysis than the Support Vector Machine technique.