Regression Analysis of Solar Flares: A Multilayer Perceptron Approach with Feature Selection Techniques
International Journal of Computers and Communications · 2020
In this paper, we are going to analyze and test the solar flare dataset from the UCI Machine Learning Repository [10], by improving it using feature selection techniques such as stepwise regression, detecting the most effective attributes, importance ranker using k-fold and leave-one-out cross validation methods. We are going test the model by evaluating the dataset using Multi-linear regression, by looking at the P-values and the VIF to show the effectiveness of the dataset attributes. Multilayer perceptron model will be created using the holdout regression by partitioning the dataset into training and testing to model the testing dataset into an MLP model with 5 hidden layers. The model will show the mean absolute error and variance of the model to test its accuracy.