Detecting Exoplanets from light curves of Kepler Mission using Fast Fourier Transform and Recurrence Plots

The quest for finding life and habitat in the universe other than earth has been going on since long. To uncover it National Aeronautics and Space Administration (NASA) has Kepler Mission which captured data of brightness of stars which are called light curves. These are formed when a star transits from one place to another its host star. This star could be an exoplanet and this is decided by checking its brightness. This data is of a period of time and is in time series format. Using this data, Support Vector Machine is applied first. To improve the performance of models this research uses pre-processing techniques like Fast Fourier Transform and Recurrence plots are applied before feeding the data to any models. Then models like Support Vector Machine and Convolutional Neural Network with VGG16 are used. Different patterns were plotted and time series data was analysed to check if the data point is exoplanet or not. The results for with and without pre processing techniques are compared and they show how FFT and RP make a difference in getting improved performance in terms of accuracy, precision, F1 score, etc