control the space

https://docs.google.com/presentation/d/11kUL3wIG8fiNASrQfuMQPZwKN-38_mTCeqIQQfUe-ls/edit?usp=sharing

https://github.com/Faris-ML/exoplanet-light-curve-detection

Project overview

The project is analysis light curves and predict if the star has s planet or not by machine learning algorithm

Methodology

The system takes a star light curve as input from the user and predict if this star has planet or not

the algorithms used is 1-D CNN but in the future, we will do some tuning and testing on other algorithms like boosting, SVM and logistic regression

Accuracy measure

the accuracy of the model trained on 25 epochs is 0.74 without hyperparameters tuning, we are planning to do some extra evaluation measures and testing, doing hyperparameters tuning and evaluate on several algorithms to get better performance.

tools and codes

programing language used is python with some important packages

the packages:

• tensorflow - for neural networks algorithms
• scikit-learn - for machine learning algorithm and evaluation metrics
• numpy - linear algebra operations
• pandas - working with tables
• ELCA - doing light curve analysis
• lightkurve - download light curves
• matplotlib - visualization

future plans

we are planning to make our project more than exoplanet detection, what we will do in the future :

• design UX/UI
• make a 3D visualization for the detected planet
• doing hyperparameters tunning to get better performance
• trying more algorithms
• add kepler project data to the training

goals

the project goal is to make exoplanet searching mission easier and get a full overview of space from light curves.

one of the project milestones is to get 3D visualization of what the planets and space look like with some prediction via machine learning

we only train our model on TESS project dataset; we retrieve time series light curves data with its tables by weget tool

we read .fits files and transform it into dump .pkl file, after that we build function to do some preprocessing on curves.

TESS data set contain a lot of information but for now we only use light curves to predict the exoplanet appearance.

and in the near future we will train it on kepler also.

Our experience from the Spaceapps Hackathon was quite an astonishing journey, we learned a great of different varieties of things concerning the space and planets which made us realize and dig deeper on the topic. 

 our topic( Exoplanet Detection ) was inspired by NASA as it’s a project for them, we thought of creating something that would help and develop more into this topic and it successfully worked for us, even though more time was needed but we did the basics and beginning to start the project and develop it with time, which was a great idea! 

Our team worked together into finding better solutions, we researched and looked into different aspects of things and we, indeed worked our way through it. 

We would like to thank NASA and our mentors who helped us by giving us advice into making it a better project!

Thank you very much

• Time-variable Electromagnetic Star–Planet Interaction: The TRAPPIST-1 System as an Exemplary Case by Christian Fischer and Joachim Saur
• https://github.com/pearsonkyle/Exoplanet-Artificial-Intelligence
• Exoplanet Detection using Machine Learning by Abhishek Malik, Benjamin P. Moster1 , Christian Obermeier1 
• https://www.nasa.gov/content/about-tess
• https://www.nasa.gov/kepler/education/getlightcurves
• https://docs.lightkurve.org/tutorials/1-getting-started/searching-for-data-products.html
• https://exoplanetarchive.ipac.caltech.edu/cgi-bin/TblView/nph-tblView?app=ExoTbls&config=TOI

#machine_learning #exoplanet #python #deeplearning

This project has been submitted for consideration during the Judging process.