Stars Over Time

A star that changes with the passage of time is a self-illuminating gas ball that is bound by a single object by gravity and is moved by nuclear fusion, which is the core. There are trillions of stars in our universe, but they are all unique and unique. There are many stages of stellar life, including main sequence stars, red giants, white dwarf stars, neutron stars, and black holes. Every star has more change at every stage of life. The star's life begins with a gathering of nebula, huge gas and dust. When sufficient mass is accumulated in a single object, the mass forces it to fall to the center by gravity.

In this article, our goal is to create a model to predict the existence of an extrasolar planet (that is, a planet orbiting a remote star system) by giving a reading of the light intensity of a certain star over time . The dataset we use comes from NASA's current Kepler telescope in outer space. We will step through from low performance model to high performance model. The Kepler telescope was launched by NASA in 2009. Its mission is to discover that the earth is like a planet orbiting other stars outside the solar system. Kaggle published a data set containing clear observations / readings from Kepler to find outer planets orbiting other stars (planets outside the solar system).

This is how it works. Kepler observed thousands of stars and recorded the intensity (light flux) of the light emitted from that star. When the planet circles the stars, it slightly modifies / decreases the intensity of the light. As time goes by, you can see that the star's line darkens regularly (like t = 2 in the figure below), which suggests there is a planet orbiting around the planet (candidate system ). Further light studies can confirm the existence of extrasolar planets on the candidate system. Firstly, there are too few examples of extrasolar planets identified in our data. There are several techniques that help overcome highly unbalanced data sets and synthesize or create new examples. What I use here is an algorithm called SMOTE (synthesis of several oversampling methods). Instead of creating a sample copy, the algorithm creates a new example by slightly modifying the existing example. In this way you can balance the positive and negative examples in the training data set.

Gaia includes three devices to observe stars. These instruments determine the position, color, and movement of stars over time. Gaia uses parallax measurements to calculate the distance of the star that the astronomer considers to be the most important data point. Gaia does this by examining the apparent movement of the star at two different points on its orbit and separating it from the actual movement of the galaxy. The spacecraft watches the goal stars for about 70 times in 5 years. This is the planned duty period. As we are about to finish the duty, Gaia sent back a lot of data.