Astronomers are interested in finding planets on Earth of similar sizes, compositions and temperatures. Earth-like planet. However, this effort has its challenges. That's a small, rocky planet It's difficult to find This is because current methods of planet hunting are biased towards gas majors. Additionally, for a planet to have similar temperatures to Earth, it must orbit an equal distance from the host star, just like Earth orbiting the Sun. In other words, it takes about a year to traverse the stars. This raises another problem for astronomers trying to find these planets. Because searching for planets like Earth around a star involves dedicating telescopes to constantly monitor them for more than a year.
To save time spent on commercial driving surgeries, scientists need a new way to find stars, a good candidate for a thorough search, before dedicating resources to observe them. A team of astronomers investigated whether observable properties of planetary systems can demonstrate the existence of planets like Earth. They discovered that the arrangement of known planets within the system, along with the mass, radius, and distance from their nearest planet to the star, can be used to predict the occurrence of planets like Earth.
How well did the team test? Machine Learning You can handle this task. They started by creating a sample set of planetary systems with or without planets like Earth. Astronomers have only found about 5,000 stars in the sky with orbits in orbit, so the sample size is too small to train machine learning programs. So the team generated three sets of planetary systems using a computational framework that simulates how planets form. Bern model.
The Bern model starts with 20 dust clumps, 600 meters or about 2,000 feet. These masses kickstart gas and dust and accumulate on full-size planets for over 20 million years. The planetary system has since evolved into a termination state for over 10 billion years, Synthetic Planetary Systemwhich astronomers should include in the dataset. They use this model to create a 24,365 system with sun-sized stars, 14,559 system with sun-sized stars, and 14,958 system with stars. They also divide each of these groups into two subgroups, including groups with planets like Earth and groups without planets like Earth.
Using these larger datasets, the team tested whether machine learning techniques were called Random Forest Model Planetary systems can be categorized as planets that probably have planets like Earth and those that have planets that do not. In a random forest, all outputs are either true or false, and various parts of the program are called. tree,determine the various subsections of the entire training dataset. The team decided that if the planetary system could have one or more Earth-like planets, then Random Forest should be considered “true.” Researchers tested the algorithm for accuracy using a metric known as A Precision score.
They set up random forests to make decisions based on specific factors in each synthetic planetary system. These factors included the number of planets, if astronomers saw similar real systems, the number of planets in the system, whether planets more than 100 times the mass of Earth have a size and distance to the stars, whether stars have a size and distance, and astronomer placement. The team used 80% of the synthetic planetary system as training data and reserved the remaining 20% for the initial testing of the completed algorithm.
The team found that random forest models predict where planets like Earth are likely to exist with a precision score of 0.99. Following this success, they tested a model of the actual data of 1,567 stars in a similar size range where at least one planet is known to orbit them. Of these, 44 passed the algorithm threshold to have planets like Earth. The team proposed that most systems in this subset would not fall apart if there were planets like Earth.
The team concluded that their models could identify candidate stars for planets like Earth, but there is a warning. One is that the production of synthetic planetary systems is long and expensive, so training data is still limited. But the bigger warning was that they assumed that the Bern model had accurately simulated the planetary layers. They proposed that researchers rigorously test their validity for future theoretical research.
Post view: 43
