Aug 2020 - Nov 2020
In the outer space, there is a special kind of neutron star called "Pulsar" that can rotate extremely fast. The radiation of the pulsar can be observed by large radio telescopes when its emission beam points to Earth ('Pulsar' 2020). Pulsars play an important role for astronomers to explore the universe, including interstellar medium and space-time investigation (UCI 2020). Researchers have detected large amount of radio signal through the years, however, most of them are noise or radio frequency interference. According to the pulsar's special signal emission pattern, researchers can identify whether the signal they observed is produced by the pulsar. Because of the large amount of noise, it is hard to find a real pulsar efficiently.
The goal of this project is to classify candidates and identify which candidate has represented as a real pulsar on its given features automatically by using machine learning tools.
- Stage 1 --> Exploring, summarising and preparing the data
- Stage 2 --> • Data Pre-processing • Classification models implementation & Parameters tuning (SVM, KNN, Logistic Regression, Random Forest and NaiveBayes) • Hypothesis Testing (Kruskal-Wallis H-test)
The dataset is downloaded from UCI website (https:// archive.ics.uci.edu/ml/datasets/HTRU2).