Home | Contact | Sitemap | 中文 | CAS
About Us Research People International Cooperation News Education & Training Join Us Journals Papers Resources Links
Location: Home > News > Seminars
  • Events
  • Seminars
  • Research Trends
  • ARIMA, TCF and RF: A new statistical approach to exoplanet transit detection

    Seminar Title  

     ARIMA, TCF and RF: A new statistical approach to exoplanet transit detection

    Speaker:   ProfEric D. Feigelson


     (Penn State University)

    When Monday morning, July 9, 9:30 a.m.


    Room 302, No.3  building , Xianlin campus (PMO, CAS)

                             Welcome to Attend   

      ( PMO Academic Committee & Academic Circulating committee)

       Abstract: While photometric surveys of normal stars for exoplanet transits have revealed many candidate planets, the effort is limited by non-Gaussian noise. This typically arises from stellar magnetic activity for space-based missions (Kepler, Corot, TESS) and from atmospheric effects for ground-based projects (WASP, HAT, AST3-1). Most treatments of these extraneous noise use nonparametric techniques, but we have found that parametric autoregressive ARIMA models are often very effective. This talk reviews the fitting of ARIMA models to reduce autocorrelated noise, the development of a Transit Comb Filter to find periodic transits in the ARIMA residuals, and application of a Random Forest classifier to reduce False Positives. Application to the Kepler mission data set discovers several dozen new planetary candidates. Preliminary examination indicates the method should allow planet detection for densely-cadenced ground-based surveys like HAT-S and AST3-1. ARIMA modeling may also valuable for studies of autocorrelated astrophysical phenomena like stellar activity and systems where the light is dominated by accretion onto compact objects.

    Copyright © Purple Mountain Observatory, CAS, No.8 Yuanhua Road, Qixia District, Nanjing 210034, China
    Phone: 0086 25 8333 2000 Fax: 8333 2091 http://english.pmo.cas.cn