Predictive Techniques for Astrophysics

Primarily applied pandas, and scikit-learn tools on JINAbase to do:

• regression/imputation - predict missing microturbulence values
• classification - classify stars based on stellar types

Report of this project can be found here. Requirements can be found in the YAML file here

Exploratory Data Analysis

    

    

Performance

Regression

    

    

Classification

Random Forest Classifier Class Precision Recall F1-Score Support HB 1.00 1.00 1.00 23 MS 1.00 1.00 1.00 55 RG 1.00 0.99 1.00 227 SG 0.97 1.00 0.98 63 Accuracy 0.99 368 Macro Avg 0.99 1.00 0.99 368 Weight Avg 0.99 0.99 0.99 368

Support Vector Classifier Class Precision Recall F1-Score Support HB 0.95 0.91 0.93 23 MS 1.00 0.96 0.98 55 RG 0.99 0.99 0.99 227 SG 0.94 1.00 0.97 63 Accuracy 0.98 368 Macro Avg 0.97 0.97 0.97 368 Weight Avg 0.98 0.98 0.98 368

KNN Classifier Class Precision Recall F1-Score Support HB 1.00 0.83 0.90 23 MS 1.00 0.87 0.93 55 RG 0.99 0.99 0.99 227 SG 0.86 1.00 0.93 63 Accuracy 0.96 368 Macro Avg 0.96 0.92 0.94 368 Weight Avg 0.97 0.96 0.96 368

XGBoost Classifier Class Precision Recall F1-Score Support HB 1.00 1.00 1.00 23 MS 1.00 0.96 0.98 55 RG 1.00 0.99 1.00 227 SG 0.94 1.00 0.97 63 Accuracy 0.99 368 Macro Avg 0.99 0.99 0.99 368 Weight Avg 0.99 0.99 0.99 368

MLP Classifier Class Precision Recall F1-Score Support HB 0.96 0.96 0.96 23 MS 1.00 1.00 1.00 55 RG 1.00 0.99 0.99 227 SG 0.97 1.00 0.98 63 Accuracy 0.99 368 Macro Avg 0.98 0.99 0.98 368 Weight Avg 0.99 0.99 0.99 368

Predictions

Interactive 3D plot can be found here.

References

[1] Trevor Hastie, Robert Tibshirani, and Jerome Friedman. Random Forests, pages 587–604. Springer New York, New York, NY, 2009.

[2] Support vector machines — scikit-learn 1.2.2 documentation. https://scikit-learn.org/stable/modules/svm.html, 2023. Accessed:17-04-2023.

[3] Cosma Shalizi. CMU Statistics 36-462/662: Methods of Statistical Learning. Lecture Notes 11: k-Nearest Neighbors, 2022. URL: https://www.stat.cmu.edu/˜cshalizi/dm/22/lectures/11/lecture-11.pdf.

[4] Introduction to boosted trees — XGBoost documentation. https://xgboost.readthedocs.io/en/latest/tutorials/model.html#introduction-to-boosted-trees, 2022. Accessed: 15-04-2023.

[5] Roger Grosse. University of Toronto CSC 411: Machine Learning and Data Mining. Lecture 5: Multilayer Perceptrons, 2019. URL: https://www.cs.toronto.edu/˜mren/teach/csc411_19s/lec/lec10_notes1.pdf.

[6] Abdu Abohalima and Anna Frebel. Jinabase—a database for chemical abundances of metal-poor stars. The Astrophysical Journal Supplement Series, 238(2):36, oct 2018.

[7] Laurens van der Maaten and Geoffrey Hinton. Visualizing data using t-sne. Journal of Machine Learning Research, 9(86):2579–2605, 2008.

[8] C. Chatfield. Problem Solving: A statistician’s guide, Second edition. Chapman & Hall/CRC Texts in Statistical Science. Taylor & Francis, 1995.