Machine Learning applied to Astroinformatics

Inspect a large astronomical data frame, determine its shape, and identify distinct targets. Filter by object id and plot flux versus time (mjd) to visualize a light curve.

Select the columns to use, create a new data frame with six features (object id, mjd, flux, flux error, and target), and verify the data shape for downstream machine learning.

Iterate over each object id to extract features from one light curve using the Fitz library, filtering by object id and bandwidth, then compute nine features.

Create a dataframe from computed astro features by concatenating arrays into a features set with target. Name the columns and verify shape to support time-series feature extraction for astro objects.

Train a supervised logistic regression classifier with two features (amplitude and another feature) to classify objects into classes, and evaluate with F1 score before predicting new data.

Expand logistic regression to multiple features in a supervised classification setup, training a model with additional features and evaluating with F1 score and accuracy to compare against a two-feature baseline.

Apply unsupervised learning with principal component analysis to reduce dimensionality from nine dimensions to three, transforming data into a PCA space and evaluating explained variance.

Apply logistic regression to PCA-reduced data for supervised classification in astroinformatics, using principal component data to train and evaluate with f1 score and accuracy, enabling visualization.

Build a concise report that communicates principal insights from data analysis in astroinformatics, comparing supervised and unsupervised learning on over seven thousand light curves with nine features.