/linear-classifier-with-computational-geometry

A 2D linear classifier that wraps each group of points in a tight shape and then draws a separating line

Primary LanguageJupyter NotebookMIT LicenseMIT

Practical Project 01 | DCC207 - Algorithms 2 | October 2023, Federal University of Minas Gerais

Doing Linear Classification with Computational Geometry

This project aims to use computational geometry algorithms to create a linear classification model, which uses a straight line to classify data points into two categories.

Implementation objectives

  • A Convex Hull algorithm
  • A Sweep-Line Segment Intersection algorithm
  • A Closest Pair of Points algorithm
  • Find the classification line
  • Evaluate with metrics 10 datasets from online databases

Steps: Starting from a training dataset, the algorithm must find an envelope (boundary) that delimits a classification region for a dataset (points in the Cartesian plane). From two of these envelopes, it will be possible to determine if there is separability (or if they have overlapping areas) and then determine a line that separates the two datasets.

Students: Juan Braga, Lucas Almeida, e Luiz Romanhol