By Mohamed Ehab Sabry

feel free to reach me out on LinkedIn

www.linkedin.com

k-NN in short

• non-parametric to know what does that mean, read this :

Parametric vs Nonparametric Machine Learning Algorithms

• uses proximity
• typically used as a classification algorithm
• it works based on the assumption that similar points can be found near one another .
• in classification, it depends on “plurality voting” not majority vote, as majority vote requires a +50% of the votes, which is not necessarily the case when you have multiple classes, you can help yourself out with less than that.
• we must choose the way to calculate the distance between points at first, there are many, the most widely used is Euclidian Distance.
• it is part of lazy learning models ( just stores a training dataset, doesn’t actually train anything)
• Also called instance-based or memory-based learning method (since it relies heavily on memory to store all it training data)
• used in:
  • simple recommendation systems
  • pattern recognition
  • data mining
  • financial market predictions
  • intrusion detection كشف اختراقات المواقع
  • more
• avoid it when:
  • dataset is becoming “huge”
    • because this will make it more and more inefficient

Compute k-NN: distance metrics

Determine your distance metrics:

عشان نحسب المسافه، محتاجين نحدد هنحسبها ازاي، تحديدها بيساعدنا نحدد ال Decision boundaries، والي معناه لو النقطة دي في أني مكان بظبط تبقا تبع أني كلاس ؟ ودي بقدر أعملها visualization عن طريق Voronoi diagrams