Original repository https://github.com/modulus100/cluster-analysis-R/
This project is intended for learning a fundamentals of machine learning and data mining
algorithms. All the code is written using a pure R programming language.
The project contains an implementation of the following algorithms.
- Mahalanobis distance
- Minkowski distance
- Canberra distances
- k-means algorithm
- k nearest neighbor’s algorithm
The kNN_data1.RData file contains all the required data to run algorithms. Please specify your path for kNN_data1.RData in Exercise1.R before you run the project.
load(file="/yourPath/kNN_data1.Rdata")
Distances
mahalDist <- mahalanobisDistance(x[2,][1:2],x[4,][1:2], x[,-dimension])
minkovskiDist <- minkovskiDistance(x[2,],x[4,], 5)
canderraDist <- canderraDistance(x[2,],x[4,])
Kmeans, x - matrix with the data, k - number of clusters, distance method - name of a distance method ( Minkovski is default)
kmeansData <- kmeansAdvanced(x, k, distanceMethod)
plotKmeansData(kmeansData)
silhoutte ratio x - matrix with the data, k - number of clusters, distance method - name of a distance method ( Minkovski is default)
ratio <- silhouetteRatio(x, k, distanceMethod)
kNN classification, x - matrix with the data, kmeansData$dataset should be used, x[100,] - vector or a point, k - number of nearest points, distance method - name of a distance method ( Minkovski is default)
nearestVectors <- kNNClassificationAdvanced(x, x[100,], k, distanceMethod)
classificator <- fetchClassificator(nearestVectors, k)
computeNewCentroids - recalculates and returns new centroids whithin each iteration during the kmeans computing
covarMatrix - computes a covariation matrix
plotKmeansData - plots the kmeans data
kNNClassificationAdvanced - returns kNN k vectors
fetchClassificator - returns a classificator according to kNNClassificationAdvanced vectors
silhouetteRatio - silhouetteRatio
https://www.youtube.com/watch?v=4HKqjENq9OU
https://stackoverflow.com/questions/38173136/euclidean-distance-for-three-or-more-vectors
https://stackoverflow.com/questions/30195806/k-means-stopping-criteria-in-matlab
https://nlp.stanford.edu/IR-book/html/htmledition/k-means-1.html
https://kevinzakka.github.io/2016/07/13/k-nearest-neighbor/
https://habr.com/post/199060/
https://habr.com/post/101338/
https://www.coursera.org/lecture/cluster-analysis/6-2-clustering-evaluation-measuring-clustering-quality-RJJfM
