Based on the data provided, the report deals with simplifying the dataset for further analysis through clustering, specifically using the K-means algorithm. Since the algorithm requires a user-input number of clusters, clustering using a range of number of clusters is done to compare each of its sum of squares to find the optimal K number of clusters.
- Multivariate statistics
- K-means clustering
- Sum of squares(SS) analysis
- LateX
- R
The data used for this project can be obtained from Palmer Penguins with 4 variables, bill_length_mm; bill_depth_mm; flipper_length_mm; and body_mass_g. With the n = 333 observations, the first section explores the dataset through descriptive statistics to determine the user-input K number of clusters. The covariance matrix is also obtained and studied to scale the data before the clustering can be done.
The second section explains the methods used to cluster the data, through an overview of the algorithm, as well as the commands used in R. How the optimal number of clusters is recognised is further detailed as reference for the dicussion of this report.
The third and final section of the report discusses the results of the methods in the previous section, using user-input K = 3 which is chosen when analysing the scatterplot of the variables against one another. The counts of the clustering done is compared to two of the known clusters provided in the dataset, L.species; L.islands, to see how the variables affect the clustering. These counts are then visualised through scatterplots of two of the variables, where the different colours indicate the K-means clusters while the point characters separate the data points by its known clusters. It's found that the clustering done matches up more with the species classification due to the smaller number of mislassified samples in comparison to the island grouping. The discussion is then concluded with plot of the within group and between group variation against number of clusters, telling us that the chosen K is the optimal number of clusters.
The final report can be read here.
- Machine learning
- Writeup
Nurfahimah Mohd Ghazali