A-Technical-Review-of-Clustering

This project analyses different clustering methods over three different datasets

Functionality

1. Cleaning and Preprocessing of datasets: Plotted pairplots, heatmaps and histograms for the three datasets to pre-analyse the data and to identify which variables to exclude
2. Used the elbow method to use the K-means algorithm to analyse the three datasets
3. Plotted a dendrogram to use the Agglomerative Hierarchical algorithm to analyse the three datasets
4. Plotted a k-distance graph to use the DBScan algorithm to analyse the three datasets
5. Compared and evaluated the three methods to identify the advantages and disadvantages of using each clustering method

Documentation

All documents can be found under the Documentation folder

1. Detailed report of the project: A Technical Review of Clustering.pdf

Datasets

All datasets can be found under the Datasets folder

1. Frequent Flyer Program: This dataset contains information about the behaviour of NZ Airline’s FFP customers. We have dropped two variables: PartnerTrans and FlightTrans. The models are built with the following variables: AwardMiles, EliteMiles, PartnerMiles, FlyingReturnsMiles, and EnrollDuration.
2. Mall Customer: This dataset contains the basic information about the customers. None of the attributes has a good correlation among them and hence we used all the numeric variables when building the clustering models.
3. Wine: This dataset contains information about different types of wines. Total Phenols, Ravanoids, Hue, OD280 and Proline show a strong negative correlation with the class label. Ash_Alcanity has a positive correlation with Ash. Therefore, we dropped the variables - Ash_Alcanity, OD280, and Proanthocyanins- when we built the clustering models.

Developed by

1. Ang Shu Hui
2. Bachhas Nikita
3. Srinivas Shruthi
4. Unnikrishnan Malavika