Unsupervised Learning - Kmeans Week 11 Challenge
Answer no.1
When dealing with categorical data in clustering algorithms like K-means, you need to use a similarity or distance measure that is appropriate for categorical variables. Two common distance measures for categorical data are the Hamming distance and the Jaccard similarity coefficient.
-
Hamming Distance: The Hamming distance is used to measure the dissimilarity between two categorical variables. It calculates the number of positions at which two categorical variables differ. For example, if you have two variables 'A' and 'B', the Hamming distance between them would be the number of positions where 'A' and 'B' have different values.
-
Jaccard Similarity Coefficient: The Jaccard similarity coefficient measures the similarity between two sets. In the context of clustering categorical data, each variable is treated as a set of categories. The Jaccard similarity coefficient is calculated as the intersection of the two sets divided by the union of the sets. It provides a measure of the proportion of shared categories between two variables.
These similarity measures can be used in K-means clustering for categorical data by defining appropriate distance metrics. In some cases, you may need to convert categorical variables into numerical representations before applying these distance measures. One common approach is one-hot encoding, where each category is represented by a binary variable (0 or 1).
It's worth noting that K-means is primarily designed for numerical data, and applying it directly to categorical data may not always yield meaningful results. There are other clustering algorithms specifically designed for categorical data, such as k-modes and k-prototypes algorithms, which can handle categorical variables more effectively.
When working with categorical data in clustering, it is important to carefully consider the specific characteristics of your data and choose an appropriate similarity measure or clustering algorithm accordingly.
Answer no.2 submitted in .ipynb form