Objectives: Data Mining using Unsupervised and Supervised Learning Approaches
Assume that a local company has collected a data set from their ecommerce website and ask you to analyze it. However, the company didn't provide much of background information about the data itself, e.g., the nature of attributes for the data set. However, based on the discussion with the people who collected the data and your observation on the data set, you felt that the first or second column, X1 or X2 may be decision column.
The basic strategy you will use is first to determine the decision column (or class attribute) using K-means clustering algorithm (unsupervised learning approach) to verify if the result of clustering is consistent with either attribute X1, X2, or both X1 and X2. Once the decision column(s) is determined, you build a model (or concepts) using supervised learning approach hoping that you will be able to offer an advice to the company for their business. To successfully complete the data analysis using this strategy, perform the following tasks:
(a) Use K-means algorithm (unsupervised learning) to cluster the data set and to verify the class field(s).
(b) Using the class field(s) determined in step (a), perform a supervised learning using any of those learning algorithms discussed in class such as Version Space, Decision Tree, and Neural Network, and build a model.
To perform above tasks, you are allowed to use either an existing system or program you implemented. However, in order to receive the maximum bonus points your program should work properly and must be powerful enough for effective data analysis. Otherwise, only a partial bonus point may be given. Therefore, it is more important to complete the above tasks (a) and (b) than implementing your own program.
Write a brief report that summarizes your data analysis activities and results including (1) your name(s) and contact email addresses; the percentage contribution to this assignment if the assignment was completed by a team. If a team cannot reach a consensus on the individual contribution, include the individual's claimed percent contribution with a brief description on specific tasks performed, (2) the language used for K-means algorithm implementation or the source of the software used, parameter settings such as K specifying how you determined the best K, clustering results, verified class field(s), and other relevant information to the task, (3) the name of the supervised learning algorithm used, the source of the implementation or software, parameter settings if any, the result of learning including the learned model and other relevant information, (4) the results of your data analysis, useful advice to the company's business, etc., and (5) other relevant discussion about your experience and data analysis results.