Describe, with motivations, any data pre$-$processing that you have applied to get the data ready for SOM training.

Find the architecture and parameterization of the SOM that provides you with the best possible feature

map. In your document provide detail on the performance measure$($s$)$ that you have used to determine

the best SOM, as well as the process that you have followed to decide on the best SOM configuration.

Provide full detail on the selected architecture and parameterization of the SOM. $(20)$

he next part of the assignment is the most important part, and will test your ability to explore relationships among the features of this data set. Provide descriptive statistics for the differmarksent clusters

in your feature map. Use these descriptive statistics and the component maps to identify patterns from

the data. In your pdf document, present and discuss all of the patterns that you can identify. Provide

motivations for these patterns, referring to the descriptive statistics and component maps. As a final step,

indicate if any of the included features can be considered irrelevant or redundant. $(50)$

Now, use the code vectors as input to a rule induction algorithm and extract rules to describe each cluster

produced by the SOM. Label each cluster with the class value that occurs most frequently among the

instances that are assigned to each cluster. Provide the extracted rules, with performance measures per

rule and for the entire rule set. $(20)$

$$ For the final part of the assignment, you will compare the quality of the rules extracted from the step

above with rules extracted from a classification tree and any rule induction algorithm of your choice. In

your report, please provide details on the classification tree induction algorithm and the rule inducation

algorithm that you have selected. Discuss and data$-$preprocessing that you had to apply for the selected

algorithms. Provide the performance measures that you will use to compare the rules extracted from the

SOM, the classification tree, and the rule induction algorithm. Present your results and discuss these

results, working towards a conclusion on which approach provided the best rules. $(50)$