For this assignment, find a suitable dataset on which you can apply various classification algorithms.
(1) A description of your dataset: what are the attributes, what are the targets, how many datapoints do you have, and some sample datapoints from the dataset. State what you are trying to predict with the data.
Note: marks will be given for an interesting choice of dataset.
(2) A description of how you structured your inputs/targets and normalised and preprocessed the data, and the split into training/validation/test data.
(3) A list of classification algorithms you applied to the data, together with the details of each implementation and the error on the test set. Also provide details on how and why you selected the hyperparameters you did. Present the errors at least in the form of a confusion matrix.
For example, if you used regularised linear regression:
• Why did you choose this algorithm, and why did you add regularisation?
• What value of λ did you use? Why was this a good choice (with evidence)?
• What basis functions did you use? Why?
• How did you train the model? e.g. gradient descent with α = 0.2. Why did you choose this?
(4) A brief discussion of your results from the various algorithms. E.g., what worked best/worst and why you think this is so. What is the best possible performance you can achieve on this dataset? How did you do that? What would you recommend someone else try if they were interested in working with this data?
