Assessed Lab ICE-4701 Weight: 20% of the total mark of the module. Total number of points: 100 Deadline: 13th March, 8pm. The submission point will close automatically thereafter. Submission: Submit one pdf file with your Lab report. List in the report your solutions including code, comments and output. Notes: Plagiarism: DO NOT copy from one another. Changing the variable names does not make the code your own! A mark of zero will be awarded for assignments which are too similar. I reserve the right to interview you about your submission and change your mark if I discover that you cannot explain your code in enough detail. 1. Your data. 1 (a) Write code (preferably in Python) to upload your data, and create array “data” and array “labels” from it. [5] 1 (b) Format and print the characteristics of the data: N, n, c, the number of objects in each class. Estimate the prior probabilities for the classes. [10] 1 (c) Prepare and show a scatterplot of the data. An example is shown in Figure 1. [6] Figure 1. Example of an expected plot for 1 (c) (Note that you all have different data sets!) 2. Classification For an example on classification using Python, look up this link: https://www.kaggle.com/abhikaggle8/pima-diabetes-classification 2 (a) Write code to apply the Largest Prior classifier to your data set. Calculate and show the error rate of this classifier. Prepare and display the respective confusion matrix. Format and include the confusion matrix in your report. [15] 2 (b) Apply the decision tree classifier to this data set using a 10-fold cross-validation. Show the confusion matrix in your report. [10] 2 (c) Plot the classification regions for the decision tree classifier. An example of the expected outcome is shown in Figure 2. [10] Figure 2. Example of an expected plot for 2 (c) 2 (d) Using Weka, run a 10-fold cross-validation on your data for the following classifiers: [15] • 1 NN (nearest neighbour) • 3 NN (3- nearest neighbours) • Decision tree • Bagging ensemble • AdaBoost ensemble • Random Forest ensemble • Rotation forest ensemble Prepare a table with the classification accuracies of the classifiers. Give a comment on the results. 3. Classifier ensembles You can solve this question by hand, using Excel or writing a piece of code. In all cases, show your work/spread sheet/ code. Take the first 5 points in your data set. Run them through a classifier ensemble consisting of the following three classifiers, each one shown with its discriminant functions: Classifier 1 Classifier 2 Classifier 3 1() + 2 − 0.1 0.02 2() − 2 −2 4 − 0.3 3() + − 0.4 2 + 0.1 ( + )2 4() 2 − − 0.1 ( − )2 3 (a) Calculate the majority vote label for each of the 5 objects, then prepare and show the confusion matrix. [18] 3(b) Calculate the class labels for the 5 objects using the following combination rules: [11] • Average • Minimum • Maximum
