Consider the points/classes below and the perceptron algorithm taught in class.
Class 0:
x1 = (-2, 1)
x2 = (1,3)
Class 1:
x3 = (2,0)
X4 = (2,2)
(a) what is the equation of that separates the two classes
(b) graph this equation
Consider the same initial problem as described in (1) above BUT change the order the examples are presented to the perceptron to the follwoing order:
x1,x3,x2,x4
(a) what is the equation of that separates the two classes
(b) graph this equation
(c) give a single NEW point that causes the perceptron NOT TO CONVERGE and clearly show this point on the graph in (b) above
Consider the points/classes below and the perceptron algorithm taught in class.
Class 0:
x1 = (0 ,0,-4,1)
x2 = (2,3,-4,1)
x3 =(12,14,-4,1)
Class 1:
x4 = ( 0 ,5, 1, 2)
x5 = (2,3,5,1)
x6 = (12,14,5,1)
(a) what is the equation of that separates the two classes
4. Select the statement that best describes BAYESIAN reasoning that I emphasized in class: BAYESIAN REASONING
(a) works with 2 or more classes
(b) can accommodate missing or partial information
(c) runs much faster than a perceptron
(d) needs much fewer examples than the J48 algorithm
5. A person has swollen glands represented as symptom G. There are three possible diseases this person has - disease A, disease B , or disease C.
The relevant probabilities are given below:
p(A) = 0.35
p(B) = 0.35
p(C) = 0.30
the conditional probabilities are as follows:
P(G|A) = 0.015
P(G|B) = 0.010
p(G|C) = 0.020
What is the probability the patient has
(a) Disease A
(b) Disease B
6. Use the data in the exam folder entitled credit-g and Weka. Under the classifier, select BAYES and NAIVE BAYES with percentage split being 0.66 (i.e., 2/3). Run the naive bayes classifier:
(a) What is the recall?
(b) What is the precision?
(c) do the identical problem EXCEPT using the j.48 classifier with the same data and percentage split. What is the recall now?
(d) What is the precision?
The problems under 7 below are real. Don't expect all kinds of clean answers and obvious explanations. There may be missing data and techniques like ID3 may actually run out of attributes and NEVER create a single homogeneous classification. Be Careful. Here's how this will be graded: There is no "right" answer I am seeking. I don't have a 'key' for problem 1A and you either get my answer of you miss it. Rather, I am going to grade HOW you went about solving the problem and whether what you did is reasonable. If you blindly go applying any technique without any type of analysis - that's reckless in real life as well as the final. Instead, you should perhaps spend some time thinking about the strengths and weaknesses of each technique and the problem domain itself before rushing off to implement. For example, what is the cost of missed detections? What about false alarms? DO NOT USE ALL THE DATA TO TRAIN THE CLASSIFIER - instead use 2/3 to train and 1/3 for testing! Think in terms of training and test sets. DON'T just run your techniques on the entire dataset. Instead, when it makes sense to you, divide your dataset into two pieces (not necessarily equal) and SEE FOR YOURSELF HOW YOU'RE PERFORMING. If you train on one set of data and then test on another - and you correctly classify the test set -> that indicates a high confidence in the result. It's hard to criticize that type of method. Also think about the results? When are the results acceptable? When are the results possibly really misleading? If use use J48 - just submit your interpretation of the tree. We'll forgo implementation in CLIPS for another class. Otherwise - just use WEKA or your own perceptron code to create a classifier.
7. SELECT TWO PROBLEMS from the problems below (A,B,C, etc) to work on.
(A) Wine classification: use the WINE DATA in the link on the DATA FOLDER and attempt to classify wine origin based on its chemical properties.
(B) Here is a typical problem dealing with a real corporation and real people. The problem is credit approval. A Japanese company uses a complicated technique to approve credit and want to see if they can use some data mining technique to classify people as + (grant credit) or - (deny credit). So they turn over a portion of the database to you with all the attributes "coded" to protect anonymity. In other words, they may change something like the person's job title to "a", instead of "bank clerk". They're consistent in that all 'a's" are bank clerks but all that appears in the data for the 'occupation' attribute is "a", "b", etc. Worse, they won't even tell you WHAT the real attributes are. ALl you know is that + meant they approved, and - meant they denied. So your job is to see whether you can do to predict a + or - and how well you can do that. A solution to this problem is just going to be a technique and an estimate of how accurate it is. You won't be able to make much domain sense of your answer by saying something like "if occupation is bank clerk, then deny". SEE THE CREDIT SCREENING DATA in the data link given in the Data Folder under CREDIT APPROVAL.
(C) Assume a corporation needs to know the yearly income of a customer and does not want to annoy the customer by asking this question directly. Instead, the desire is to ask the consumer some "innocent" questions that can predict yearly income. The corporation gives you US Census data and asks you if this can be used to classify people as to whether they earn above or below $50,000 yearly. What are some questions that might be asked given this data? Data is found here under ADULT DATABASE in the data link in the DATA FOLDER: Assume your job is to come up with an automatic way to classify consumers if given the same data as appears in the data file. Justify and test your approach. What happened? How accurate were you?
(D) What can you tell me about the MUSHROOM DATA given in the data link in the DATA Folder. If you can build a classifier - do that. Do whatever you can with this data and tell me what you did and why. Justify your results and approach.
(E) Arrythmia is a heart ailment there is a database of people who either are normal or suffer from some type of arrhythmia given on the data link in the DATA FOLDER. Assume your job is to come up with an automatic way to classify patients if given the same data as appears in the data file. Justify and test your approach. What happened? How accurate were you?
(F) Select any data set from the provided data folder link and apply any technique you wish or a mix of techniques to see how they agree (e.g., J.48 and Bayes) Your choice - whatever interests you.
(G) Solve using any technique you wish but JUSTIFY your choice in a paragraph explaining why you selected that technique and how you tested the classifier!
Use the LUNG CANCER DATASET (Attribute 1 is the class variable). NOTE THAT THERE ARE MISSING VALUES IN THIS DATA!!!!
To receive credit you must present the classification results