Dds Efs
Premium Setting and Claim Prediction for Motor Insurance 1. Go to the assignments page on the course web site and down load the two .csvfiles for assignment 2. Store them in your own directory 2. Have a quick look at them in Excel. You will see that they both contain data describing motor insurance policies. a. The fileMotorPremiums.csv contains policy details and the premiums charged on those policies. b. The file MotorClaims.csv contains the same policies with a field indicating whether or not a claim was made. 3. Put thepremiums and the claims columns together in a new file and use a chart to see how well the premiums predict claims. 4. Your first task is to use the Premiums file to reproduce the insurance companies ownset of pricing rules using the examples of prices in this file. Use a multilayer perceptron (MLP) in Weka with 50% test data. How well does the MLP reproduce the prices in the test data? 5. Next youneed to use the claims data and produce a better pricing system. a. Using Weka, build a model that classifies claims as Yes or No. Try a few models including an MLP and an ID3 tree. b. When you have amodel you like, verify it using cross validation. c. Look at the confusion matrix for your model – what is it telling you? 6. Now we are going to look at lift curves. Right click on your chosen modelin the result list and choose Visualize Threshold Curve from the pop-up menu. Below this option will be two more – Yes and No. These are the names of your output class values. Choose ‘Yes’. a. Thewindow you get allows you to plot various results from your test data to see how good the model is. You can vary what is plotted on the X and Y axis and also what the colour of the points tells you. Thedata is sorted by the probability of the output value you chose (‘Yes’ in this case). So, instances that are most likely to be Claim = Yes are plotted at the left hand side of the curve. The curves...
Regístrate para leer el documento completo.