Skip to content

README.md

Latest commit

 

History

History
44 lines (24 loc) · 2.18 KB

README.md

File metadata and controls

44 lines (24 loc) · 2.18 KB

Bayesian-Classifier

Bayesian Classifier

In this assignment, you are supposed to build the Bayesian classifiers for the datasets assigned to your group. This assignment is more focused on analysis of classification techniques and getting used to handling data in Machine Learning instead of getting classification accuracy (as you will be getting pretty well accuracy in most of the cases).

Bayesian Classifiers: Builds Bayesian model using the given dataset for the given case numbers.

CaseNumber: 1 -- Bayes with Covariance same for all classes

                   2 -- Bayes with Covariance different for all classes

                   3 -- Naive Bayes with C = \sigma^2*I.

                   4 -- Naive Bayes with C same for all classes.

                   5 -- Naive Bayes with C different for all classes.

Datasets Details: Dataset for each group can be found here. It also has the sample plots required.

  1. Linearly separable data:

2-dimensional artificial data for 3 classes is given for each group. The format is "groupNo_ls.txt" Each class has 500 data points (First 500 is class1, next 500 is class 2 and last 500 for class 3). Divide the points randomly for training(70%), validation(20%) and testing(10%)

  1. Non Linearly separable data: Again each group has 3 classes to work with. It is already divided into training, validation and testing data and kept in a folder named with GroupNo.

  2. Real Data: Real world data of 3 classes is given. The format is "group_GroupNo.txt". Divide the points randomly for training(70%), validation(20%) and testing(10%)

Plots Required (refer "Sample Plots" folder in the above link):

  1. Plot of PDF (Gaussians) for each of your class.
  2. Decision boundary and decision surface.
  3. Constant Density Curves and EigenVectors. (can be shown in plot1 or in plot2. No need for a separate plot. Refer images 1 and 2 in "Sample Plots" for details)
  4. Confusion matrices. (you may not show all in report)
  5. ROC and DET curves (they help you to compare your model). Plot them in a comparative way i.e the ROC for the models that you want to compare should be on the same plot. Similarly with DET. (Again, you may not show all of them in your report)