Skip to content

Latest commit

 

History

History
39 lines (28 loc) · 2.81 KB

README.md

File metadata and controls

39 lines (28 loc) · 2.81 KB
Raw

CBA

This is my implementation of algorithm CBA (Classification Based on Associations). It's my final project of data mining course, taught by Prof. Chen Lin. It's written by Python 3.6. You can read the paper Integrating Classification and Association Rule Mining by Bing Liu et al. for details.

What's in datasets?

The datasets are chosen from UCI Machine Learning Repository and processed simply. We download 30 datasets including some classical datasets like iris.

Each dataset has 2 files:

  • *.data contains many instances. Each line represents a sample. The attributes of the instance are divided by comma mark (,) without the space ( ). The last attribute is the class label, both number or literal string are vaild.
  • *.names contains two lines. The first line is title line, describing the name of each attribute. The last word must be class, representing the class label. The second line is the type of each attribute, only numerical and categorical are acceptable. The last word must be label. The words in both lines are all divided by comma mark (,) without space ( ).

Moreover, the names of two files must be the same.

You can open iris.data and iris.names under datasets directory to understand the rules above.

How to run it?

The entry of code is in the file validation.py. You can modify the test_data_path and test_scheme_path at the end of file. All datasets can be found in datasets directory. I provide 4 running modes: CBA-CB M1/M2 with/without pruning. Choose one mode you want to run, keep that line available and comment out the other three lines.

For example, if you want to take iris dataset as test data, just let test_data_path = datasets/iris.data and test_scheme_path = datasets/iris.names. And if you want to test CBA-CB M1 without pruning, you can prefix the last three lines with hash mark (#), like

# just choose one mode to experiment by removing one line comment and running
cross_validate_m1_without_prune(test_data_path, test_scheme_path)
# cross_validate_m1_with_prune(test_data_path, test_scheme_path)
# cross_validate_m2_without_prune(test_data_path, test_scheme_path)
# cross_validate_m1_with_prune(test_data_path, test_scheme_path)

Then you can run the program.

Acknowledgment

Dasong Chen and Lujing Xiao assisted me to complete this project. Thanks for their effort.

Reference

[1] Liu, Bing, W. Hsu, and Y. Ma. "Integrating Classification and Association Rule Mining." Proc of Kdd (1998):80--86.