Implementation of "A Correlation-Based Feature Weighting Filter for Naive Bayes" (Liangxiao Jiang et al.)

Getting started

Run the following command to install the dependecies:

pip install -r requirements.txt

The CBFW classificator was implemented following the standards of Scikit learn. For a example of usage please refer to the notebook example.ipynb.

Results

Accuracy results (%):

Dataset	Paper	Mine
anneal	98,5	91,87
anneal.ORIG	94,60	91,90
audiology	74,22	70,15
autos	77,95	72,81
balance-scale	73,76	90,72
breast-cancer	72,46	72,49
breast-w	97,14	97,34
colic	83,34	83,22
colic.ORIG	73,70	73,05
credit-a	86,99	86,59
credit-g	75,70	75,11
diabetes	78,01	66,82
glass	73,37	97,79
heart-c	82,94	82,58
heart-h	83,82	83,99
heart-statlog	83,44	82,78
hepatitis	85,95	84,29
hypothyroid	98,56	98,93
ionosphere	91,82	91,39
iris	94,40	94,20
kr-vs-kp	93,58	93,52
labor	92,10	89,75
letter	75,22	75,66
lymphography	84,81	81,72
mushroom	99,19	99,88
primary-tumor	47,20	45,05
segment	93,47	86,90
sick	97,36	97,26
sonar	82,56	75,39
soybean	93,66	92,40
splice	96,19	96,13
vehicle	62,91	61,08
votes	92,11	92,14
vowel	68,84	62,33
waveform-5000	83,11	82,10
zoo	95,96	96,45

Elapsed time (seconds):

Dataset	Paper	Mine
audiology	5,28	2,18
balance-scale	0,09	0,0053
breast-cancer	0,18	0,051
breast-w	0,21	5,56
colic,ORIG	0,99	0,43
credit-a	0,46	0,15
credit-g	1	0,36
diabetes	0,23	12,87
glass	0,19	0,015
heart-c	0,22	0,031
heart-h	0,25	0,024
heart-statlog	0,11	0,0093
hepatitis	0,25	0,029
ionosphere	1,69	0,15
iris	0,12	0,0021
kr-vs-kp	6,82	1,21
letter	29,13	5,6
lymphography	0,35	0,032
mushroom	12,08	7,78
primary-tumor	0,43	0,088
segment	4,41	0,05
sonar	1,59	0,069
waveform-5000	17,4	0,46
zoo	0,27	0,015