Software Fault Prediction with Naive Bayes

To make your test phase more efficient

This is a basic Software Fault Prediction (SFP) program, written in Java.
It uses a Naive Bayes Classifier for predicting software faults.

A custom, rather small dataset is used (dataset.txt).
The structure is as follows:
Please note that the 'module' column is not part of the dataset, it is just shown to make it easier to understand.

module	f1	f2	...	fn	Faulty
1	m1_f1	m1_f2	...	m1_fn	0 or 1
2	m2_f1	m2_f2	...	m2_fn	0 or 1
...	...	...	...	...	0 or 1
k	mk_f1	mk_f2	...	mk_fn	0 or 1

Where
       f1 stands for feature 1, f2 = feature 2, ...
       m1_f1 stands for first feature of first module
       n is total number of features for each module
       k is total number of modules in dataset (total number of lines)

       A module can either be faulty (=1) or non-faulty (=0)

The used equations can be seen below.

Equation 1: Naive Bayes Classifier

Equation 2: Standard Deviation

Equation 3: Gaussian Distribution Function

Equation 4: Arithmetic Mean

Equation 5: Category Specific Set Size

Run from command line:

First compile the project:
       javac model/SFPModel.java
Then run it:
       java -cp . module/SFPModel

Notes

A second dataset file is included "nasa_cm1_dataset.txt" which is a dataset from NASA, specifically for CM1. It was taken from PROMISE.

Full link: http://promise.site.uottawa.ca/SERepository/datasets/cm1.arff

The dataset was slightly modified by changing each last value of each module from boolean values into integers (True = 1, False = 0). This makes it easier to process.