search

ai-se / bellwether_community

Bellwether Community detection with JS projects using r2c
GNU General Public License v3.0
2 stars 0 forks source link

CommitGruy15): does your cfs work? #13

Open timm opened 5 years ago

timm commented 5 years ago

for 10 data sets, showme classification results before and after cfs

Suvodeep90 commented 5 years ago

results: CFS by weka: F1-score {'ant-1.5': [0.83, 0.93, 0.95, 0.88, 0.88], 'ant-1.3': [0.77, 0.87, 0.79, 0.96, 0.77], 'camel-1.0': [0.94, 0.93, 0.93, 0.95, 0.96], 'camel-1.6': [0.72, 0.73, 0.77, 0.76, 0.8], 'velocity-1.5': [0.65, 0.67, 0.79, 0.74, 0.78], 'velocity-1.4': [0.86, 0.87, 0.78, 0.78, 0.86], 'jedit-4.2': [0.86, 0.86, 0.89, 0.91, 0.81], 'jedit-3.2': [0.82, 0.77, 0.74, 0.77, 0.84], 'log4j-1.1': [0.82, 0.77, 0.85, 0.84, 0.9], 'log4j-1.0': [0.86, 0.84, 0.92, 0.8, 0.81]}

Median Values(f1 score) ant-1.5 0.88 ant-1.3 0.79 camel-1.0 0.94 camel-1.6 0.76 velocity-1.5 0.74 velocity-1.4 0.86 jedit-4.2 0.86 jedit-3.2 0.77 log4j-1.1 0.84 log4j-1.0 0.84

CFS python: F1-Score {'ant-1.5': [0.84, 0.81, 0.92, 0.83, 0.88], 'ant-1.3': [0.77, 0.91, 0.84, 0.84, 0.75], 'camel-1.0': [0.94, 0.93, 0.93, 0.94, 0.96], 'camel-1.6': [0.72, 0.75, 0.75, 0.74, 0.76], 'velocity-1.5': [0.74, 0.73, 0.79, 0.79, 0.73], 'velocity-1.4': [0.87, 0.9, 0.78, 0.8, 0.87], 'jedit-4.2': [0.83, 0.92, 0.89, 0.91, 0.81], 'jedit-3.2': [0.82, 0.77, 0.71, 0.74, 0.71], 'log4j-1.1': [0.86, 0.64, 0.8, 0.84, 0.9], 'log4j-1.0': [0.76, 0.8, 0.73, 0.83, 0.75]}

Median values(f1-score): ant-1.5 0.84 ant-1.3 0.84 camel-1.0 0.94 camel-1.6 0.75 velocity-1.5 0.74 velocity-1.4 0.87 jedit-4.2 0.89 jedit-3.2 0.74 log4j-1.1 0.84 log4j-1.0 0.76

timm commented 5 years ago

i want you to start showing me data like https://docs.google.com/spreadsheets/d/10BGV8CqiOciR3HtNjvSIbUePK1ZqfsUerpGiMtGMRsE/edit?usp=sharing

see the comparison columns on RHS? very clear that the numbers are very similar (all so close to 1).

timm commented 5 years ago

in any case, this cfs works.

what are its hyperparameters? this is a beam of one and a horizon of five?

This is a beam of one and horizon of five attributes of non-improvement for cut-off

Suvodeep90 commented 5 years ago

There is python wrapper for weka called "python-weka-wrapper3". using which we can directly use weka's implementations in our python code. I am trying to install that but there is something called javabridge, which is giving me a problem while installing.

timm commented 5 years ago

where is the CFS code for this one?