ptnplanet / java-naive-bayes-classifier Goto Github PK

I tried to serialize a classifier I train on my data set but it turn out that it uses a non serializable object. The "Classifier" class in itself implements serializable but I am not quit sure which class it uses does not.
Thank you and great work with the naive Bayes classifier.

via Classifier code i guess the featureProbability should not bigger than 1.0, but I found that if have 2+ "I" in the trainning material, I got the featureProbability more than 1.0+. following is the code snippet:
in test change the learn material like:
final String[] positiveText = "I happy now, I like you".split("\\s");

print some info in featureProbability :

float featureCount = (float) this.featureCount(feature, category);
float categoryCount = (float) this.categoryCount(category);
System.out.println("raw p(f|c) " + String.valueOf(feature) + featureCount + "->" + String.valueOf(category) + categoryCount + "->" + featureCount / categoryCount); return featureCount / categoryCount;

then the output:

raw p(f|c) I2.0->positive1.0->2.0

does the result is correct? or can you clearfy what the math model to calc p(f|c)?

Hello,

I know you haven't worked on this in a while but was wondering if you had any idea why I keep seeing this issue. I have added about 25 categories to the model with lots of data in each category. For the majority of the categories no matter what I feed in when I classify a chunk of text most of the categories return a probability of infinity.

ex.

Classification[
category=friends_gatherings,
probability=Infinity,
featureset=[
after,
school,
soccerabout,
this,
...
--
]
]

I am using double[] as feature instead of string in runnableExample, but the feature of class -ve is still classifying as +ve. And in featureCount, I am not getting any value other than null.

Hello.
Is there any way to store results of the learning as experience data in my database?

Forgive me if I am wrong, but on this line

would not
return this.getFeatureCount(feature, category) / (float) this.getFeatureCount(feature);
return P(category | feature) rather than P(feature | category)?

For each learn call, the probability that a category happens is set to 1.0. This in turn makes every classify be whatever the last learn's category was.

Please could you add a (MIT?) license?

Thanks

Hi, thanks for simple and well working code.
I have some question featureWeighedAverage method.
What does these parameter do? Is it like Laplace smoothness term?

• weight
• assumedProbability

It would be helpful for me if there is a paper to refer to the weighted average formula used in the featureWeighedAverage method.

It is better to have a save and load for model to be reused later

The documentation is referenced in the readme file but there is no link to it? pls help thanks (also kinda new to java so might be a stupid question - any help is well appreciated)

getprobability() return value more than 1 (not between 1 and 0)
what should i do?
tnx

Is it possible to save/load (maybe some kind of serialization/deserialization) the trained classifier?