Good day everyone,
I must thank the organizers of this forum for their efforts, it makes the
machine learning experience very encouraging.
I have a dataset i want clarification on.
Having 12 attributes inclusive of one class column trying to predict two
classes, Pass or Fail.
These are all results from a local ping pong tournament that has been going
on since 2010 till date though most weeks results are missing in the
The experiment is to predict the most rows that will have a PASS.
Over the past few weeks, I have tried the classifier BayesNet on the data
and the predicted results is not too encouraging, always predicting at
least a 3rd of the actual results accurately.
What i mean is that like last week which was week 16 in the ping pong
tournament, running a Bayes classifier resulted in 17 rows that had PASS,
but the actual result in the tournament for that week had 12 rows having
PASS. (Actual result means real life result)
Out of these 12, my Bayes experiment predicted 5.
What i want someone interested in my data to do is tell me how to tweak
WEKA to at least predict more accurately. Maybe getting 8-10 accurately or
at least 70% of the real life prediction correctly.
In other words, since my bayes classifier predicted 17 of which 5 was
accurate predicted in the actual results, how can i tweak weka to predict
8-10 or even the whole 12 PASS out of the 17?
I will include week 17 tournament fixtures in the email attachment i intend
to send, can you make use of it and come up with more accurate results.
I will provide the actual results on Monday so we can compare.
Let me know where to send the arff data.
There is no noise in the data, it is ready for immediate use.
Oluwakayode Iretunde Adeyemi-Kayne
'SHUGA' - Simple Humble Unassuming Good-natured African
Goos Media Concepts
Show replies by date