Generally SVM and its variants like laplacian SVM are good predictors. Do you have an idea about the size and number of predictors in the data?

Thank you for answering my question, such a great thing for me to get answer from you all, i will try the SVM classification model and find the appropriate size of predictors.

Hi,

You could also try the Naive Bayes method (available in the Weka software). Depending on the data (especially if it contains some noise) Bayes can give better results than SVM. Maybe you could try both to see which gives the best results for your data. In my paper:

David F. Nettleton, Albert Orriols-Puig, Albert Fornells. A study of the effect of different types of noise on the precision of supervised learning techniques. Artificial Intelligence Review 01/2010; 33:275-306. 

Naive Bayes and SVM were compared.

Regards,

David Nettleton.

Article A study of the effect of different types of noise on the pre...

Thank you so much David F. Nettleton, i will read your paper to add new knowledge

Regards,

Pandu Dwi Luhur Pambudi

Dear Sir. I recomend to keep it simple. The mos used method are standars fro WHO and PAHO and similrs  based in monthly calculated Number of cases per 1 thousand unhabitants IPA this can be collected form monthly bulletings of health organizations en each coutry the map the to analize spatila distribution and sopt s of haiher inience and persistence, the spatial distributions can also be  statisticl analized with climatic consitions and foloow up of control campains.

Dear Pandu

It sounds like you are trying to build a Malaria Early Warning System (MEWS). My advise would be to try and align your methods with the WHO standards and techniques. This does however not mean that you cannot play with the data to obtain better predictive patterns and values.

Good luck.

Burnie

open source social networking platform scrapes information from RSS feeds or other Internet sources and uses a support vector machine behind the scenes to tag relevant items. As they are tagged, an expert gives a thumbs up or down to help the machine learn from its own mistakes. “Consensus develops and Riff does better over time,

http://biomedicalcomputationreview.org/content/biosurveillance-text-mining-freakidemiology

Nice to meet you guys. i want some guide on project methodology, to be specific i want to work on predictive system of blood related diseases and i am targeting medical record from general hospital. I was wandering if data mining could be best route or text mining. secondly, choice methodology on best classification to be use is kind confusing me. pls need advice.

Extracting intelligence data for quick response to decision making of mass data is possible with Hadoop and Sqoop