Hi Ahmed,

In addition to imputation methods such as KNN, regression and MLP for missed data, other approaches can be applied. Ensemble classifiers such as AdaBoost and Bayesian Network can classify without imputation.

Please find the review paper as an attachment.

Hi Ahmed,

In addition to imputation methods such as KNN, regression and MLP for missed data, other approaches can be applied. Ensemble classifiers such as AdaBoost and Bayesian Network can classify without imputation.

Please find the review paper as an attachment.

Thank you all ......  very helpful.

Hi Ahmed,

I would suggest that, It would be much more useful if you calculate the expectation values for the missing values and afterwards when you will encounter the missing data you can calculate the expectation for the expected values for that data.

One thing you might consider is the computational complexity / overhead of using the different techniques if you wish to execute the classifiers in real time, particularly with internet access to the application  (for example, look at how a simple non-AI system to decide whether or not tamiflu should be prescribed was overwhelmed in 2009 http://www.express.co.uk/news/uk/116041/Flu-website-is-overwhelmed-in-minutes). Sometimes it is worth considering whether using a relatively crude measure, such as substituting the mean (median / mode) will be "good enough" in view of its low computaitonal overhead. 

hi friend

first you apply some data missing algorithms to fill it up or do some other task.

Then apply some algorithms.

As simple as that.

From

Dr.Indrajit Mandal

Dear Ahmad,

the metric I invented give similarity in the range of [0,1[, this means if a value in a spesfic feature is missing it will not affect the final distance significantly.

I just have published a paper that proposes the new metric which was invariant to the dimensionality of the feature vector, you may find the paper at the link:

https://www.researchgate.net/publication/264995324_Dimensionality_Invariant_Similarity_Measure?ev=prf_pub

Article Dimensionality Invariant Similarity Measure

Dear all,

In the context of support vector machines it is possible to also change the problem formulation in the case of missing values, see e.g.

Pelckmans K., De Brabanter J., Suykens J.A.K, De Moor B., ``Handling Missing Values in Support Vector Machine Classifiers'', Neural Networks, vol. 18, 2005, pp. 684-692.

Best regards,

Johan Suykens

----------------------

Prof. Dr.ir. Johan Suykens

Katholieke Universiteit Leuven

Departement Elektrotechniek - ESAT-STADIUS

Kasteelpark Arenberg 10

B-3001 Leuven (Heverlee)

Belgium

http://www.esat.kuleuven.be/stadius/members/suykens.html