Hi Abdalraheem Alsmadi,

Amongst several available conventional machine learning classification algorithms, I would suggest you to consider these classifier based on my experience: Support Vector Machine (SVM), Artificial Neural Networks (ANN), Linear Discriminant Analysis (LDA), and Naïve Bayes (NB). These classification algorithms perform fairly well in terms of classification accuracy for multi-class problems. However, ANN and SVM are more computationally complex that the other two classifiers thus requiring more time to output a class decision although their classification performances are sometimes slightly better that of the LDA and NB. Therefore, there is always a trade-off between classification accuracy and computational complexity of the classification algorithms which you may also need to consider in your research.

I hope you would find this tips helpful.

Wish you good luck.

Hi Abdalraheem Alsmadi,

Amongst several available conventional machine learning classification algorithms, I would suggest you to consider these classifier based on my experience: Support Vector Machine (SVM), Artificial Neural Networks (ANN), Linear Discriminant Analysis (LDA), and Naïve Bayes (NB). These classification algorithms perform fairly well in terms of classification accuracy for multi-class problems. However, ANN and SVM are more computationally complex that the other two classifiers thus requiring more time to output a class decision although their classification performances are sometimes slightly better that of the LDA and NB. Therefore, there is always a trade-off between classification accuracy and computational complexity of the classification algorithms which you may also need to consider in your research.

I hope you would find this tips helpful.

Wish you good luck.

Naive Bayes and LDA. SVM could be also an option.

SVM and Naive Bayes

I do believe that classification algorithms are actually data dependent. However, the least data dependent algorithm is usually ANN as it works as a universal approximation function for most problems. Hence, if you have enough computational resources, it should be better to explore as many algorithms as possible to get the best possible results on the problem in hand.

I suggest DNN and Recurrent Networks for large number of attributes.

Multi-Class SVM can be a good option

SVM in ENVI Software