No method is best unless you have decided your goals. A lot methods can perform classification. Selection is mainly depended upon:

These all will collectively define what is "Best" for you. Hope it will help. 

It is totally based on  the characteristic of your input data and dataset. You may used supervised machine learning(classification,regression) or unsupervised learning(clustering). Finally you need to measure the accuracy of your result. Most widely used classification algorithms are, Naive Bayes, KNN, Support vector machine, random forest, decision tree. You may follow the below papers for your help,

http://www.mecs-press.org/ijieeb/ijieeb-v8-n4/IJIEEB-V8-N4-7.pdf

https://pdfs.semanticscholar.org/fb6b/4b57f431a0cfbb83bb2af8beab4ee694e94c.pdf

http://www.mecs-press.org/ijitcs/ijitcs-v6-n11/IJITCS-V6-N11-10.pdf

Feature extraction based on hybrid method can improve the system recognition.