Which Machine learning classifier should I use?

i am training an algorithm to differientiate between two behavioural variables.

I had 100% accuracy for ensemble bagged trees, and 97.6% for SVM cubic and 97.6% for KNN.

this will be used in unknown data sets. do i go with the most accurate? or are certain classifiers better for different situations.

thanks!

Answer 1

[Abhishek Singh]

For your further query you can look into these links:

http://blog.echen.me/2011/04/27/choosing-a-machine-learning-classifier/

https://stackoverflow.com/questions/2595176/which-machine-learning-classifier-to-choose-in-general

In short:

Boosting - often effective when a large amount of training data is available.

Random trees - often very effective and can also perform regression.

K-nearest neighbors - simplest thing you can do, often effective but slow and requires lots of memory.

Neural networks - Slow to train but very fast to run, still optimal performer for letter recognition.

SVM - Among the best with limited data, but losing against boosting or random trees only when large data sets are available.

Hope this helps,

[Katrina Anderson]

Awesome! That helps a lot!!

Thankyou :)

Answer 2

[Abhishek Singh]

Behavioral variables are just numbers to differentiate between customer or user behaviour

The division of a market into groups according to their knowledge of, and behaviour towards a specific product. Behavioural dimensions commonly used to segment markets include benefits sought, user status, usage rate, loyalty status and buyer readiness stage. Also called Behaviouristic Segmentation.

Certain classifiers are indeed better for different situations.

It depends upon what you want to use the classification for. Since you have not explained your use case, it will be unwise to suggest a classifier based on the accuracy alone. That is unless accuracy is what you care most about.

For example, if you care more about interpretability other classifiers are probably a better choice since they are quite close in terms of accuracy.

Again, it is impossible for anyone to advice you on what classifier to choose without any knowledge of the situation in which it is going to be used. These decisions are better left to data scientists who have in-depth knowledge not only in the business use case but also about the trade-offs between different classifiers.

[Katrina Anderson]

Thankyou,

I am not a data scientist. I am a masters student looking to classifier behaviour using an IMU. I am unsure about the trade offs for different classifiers. Where could I research this?

Thanks Katrina

[Abhishek Singh]

I'll write a new answer for you