To-dos to settle down

[shinhokang]

To-dos to settle down

[omartrinidad]

• The account is working.
• Not sure about the method, maybe we will create several models and choose the best one. Or merge them.
• I think we will work with Python. I like to work with Tensorflow. But for sure we can use Scikit, Pandas and also R to explore data and that kind stuffs.

[baldmer]

For a classification method.

Is recommended to compare the performance of several algorithms to select the best model for the problem.

consider:

• the number of features or samples
• the amount of noise in the data set
• the classes are linearly separable or not

[baldmer]

I have been reading about NaiveBayes classification, it is the most popular method for classification, it is easy to implement, I think everybody is using it. I read that is more suitable/popular for text classification, but according to this example, it is also useful for problems similar to the one we have.

https://www.youtube.com/watch?v=ZAfarappAO0

But I don't know how computational expensive it could be.

[omartrinidad]

If you want to use NBayes, you should first do a pre processing, for example, the ages are values from 18 to 30, for NBayes is better to have labels as ''teenage", "young", "old".

After that you can try. But will be expensive. So you can use "ensemble" methods. That is, take several samples, apply the algorithm, and combine the output.

Another issue is, we need our output with real values, and our training dataset have binary output. So, maybe a modified NBayes should help.

[baldmer]

About Ensembles

Ensemble methods use multiple models to obtain better performance than any single constituent model. When working with large datasets and quick response times, this can be a significant developmental bottleneck.