Selecting an algorithm for processing training data

Select an algorithm

Before proceeding, I hope you have covered our previous articles:

• Starting in Machine Learning.
• Preparing training data in Machine Learning.

In the above second article, we have prepared the training data now we are at the step where we will make the selection of the algorithm to get the trained model so that we can predict which people will develop diabetes in future?

In ML we have over 50 algorithms to choose from but which one to select, it depends on several factors:

• Learning type: Here, in our problem, we are looking for a model which can predict which people are likely to develop diabetes. So, we are looking for a prediction model and we can say:

Prediction model => Supervised machine learning

• Result type: In this type, we have two classifications:

1. Regression: In this, we want to predict e.g. the price of the house which depends on several values such as bedroom size, rooms available and size, etc and these values vary for each house.
2. Classification: In this, we are interested in discrete values such as small, medium and large or true or false value.

And, we are interested in the classification result type because we want to predict whether people will develop diabetes or not.

Hence, the algorithm much supports binary classification.

• Complexity: We must choose the simple algorithm and should not consider ensemble algorithms which are kind of nested algorithms and can be very difficult to debug.

Based on the above-considered factors we have selected some famous algorithms in ML:

But still, we need to make the selection which one to choose.

Naive Bayes:

Here feature represents columns of the table in our data.

Logistic Regression:

Decision Tree:

Now, we can decide “Naive Bayes” can easily fulfill our purpose because:

• It is simple and easy to understand.
• Fast- up to 100x faster than other algorithms because it supports features having the same weight and we have all those columns in our data.
• It is stable to data changes.

These points will be more evident in the upcoming articles where we will work on training the model.

Thanks for reading this article and if you have liked it, please give us a clap and stay with us for more.

Happy Machine Learning :).