Think before you act! Choosing Machine Learning algorithms
Question first then look for the answer… not the other way around! Before looking at the data it is important to understand which problem you are trying to solve and how, in the end, the business will benefit from your model. This first brainstorming helps to determine how to frame the problem, what algorithms to select and measure the performance of each one.
Importance of Problem Statement
Before tackling any data science challenge, in your company or at any hackathon, it’s important to spend some time identifying which algorithm we should apply for the business problem at hand. First, create a Problem Statement from the problem given.
Here are some pointers you can follow to help build your own Problem Statement:
- What is the type of problem you are working on?
- What data was made available to you and what is the expected outcome?
- Your statement should be concrete without many sentences.
Your Marketing Director wants to launch a new online campaign for the recent Computer Sciences courses. He asks you to identify which users are more prone to be positively influenced by the campaign. You are given a dataset of a recent online survey they sent out to their users.
You’re expected to build a model that will allow him to understand which group of people, depending on certain characteristics, will enroll or not in the offered course.
Once you have the problem statement define you’re set to identify which algorithms you should choose to answer the business problem.
- Classification algorithm
- Regression algorithm
- Clustering algorithm
- Recommendation algorithm
Types of Machine Learning Algorithms
There are tons of different Machine Learning systems! Therefore in order to classify them according to their approach, it is useful to cluster them in broader categories based on:
a) Whether or not they are trained with human supervision
- Supervised Learning;
- Unsupervised Learning;
- Semisupervised Learning;
- Reinforcement Learning.
b) Whether or not they can learn incrementally
- Online Learning;
- Batch learning.
c) Whether they work by comparing new data points with already known information or detect patterns in the training data and use that to build a predictive model
- Instance-based Learning;
- Model-based Learning.
Note : These criteria are not exclusive; i.e. an ML system can combine three of the categories.
For more insight on each one of this broader categories follow go to:
If you liked it, follow me for more publications and don’t forget, please, give it an applause!