Data Science Project | Bike Rental Count | Part 5 | Regression, Selecting right Algorithm

The world is now awash in data and we can see consumers in a lot clearer ways by Max Levchin, Co-founder of Paypal

Source: bbva

Selecting right Algorithm

First, we need to understand the nature and objective of our model. Target variable decides which algorithms to use. Our target variable is a continuous variable that means it is regression problem. So, we have to choose algorithms that works under regression. Let's look at some of regression algorithms.

 Linear Regression/Polynomial Regression
 Ridge Regression
 Lasso Regression
 Elastic Net Regression
 Decision Tree
 Random Forest
 kNN(k-Nearest Neighbors)
 XG Boost
 Gradient Boosting Machines(GBM)
 Light GBM
 Adaboost
 Neural Networks

As you can see, I have listed some of algorithms which  work for Regression problems but you can't use them all(I mean you can ) but every algorithm has their advantages and disadvantages and comes with their own constraints. Like Ridge, Lasso are the modification of the Linear Regression algorithm and by combining some of properties of Ridge and Lasso, Elastic Net born. 

So, according to the situations and objective, we have to make sure we are using right algorithms for getting the best results. Surely, we can apply trial-error method(you should).

The thing is some of algorithms used for predicting and some of used for tuning the performance of the model and some can be used for both classification and regression problems. Some are simple algorithms and some comes up with complex functions. It's a hectic task to decide but with practice comes experience, this is the only way.

For now, we will start with Linear Regression and then compare with the Random Forest algorithm.


Feel free to comment your opinions💓