I’ve completed the 2nd course of the Machine Learning Specialization. Now I’m thinking that for supervised you have two options either go for decision tree or neural network. So do we ever use standalone algorithm like linear regression/logistic regression in our project?
The algorithm you use for a project depends on many factors like training budget and the desired performance. Feature engineering is crucial in getting the best out of the dataset and the underlying algorithm.
There’s just one guiding principle when it comes to where you start. For structured (aka tabular) data, start with classical machine learning algorithms before diving into deep learning methods. For unstructured data (text / images / audio), start with neural networks and fine-tune to your dataset if required.
2 Likes