Far too often, folks skip the most crucial part and launch into the steps above. Before you start doing anything, you have to know your "why." If I don't know where I am going, I won't know the best course of action.
Train a model right quick by following a few steps.
- Explore your data (EDA)
- Prepare your data
- Clean your data (Data Wrangling, Munging, Cleaning)
- Preprocess your data - scale it, normalize it, lemmatize it, etc
- Feature Engineering (make new features or combine them or drop some
- Test your features (which features are significant and junk)
- Train, test, split
- Pick your model
- Supervised vs unsupervised learning
- Classification vs regression
- Train your model
- Evaluate your results
- Tune your hyperparameters
- Do your prediction(s)
Easy, right? Nope!
KNOW YOUR WHY
Before I take on any problem, I have to know my why.
W- Where am I going?
H- How am I going to get there?
Y- What's my y (target variable)?
If I don't know where I am going, I won't know the best course of action. If I don't know my target, then I won't know what success looks like. Overall, if I don't ask myself questions BEFORE I start my projects, I will waste time and energy going down rabbit holes.
So, before you sit down for that marathon coding session, or before you do that Kaggle competition, ask yourself, do you know your WHY?