a. Understand the data(exploration, type of features, numerical vs categorical)
b. Understand the metric to optimize
c. Decide cross validation strategy
d. Start hyper parameter tuning which furthermore includes:
i. Data transformations (like scaling, remove outliers, treating null values, transformation categorical variables, do feature selections, create interactions)
ii. Choosing algorithms and tuning their hyper parameters
iii. Saving results
e. Combining models (ensemble), possibly on multiple levels
nice blogs:
http://blog.kaggle.com/2016/02/10/profiling-top-kagglers-kazanova-new-1-in-the-world/
https://mlwave.com/about/
http://fastml.com/
https://www.analyticsvidhya.com/
