Have you ever considered changing the world with data science models? Or maybe you want to learn some machine learning techniques just to supplement your development skills or a project you are working on?
Magda, a Senior Developer at Freeport Metrics has just shared a few extremely useful tips for building data science models.
To learn data science, find a dataset that fits your interests. For this particular tutorial, we used a dataset from a competition called “Histopathologic Cancer Detection.” Freeport Metrics has quite strong experience in data analytics-related projects for the HealthTech industry, so it was a natural choice for one of our developers.
First of all, we need to choose a game plan, meaning to decide if we are going to use a supervised or unsupervised type of machine learning. In our case, we have a training set and ground truth–train labels. The goal here is to learn a function based on example input-output pairs.
Once we set up a game plan, it’s time to start by importing the files we need to Google Colaboratory ( we decided to use explore Google Colaboratory which is a free Jupyter notebook environment with Python 3).
The process we described was a part of a competition, thus there are winners and losers. The main takeaway we have is that we should improve the predictive power as much as possible and learn how to pick the techniques that actually will improve our score.