Skills you need to launch your Machine Learning Rocket in SEM

Published by Patrick Mebus on

The most important thing you should know: Machine Learning relies on data. The most important thing you should learn: how to handle this data and how to adjust the right tools and models. What you don’t need: A degree in math and information technology. Here are the skills you need to start with Machine Learning in Search Engine Marketing.

Machine Learning skills you should develop in SEM
1. Python Basics

To understand the basic programming concepts and the rules of code execution is a key skill of Machine Learning (and in Search Engine Marketing in general).

The good news for our Machine Learning ambitions: The most commonly used ML-programming language Python has a quite easy to understand syntax and data structure.

Python linear regression example for Machine Learning

That makes it flexible and easy to implement. Most probably this is the reason why Python became that popular in data science. It’s great for experiments and frequent adjustments.

There are a lot of programming languages outside, that are useful for Machine Learning, like R, C++, Java or Julia. But there is another big advantage that let me recommend to use Python at this point: most of the standard libraries and frameworks in Machine Learning support Python.

As a consequence you’ll find  the most Machine Learning examples and documentation written for this language by a giant user-community that is willing to share knowledge and solutions.

Fun fact about Python: The name has nothing to do with snakes. It’s named after the BBC comedy show Monthy Python’s Flying Circus. Something  really love ; ).

2. Fundamentals of Statistics & Probability

Let’s talk about the heart and the soul of Machine Learning. This is actually Statistics and Probability calculation.

No worries: You don’t need a university degree in mathematics. But Machine Learning can only improve your SEM-campaigns, if you’re able to prepare and validate your models properly.

All of these algorithms we use in Machine Learning, like Linear Regression or Logistic Regression, are deeply rooted in Statistics. 

Probably the most important corner stone of ML: statistics

I actually write this very often in my posts: start by learning just the things you really need for your specific tasks.

If your challenge is to predict future conversions, focus on learning the concept of linear regression.

Develop a basic skill-set that allows you to improve your urgent and most important tasks. From this point you can proceed and expand your knowledge in Machine Learning and Statistics.

 

3. Essentials of Linear Algebra

Linear Algebra is one of the key-concepts of Machine Learning. If you upload your dataset in to spreadsheet, you’ll find your data sorted in rows (x) and columns (y).

What your algorithm basically does (depending on your model) is to multiply the rows with the columns.

That’s what we calle a matrix. To develop a basic-knowledge for matrices and vectors will help to improve your understanding of processes in Machine Learning.

4. The algorithm-classes you really need for your tasks

“There is no free lunch”. What sounds like the worst message you can get around noon-time, is actually an important principle in Machine Learning.

Basically the theorem describes, that there is no universally optimal algorithm in general. An algorithm that’s very good in clustering audiences might be a worst decision to predict future sales.

The scikit-learn cheat-sheet helps to find the right algorithm-class for your purpose (source: scikit-learn.org)

But also the algorithm you use for sales-prediction usually is not supposed to be the forever-perfect-solution. There is simply no “best” algorithm.

Do you remember Google changing its algorithm frequently back in the early days. They tested different models of algorithms and tried to find the best solution dedicated to their specific task (providing the most relevant search results).

Nevertheless there are several classes that include a bunch of algorithms that are commonly used for a specific project ( e.g. Regression or decision trees for predictions).

5. Deep knowledge in Search Engine Marketing

Sounds trivial. But if you can’t describe your challenges and tasks in SEM, the world’s best algorithm can’t save you.

You have to know about your metrics and your Key Performance Indicators. You need to know your product, your market and your customers (!).

Be aware of your competitors and your channel specifics. That’s the homework you have to finish before you can make the step forward.

Recommended Resources

Getting started with Python: https://automatetheboringstuff.com/chapter1/

More Python practicing: http://www.practicepython.org

Statistics for Data Science: https://elitedatascience.com/learn-statistics-for-data-science

Types of Machine Learning algorithms: https://towardsdatascience.com/types-of-machine-learning-algorithms-you-should-know-953a08248861

Top 10 Algorithms for ML-Newbies: https://towardsdatascience.com/a-tour-of-the-top-10-algorithms-for-machine-learning-newbies-dde4edffae11

The No-Free-Lunch-Theorems in Machine Learning: http://www.no-free-lunch.org/

exited about this post? feel free to share

Patrick Mebus

I’m a Digital Marketer with deep passion for Search Engines, Automation and AI. I’m here to make Machine Learning more feasible for Search Engine Marketers.

Leave a Reply

Your email address will not be published. Required fields are marked *