Don't visualize all the data together in a single plot.

Learn how to differentiate among the categories by using "facets".

Learn how in the following short tutorial:

https://youtube.com/shorts/cmERbDabg64?feature=share

Facets to visualize data

If you have many columns that indicate the same information by year, think about something other than a for loop to plot them.

Instead, you can use the melt function.

Learn how:

https://www.loom.com/embed/d8780bdd031d45f2b0a3719f382a75dd

Melt to plot yearly columns

An intern asked me: "why do we use Max Pooling on Neural Networks?"

🧵

Before getting to this, you need to understand:

1. Convolutional Layers

A neuron on the neural network represents a coefficient of the mathematical equation.

Convolutional Layers make each neuron a matrix (3x3 "coefficients") of the Mathematical Equation.

Watch the following video from @deeplizard to visually understand Convolutional Neural Networks (CNNs):

PS: they are the best channel to explain Deep Learning abstract concepts with easy-to-understand examples.

https://www.youtube.com/watch?v=YRhxdVk_sIs

2. Relevance of the pixels to predict the categories of your data.

Why would we have 16 pixels (4x4) to determine the border of the seven?

We could simplify the image by taking the Maximum number of 2x2 pixels making it smaller and optimizing computation (aka MaxPooling2d).

Watch the following video for a visual explanation:

https://www.youtube.com/watch?v=ZjM_XQa5s6s

So, are we done by just computing a Machine Learning model and getting a good score?

- Hell no!
- We should strive to tell a story with data.
- A visual story.

Learn how to develop the Python code to create a visual story ↓ 

We load the following data where each row represents a USA State, and the columns represent characteristics related to car crashes. 

We compute the Mathematical Equation to predict the total number of accidents based on alcohol.

Later, we create a filtered DataFrame called *df_pred* with the columns involved in the model. 

We have gathered all the information in a DataFrame, but we want to create a Visualization to tell this data's story better.

We can concatenate different seaborn functions line by line, according to the figure we want to represent, and gather them in the exact figure: 

To make this Visualization better, we should add the meaning of each figure with a legend: 

Fair enough, but it'd be great to add the Mathematical Equation of the model to understand how much the number of accidents increases based on the coefficients.

Let's create an f-string with the model's attributes and add it to the legend: 

Now, we can better tell the story:
    
For each unit increase in alcohol, the total accidents will increase 2.03 units.

What to do next?

1. Follow me @jsulopzs for more of these
2. RT the tweet below to share this thread with your audience
3. Write the code with your data https://twitter.com/1270439165658828805/status/1587392233816444931

Data Visualization | Linear Regression Model

Introducing the #PythonResolver Roadmap to learn #Python + #DataScience 👀

During 3 years working as a Private Tutor, I have tried a variety of materials with more than 500 students from several industries:

· Medicine
· Finance
· Energy
· Psychology

Check it out ↓

#1 Resolving #Python for Data Science

Please, read the README files of every GitHub repository you want to interact with. They explain how to use the materials.

Otherwise, you'll go over the materials like a headless chicken 🍗 

https://github.com/jsulopz/resolving-python-data-science

If you have any doubts, please reply to the tweet you doubt with.

I will be more than happy to help you because you give me feedback to improve my content and grow as a content creator!

Win-win 😉

- Everything nowadays is information.

- Information is data.

Where is the data gathered to be later analysed?

- In a Data Table

How does Python work with Data Tables?

- The #DataFrame from Pandas library

How can you master the manipulation of the DataFrame?

#2 Practice with #Pandas Exercises

Guilherme Samora created this GitHub repo. It contains a sequential list of exercises that increase the difficulty as you advance the chapters.

Please, read the README to learn more about how to use them!

https://github.com/jsulopz/pandas_exercises 

It would help if you aimed to be pragmatic:

Apply the concepts you learn in #1 to #2 exercises simultaneously.

It would help if you aimed to be curious:

Don't focus on results; you have got the solutions already. Instead, focus on taking notes and solving your doubts.

Once you have learnt #1 Resolving Python for Data Science and have gotten used to working with the DataFrame in #2 Pandas Exercises, you are ready to jump into #MachineLearning

Companies, Organizations & Institutions are making decisions all the time.

They make decisions based on the data their Stakeholders generate. Then, they manipulate the DataFrame and create cool visualisations to understand what's going on.

Once they have understood, they make decisions based on guesses about what may happen in the future.

Nevertheless, it's tough for a human to predict what may happen in the future because you need to consider more than 10 factors/variables.

What can a human do then?

- Rely on #MachineLearning

Machines learn from historical data to create a Mathematical Equation that predicts the future.

https://twitter.com/jsulopz/status/1449735653328031745?s=20&t=AcDjccG585cUjlCRNIhS2Q

This Mathematical Equation can gather as many factors/variables as you want. Therefore, the machine will make better predictions than a human.

How can I learn #MachineLearning?

As I was teaching people "à la carte" when they asked me to learn, for example, the Decision Tree algorithm that the University assignment required, I realised I needed to explain more concepts to understand the whole picture.

The following materials cover the whole picture.

#3 Machine Learning

This GitHub repo is the fruit of all the University assignments I have helped my students with.

They gather the different types of Mathematical Equations the Machine can learn in sequential order of difficulty.

https://github.com/jsulopz/machine-learning

After you are done with this GitHub repo, you will know how to apply different Machine Learning algorithms based on the variable you'd like to predict.

What's next?

- Advanced Machine Learning for feature interpretation

- Deep Learning