It's important to understand basic concepts such as probability, descriptive statistics, statistical inference, and mathematics such as linear algebra and optimizations.

This requires an understanding of data structures which are Series and Dataframe and knowing data preparation techniques using tools such as Pandas and Numpy. When comes to machine learning, people think that creating a model is the most important part. But let me tell you the reality.

Actually, data processing is the most important part. Why? Because if your data are not well-cleaned, you will have a poor model. So pay attention to empty values, bad formatting, etc. Turns out, this is the most tedious and time-consuming part, but in the end, so satisfying!

Data visualizations are a crucial tool for understanding and interpreting data. It's important to know how to use tools such as Matplotlib and Seaborn to display plots, histograms, bar chat, etc. It can be hard to know how to display such code.

Seaborn it's easier to implement because the syntax it's simple, but I recommend you to have snippets to help you understand how to use these libraries, because, usually, you use both in your code. AI tools like ChatGPT can also help you code your charts.

Today, we don't need to create our own algorithm, except if we have a very specific need or if we do a deep tech project. But libraries like Scikit-learn, Tensorflow, and Pytorch help us implement these algorithms very easily to create our models.

However, it's important to understand the main machine learning algorithms such as linear regressions, decision trees, neural networks, and clustering algorithms, understand how they work in general, why and when to use them, with which data, and the best practices.

Understanding model evaluation methods such as cross-validation, confusion matrix, and performance measures such as AUC and F1 score is crucial to make sure that our model can be generalized.

Also, knowing exactly why using accuracy allows us to avoid errors of interpretation. This requires a deep understanding of your data. For example, if our dataset is imbalanced, it's better to use the F1 score to avoid bias.

Of course, practice is mandatory to mastering data science skills. I recommend doing classic projects like Titanic, image classification, and spam detection to understand the underlying concepts and use best practices.

However, don't hesitate to go further by participating in data science projects on online platforms such as Kaggle, Omdena, DrivenData, MachineHack or working on personal projects is recommended. For your personal project, my advice is to choose a subject related to your passion, for example, music or ecology. Or perhaps related to your culture or very important like women's rights. It will be so valuable to your portfolio. And remember, try to document everything. What is the best place? Your notion for example, but even better, blog posts! If you want help with that, don't hesitate to join our membership.

I know that it could be so complicated to start a data science project. Which libraries do I need to import again? Should I use a histogram or a bar chart? You know what? Practice makes you better. So what are you waiting for? Let’s go!

Data science is a constantly evolving field. It's important to stay up to date with the latest trends by following data science blogs like KD Nuggets, attending conferences like NeurIPS, and reading professional literature like Paper With Code.

However, subscribing to newsletter can be so overwhelming. My advice? Create filters in your email box and set a day in the week when you will read everything. And most importantly, don't subscribe to everything. it can be tempting, but I'm sure that you already have a busy life. So choose 3 newsletters max.

This is a starting point for learning the basics of data science. It is important to persevere with learning and never stop growing by following trends and practicing regularly.