90% accuracy misses the mark

Let me tell you the story of a data scientist intern, so happy to use machine learning algorithms on real-life data after a year studying Iris dataset, comparing Sepal.length and Petal.width.

This intern was assigned his first task in his new company. From browsing data, predict the probability for a visitor to come back on the website within the day. The goal of this prediction was to propose a different experience for a user who’s likely to come back.

Clic, clic, clic, shift+Enter : Notebook cell launched, algorithm doing his magic, accuracy calculating…

After a…

Create, write, update, update permission, move to folder with Google Sheets API and Google Drive API (and a cup of tea).

Google Suite is maybe the most used tool in companies. I wonder why it was such a struggle to manipulate the API using Python. Google documentation is complete but misses clear examples. Even the Python Quickstart is complicated for someone who only wants to create a spreadsheet…

I wrote this simple cheat sheet where I will create, write and read a spreadsheet, manage permissions and move it in Google Drive.

For this tutorial, I used Google Colab, you can create a new notebook right now and follow along.

Create a service account

To use Google Sheet & Drive API, we need to create a…

and a cup of tea

This article shows how to use a set of tools available in Google Cloud Platform that allows your team to set up a development + automated build + continuous delivery pipeline using Google Cloud Engine + Google Cloud Build.


  • A server running on Google Cloud Compute Engine.
  • PM2 installed and running

If you don’t have that, follow this brilliant and easy tutorial to deploy a node application online.

What is CI/CD ?

CI/CD stands for Continuous Integration and Continuous Deployment.

Big words for an easy concept : Automated deployment and testing.

Continuous integration improves collaboration and quality.

Teams using…

Getting started with a minimal Node.js “ping” app

This post shows how to deploy a minimal Express.js application to production from zero. It could be the starting point of any node project.

Minimum installation. Minimum code.

Skeleton code

I hate tutorials like “How to build <a-super-long-and-difficult-task> in only 5 minutes !” when I actually have to clone a repository and run npm start.

OK, it works (sometimes it doesn’t).

But I didn’t learn anything.

Therefore, I provide the minimal Node.js skeleton to start with simple stuffs.

This server only returns “pong” when http://localhost:8080/api/v1/ping is called.

That’s all.

Go 100x faster for simple detection tasks.

For a recent project, I had to detect IMEIs (International Mobile Equipment Identity) from an image.

The IMEI could be printed or handwritten. I had to think about every possible cases.

My first idea was to build a neural network using tensorflow : build a model, define the architecture, train it on the MNIST dataset, think about the best parameters, export the learning weights… Then I would have to receive and process the input image, detect digits among the text and recognise them to somehow extract what should be an IMEI. …

Going through the process, solving common errors

I started many React Native apps from scratch using Windows and Mac OS for Android and IOS devices. My first advice : be ready for some tough time.

What is React Native?

React Native is an open source framework for building Android and iOS applications using React.

A key concept is Native Components. Before React Native, there was Android developers and iOS developers, developing in Kotlin, Java, Swift or Objective-C. Now, a Javascript developer can create a mobile app running on iOS and Android. …

Start from the ground. Finish in the cloud.

No BS: in few minutes you will have a React app running online.

No coding knowledge required, except knowing what “terminal” means.

We will build, from scratch, an app that send a temperature to a Cloud Function and receive a response message, displayed to the user ONLINE.

Very basic things. I believe starting with simple case is the best way to really understand concepts before going further.


I recorded the full process here.


  1. Very very fast version : Command lines only
  2. Very fast version : Step by step with screenshots
  • 2.1 Requirements

Photo by Dan Dimmock on Unsplash

Il y a maintenant 5 ans, je commençais mes études d’ingénieur, plein de doutes, de craintes, sans savoir ce qui ressortirait de cette machine à façonner des cerveaux. Pendant 4 ans et demi d’élevage industriel, j’ai finalement produit un foie gras de connaissances qui m’a permis d’apposer le tampon “Data Scientist Intern” (notez les majuscules) sur mon CV. Le Saint-Graal pour les étudiants comme moi qui ont façonné un imaginaire du monde des données.

Après avoir englouti des manuels de PascalCC++SQLHTMLPHPCSSPostgreSQLJavascriptReactNodeGitMongoDBPythonRAWSRESTBERTCNN et autres gros mots de l’informatique, j’ai pu découvrir la vraie utilisation de ces outils dans un cadre professionnel.

Quand notre président a un coup dans le nez…

Françaises, Français, Mes chers compatriotes. Je veux ce soir remercier les services de sécurité intérieure par l'amélioration des mécanismes existants pour les contrôles frontaliers et douaniers dans les zones frontalières communes. À cet égard, une avancée significative a été réalisée en mars 2019 par la signature d'un accord cadre intergouvernemental visant à favoriser la coordination et l'harmonisation stratégique de leurs activités respectives dans le secteur public. Donc c'est en réflexion très active là ces jours ci. Réflexion très active ça veut dire quoi ? Ça veut dire que ce seront véritablement…

Oslo ❤ — Photo by Arvid Malde on Unsplash

Détecter automatiquement les sujets d’un corpus, c’est possible avec le CorEx (Correlation Explanation) !

Un compromis entre le topic modeling non supervisé et la classification supervisée, qui suppose une labellisation préalable : le topic modeling semi-supervisé.

Le topic modeling semi-supervisé permet aux data scientists d’ajouter leur compréhension, leurs connaissances des données, et d’aiguiller l’algorithme de reconnaissance de sujets avec des mots-clés.

A la différence des LDA et NMF, connus, CorEx permet l’ajout d’ “anchor words” qui représentent des sujets particuliers. …

Beranger Natanelic

Data scientist jr & Javascript + Python full-stack developer… ok I like tech and do all kind of stuff

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store