Projects | Joffrey Mayer

Visualizing a Database with a UI | Scraping + NLP + Web Development

Tue, 14 Feb 2023 00:00:00 +0000

Full Data-Science Project | Data Analysis + Data Visualization + Deployment

Mon, 25 Jul 2022 00:00:00 +0000

⚠️⚠️ WORK IN PROGRESS ⚠️⚠️

Working with data is a tedious task.

Bringing the data into the format you need to pursue a type of analysis, calibrating your model and using the results to fine tune it can take up to months in order to achieve a satisfying outcome.

Let’s assume that we were able to generate an awesome model that is able to fulfill every possible requirements we wanted it to have. Since models are digital goods, the next logical step would be to create a dashboard, in order for other people to benefit from what you created.

This is where visualization libraries like bokeh or PowerBI come into play. While these tools are perfectly fine to use - especially when using them inside of a . As soon as publishing on the web comes into play, those tools show their limitations:

The performance on a site is slow,
The dashboard is oftentimes not useable on phones and similar smaller devices, resulting in a poor user experience.
The library’s limitations will create boundaries for our story-telling.
No interactions with a user is possible.

Wouldn’t it be great if we could built are own Dashboard from scratch? We would be able to customize it infinitely to our & the user’s taste, as well as .

Well, this is exactly the challenge that I tried to solve. Creating a dashboard that:

Is highly userfriendly, even on smaller screens (like phones),
Tells a story in a highly-amusing way, by abstracting most of the data analysis’ complexity away,
As well as enabling the user to play around and generate his own predictions on the fly.

The best part about this? I created it without any framework, solely with the help of webtechnologies (HTML, CSS and JavaScript). It works on any machine with a modern browser, even locally.

Sure, it took some time and efforts to get the dashboard. And surely, the use of a visualization tool could achieve a dashboard in less time. But would it be able to have the same impact on people, like this unique dashboard? Check it out and answer the question by yourself 😉

✨ DISCOVER DASHBOARD ✨

5 Full Website Projects | Web-Development + Digital Marketing + Deployment

Wed, 27 Apr 2016 00:00:00 +0000

✨ Visit Hifi.Zuerich ✨
✨ Visit Exxeption-Audio.com ✨
✨ Visit Hochwertige-Musikanlagen.ch ✨
✨ Visit OptimizeAudio.com ✨
✨ Visit Audiophil-Dreams.com ✨

A Critical Evaluation of a Behavioral Paper

Wed, 27 Apr 2016 00:00:00 +0000

In the field of media and communication, attention is often drawn to the special role of magazines, television or - more recently - online channels, such as the internet or social media, to disseminate information to a population / across a network. On this basis, I analyzed how important the source of an information (the „messenger“) can be when people formulate a decision - here in the context of choosing to consume normal tobacco cigarettes VS. electronic cigarettes. More specifically, the goal was to present and explain a study’s empirical findings during a 30 minutes presentation.

In the (empirical) study I analyzed, the test subjects - in this case smokers - were given the exact same information, namely: „E-cigarettes are significantly less dangerous than tobacco cigarettes: If you switch to e-cigarettes today, you are likely to live 5 years longer“, but the source of the information - i.e., „the messenger“ - varied, i.e., certain smokers received the message from:

a fictitious e-cigarette company, others from
a health organization, others again from
physicians, or
from a „neutral source“ (e.g. „no messenger“).

As you can guess from the above list of „messengers“, some of these sources seem more trustworthy than others. This is exactly the hypothesis the researchers wanted to test, namely whether the exact same information would have different marginal effects on smokers’ beliefs and their consumption decisions („buy e-cigarette VS. buy tobacco cigarettes“).

As a student, it was then my task to present this data analysis in an understandable way and to analyze the reliability of the research method used (are the final results of the study credible?).

DOWNLOAD MY PRESENTATION

Application of a DiD-Model with R

Wed, 27 Apr 2016 00:00:00 +0000

One of the main goals in criminology is to alleviate violence before it breaks out, in order to maximize a society’s safety. This is achieved by implementing various kinds of preventive measures.

Security topics have always been on the top of many countries’ agenda, especially in the U.S., where there are almost as many guns as there are inhabitants (about 300 millions). Therefore, many U.S.-cities invest in multi-million dollar projects to reduce gun violence in general. Oftentimes, disadvantaged neighborhoods - where crime is more prevalent than in other geographical areas - are targeted.

Using aggregated monthly crime data at the precinct level, I was able - along with a colleague of mine - to analyze & assess the effectiveness of a $12.7 million project in New York that aims to reduce crime by implementing preventive measures.

DOWNLOAD OUR PAPER

Application of a VAR-Model with Matlab

Wed, 27 Apr 2016 00:00:00 +0000

In the field of economic history, historians & economists are interested in the causes of the world’s greatest economic crisis - the Great Depression in 1929.

By using time series data from state archives, I was able to perform a macroeconomic analysis of France’s economy during the interwar period (1919-1938) using a „Vector Autoregressive Model (VAR-Model)“ in order to investigate empirically, whether potential causes, discussed in the literature, were truly relevant.

DOWNLOAD MY PAPER

Deep Learning and Neural Networks

Wed, 27 Apr 2016 00:00:00 +0000

During my studies, I came into contact with the programming language Python. This experience culminated in an empirical project, where a fellow student and I attempted to predict the outcome (win, loss, draw) of football matches in the four largest European leagues over the past 14 seasons with a neural network. This was achieved by using a detailed panel dataset.

DOWNLOAD OUR PAPER

Master Thesis

Wed, 27 Apr 2016 00:00:00 +0000

The main goal of my thesis was to model people’s sport choices over time. More specifically, I tried to answer questions such as why certain types of people tend to develop sport habits, starting from a very young age, while others prefer to quit / start to do sport later / earlier in their lives, respectively. To answer such questions, I used various theoretical approaches from the field of Behavioral Economics, as well as Discrete Choice Methods for the empirical analysis.

DOWNLOAD MY THESIS

Optimizing a Website's Performance by over 250% | My first Webdev Project

Wed, 27 Apr 2016 00:00:00 +0000

R Code Master Thesis

Wed, 27 Apr 2016 00:00:00 +0000

Data Scientists spend a lot of their time (up to 80%) to bring the data into the right form. This crucial step is absolutely necessary, otherwise a data analysis cannot be performed. The Swiss Household Panel, which is the dataset I used for my empirical analysis in my Master thesis, is a good example to illustrate my experience with data processing. For instance, I had to:

re-name & select the variables (derived from the theoretical models that I used),
transform the data from wide- into long-format (and vice versa) in order to distinguish the different family members of a household,
deal with missing values, but also
create new variables based on already existing variables.

As these points illustrate, a considerable part of my Master thesis consisted - among other things - putting this uncleaned dataset into a form that was adapted for the use of the quantitative methods I had selected.

DOWNLOAD MY CODE

Using $\LaTeX$ for Entrepreneurs

Wed, 27 Apr 2016 00:00:00 +0000

Nowadays, it is from utmost importance to be aware of new working frameworks and technologies all the time, otherwise competitors with better adaptability will develop much faster and - ultimately - you lose your position in the market.

In the digital age, companies must be able to establish a dialog with their customers. One possibility to increase the trust in a company and highlight very sophisticated knowledge of a brand, may be through Google reviews.

By providing a rating, today’s customers give future buyers the opportunity to better assess the performance of a company and thereby increase the overall transparency between the demand & supply side. Furthermore, the more a company is able to get high ratings on Google, the better the visibility on the internet will be.

Therefore, being able to incite customers to rate the company, is becoming extremely relevant for the future development of a brand. That’s why I created this very beginner friendly tutorial, in order to maximize the amount of Google review for a company.

DOWNLOAD MY GUIDE