| Project | Description |
|---|---|
| SzazKilencvenKilenc | Forecast Hungarian parliamentary elections using probabilistic programing and machine learning [Hungarian] |
| Chili Classifier | Identify chili type (out of 142) using its photo. Utlizing deep learning with transfer learning. Quantizing model for performance optimization and porting it to JavaScript to run on the user side. Ranking second with the solution on OTDK (Hungarian science competition for university students) in 2021. |
| Bet Master | Predict result of football matches using probabilistic programming |
| Female Diabetes Risk Predictor | Training a random forest model in Scikit-learn/Python to predict the potential risk of female diabetes then converting the model with Sklearn-porter to JavaScript in order to deploy it on the Front-End side of a web app. Click here for the web app. |
| Milan/Budapest Rental | A Scrapy/Python webspyder collects rental and sale advertisements from a major Italian/Hungarian real estate site. Raw data is processed in Pandas/Python to better understand the housing market and provide investment insights. Result are represented in a single page app using CrossfilterJS. Click here for the Milan and here for the Budapest ([Hungarian]) results. |
| ElektoR | Neat Shiny/R webapp to simulate how the modification of certain election rules would have influenced the result of the Hungarian parliamentary elections in 2014. (Note: Hungarian election law was severely modified in 2013) As a spin-off of the project I am working on an R package implementing different vote-to-mandate allocation algorithms. For the deployed version of the app [Hungarian] click here. |
| Portfolinfo | Creating effective investment portfolios. Financial analysis in Python with Monte Carlo analysis and Markowitz portfolio optimization. Click here for the corresponding notebook. |
| Game Title Evaluator | Creating a deep neural network with RNN to predict computer game review scores using only the title of the games. Resulting Flask/Python web app is deployed on heroku and in an android app (using only a simple web container). Click here for the web app, here for the android app and here for the model training notebook. |
| What’s cooking | Shiny/R application based on the “What’s cooking?” Kaggle machine learning challenge. User can input ingredients from a receipe and the app will predict the cuisine. It will also represent in a wordcloud the main ingredients of the predicted cuisine. After preprocessing ingredients are represented as a TF-IDF vector and a random forest was trained for the predictive algorithm. For the deployed version of the app click here. |
| Did He Tweet? | Retired project due to Trump’s ban on Twitter A simple VueJS page served by Flask/Python API to show if Donald Trump tweeted lately. Also displays a wordcloud based on the latest tweets and some basic statistics. A joint project with @Formatted to practice remote AGILE cooperation (@Formatted is based in US). Click here f |