Football - Web Scraper

(Github)

Development

Start: 2021.

End: In Progress

Project description

Python files to extract online player data from the database of “SoFifa” and “SoccerWiki” sites into .csv files, to be used in my Football Manager game.

The file takes data from the sites “SoFifa” and “SoccerWiki” and using Python’s BeautifulSoup looks for Html tags with the necessary information of each player such as name, age, position and potential to save in a “.csv” file. Finished, your process generated files can be imported and used in the football manager application that I am developing with more than 400 teams always updated.

Football Champions - Web Scraper (Wikipedia)

Development

Start: 06/2022.

End: In Progress

Project description

In this project I take data from all the football leagues in the world using Wikipedia and classify the teams in their respective positions. With that I can make predictions of the future positions of the clubs among other analyses.

Of course, the project is missing data as there are many leagues with confusing structures like Apertura and Clausura and knockout championships that have no defined rankings.

All Football Players Stats in History - Web Scraper (Besoccer)

(Github)

Development

Start: 12/2022.

End: 12/2022.

Project description

Project where I get data from the squads of hundreds of football teams from the besoccer website, since 1900. Then I group all this data into a .csv. There is even a jupyter notebook with a quick analysis of existing data.

Streamlit Python - ML Website

(Website)

Development

Start: 12/2022.

End: 12/2022.

Project description

Project for viewing player data with python, through an interactive and online way through the Streamlit platform. It contains data on the greatest players of all time, such as goals scored, assists, games, among other parameters, making it possible to compare the best players in history.

Football Match prediction

(Website)

Development

Start: 10/2022.

End: 10/2022.

Project description

Using Python, I use a dataset with information about the teams, home club and retrospective of the last 5 matches. From there, I filter the dataset in train_set and test_set and apply feature scaling and the random forest method in sklearn’s train_set to obtain a prediction of the results. By modifying the hyperparameters and observing the Confusion Matrix, Recall, Accuracy and F1-Score metrics, the algorithm was improved for better results.

At the end, a 3D graph is generated comparing the real result with the expected, and with another test dataset from a new round of the Brasileirão, predictions are generated.

The algorithm is still imprecise, both due to the unpredictability of the results and the small dataset with little information.

My Music Preferences Analysis - Dashboard

(Github) (Tableau Link) (Gephi PDF)

Development

Start: 06/2023.

End: 06/2023.

Project description

Interactive menu using python to load the music files. Tableau dashboard of my musical preferences.

Analysis of my songs filtering the data using Python. Obtaining genres from the Last.fm API and visualization of the results using Gephi, Tableau.

Reinforcement Learning with Q-Learning

(Github)

Development

Start: 02/2023.

End: 02/2023.

Project Description

Deep Q-learning experiment to simulate a player hitting penalties. At first a circle (player) would have to hit a ball in the middle of the screen and make the ball cross a line (goal). However, I realized that the algorithm was not appropriate because it deals better with finite spaces, such as boards. For continuous spaces other more recent algorithms must be used.

Uncertainty in Neural Networks

(Github)

Development

Start: 05/2023.

End: 05/2023.

Project Description

Experimental application of Bayesian Neural Networks and other methods to get the uncertainty and the std. dev. in point predictions.