Predicting-Soccer-Match-Winners-With-ML

Creating a Machine Learning Model to Predict the Outcome for La-Liga Soccer Matches based on Data Scraped Over The Past 3 Seasons.

Project Overview:

Skills: Python3, Machine Learning, Random Forest Classifier Algorithm, Data Scraping, Jupyter Notebook

Imports: Beautiful Soup, Pandas, Random Forest Classifier, Requests, SKLearn

Phase 0: Picking the Data

-My Favorite Soccer Team is F.C Barcelona, they play in the Spanish La-Liga leauge.
and so I decided to use Data from all La-Liga squads https://fbref.com/en/comps/12/La-Liga-Stats 
of the past 3 seasons (2018-2023)
to collect my data to use for my model I will be creating/training. 

Phase 1: Data Scraping

-Because Data is messy and unorganized, 
I scraped the Data from past 3 seasons 2018-2023 to
better extract and prepare it for the model to read information efficently. 

-Started off with getting standings url for the La-Liga season from FBREF.com
-used Beautiful Soup import to take in the data text from Fbref.com to visualize it better
-Extracting URL and HTML Statistics into our Data from Each Team. 
-Using Pandas Import to Read table out of HTML and making it into a data frame 
-Scaped Data out of the table in the standings page and to get each individual squad page into our data.
-Then Scraped to get Scores and Shooting stats for each Match.
-Combined all the data frames together.
-Make a loop to go through the past seasons data for each year and for each team and then scrape the data into a single data frame
-Turned data into a csv called "Matches.csv"

Phase 2: Creating our Model

Used Our Scraped Data from the past 3 la liga seasons 2018-2019, 2020-2021, 2022-2023.
-Used a Random Forest Classifier to create a model that takes in the data from the sets in order to make predictions on outcomes and compare actual results.
-Used Predictors from the data given to train our targeted model. 
-Increased Precision and Accuracy from the model by creating new predictors and using the rolling averages and combining --
-Each the object to increase prescision and accuracy.
-Taken into the matches played in September our model has 
an accuracy of and prescion score of about 71%, when looking at the combined data frame.

--

-Bryan Aguirre