Predicting FIFA players value

Inspiration

The main reason why I created this project is to understand what makes a football player more expensive based on different skills. Can we find the new Messi on his very early stages?

Kaggle offers excellent datasets to start diving into the data : https://www.kaggle.com/ketanp79/fifa-19-predicting-players-market-value\

Jupyter notebooks are quite good to create and run the model as you can iterate and print the result easily. To interact with the machine learning model , I have used tkinter as the main library.

Overview

This instructions will allow you to make a copy of the project locally.
Check test execution to understand how to deploy and execute the Model_for_app.ipynb.
I have created 2 extra files as a bonus so you can understand better the machine learning models and make investment decisions

You can find the following executable files:

• Model_for_app.ipynb (Main_tool) : explains the dataset, does a random forest regression and creates an app that allows you to get the value of a player based on certain parameters.
• Bonus.ipynb (Bonus_models) : explains the dataset and does more complex machine learning models.
• under_rated_players.ipynb (Bonus_models): explains the dataset, does a random forest regression and creates a csv with the value predicted vs reality. It includes a link to a tableau dashboard so you can interact with the list and take the best decision.

Folder structure

└── project
    ├── Bonus_models
    ├── Environment
    ├── Images
    ├── Main_tool
    ├── Raw_data
    ├── README.md
    │  

---

Pre-requisites:

• Python 3.7 (at least)

• Sklearn

• Pandas

• Seaborn

• Matplotlib

• os

• Time

• Conda

• Tkinter

You can find the conda environment on the file : environment.ylm

Instalation:

As a rule of thumb: I do encourage you to install conda, create a working environment and download any library (mentioned above) doing : "conda install..."

Example: "conda install numpy"

Test execution:

Any execution must be done from a jupyter notebook so both the model and the app works perfectly. The machine learning model will run automatically and after it has learned , you will be able to input data in the app.

Usage

1- Run all the jupyter notebook
2- An app will pop up, input the parameters based on the following intervals:

• Parameters range:

    - 'age' : 16-46 --> top rated players are between 23-30 years old
    - 'height_cm': 155-205
    - 'weight_kg': 49-110
    - 'nation_jersey_number': 1-99 --> best players tend to wear 1,9 or 10
    - 'shooting': 14-93 --> top forwards are close to 93
    - 'passing': 20-93 --> top midfielders are close to 93
    - 'dribbling': 22-96 --> top forwards are close to 96
    - 'defending': 14-90 --> top defenders are close to 90
    - 'debut_year': 1987-2020 --> assuming that on average they debut at the age of 18
  

3- Press the button and you will get the time of the execution and the player value
4- Known issues : the model does not accept blank inputs
5- Limitation: the model does only accept numerical values coming from the dataset. Ideally you can use dummy variables to increase R2 but you will not be able to create a usable GUI

To Do

• Better explore the player position as it will probably improve the understanding of the model

Thanks and contact info

Getting help from Julia Roch ,Javier Molins, Octavio García & Pedro Muñoz. Please do email me for further questions : jcglez93@gmail.com