30 Days ML Challenge (BreakInto Data)

---

Football Goals Prediction Model

Overview

This project focuses on building a predictive model to forecast the number of goals scored by each team in a football match. The model incorporates various data sources, including player ratings and team information, to create a robust prediction system. The process involves data collection, preprocessing, feature engineering, exploratory data analysis (EDA), model training, and deployment.

Table of Contents

1. Data Collection
2. Preprocessing
3. Data Cleaning
4. Exploratory Data Analysis (EDA)
5. Model Training
6. API Development with Flask
7. Dockerization
8. Deployment on Azure

01. Data Collection

Source

Data was collected using Scrapy from Transfermarkt.

Datasets

The following datasets were collected:

• clubs_team_players_v1.json
• national_team_players_v1.json
• matches_v1.json
• players_rating_v1.csv

02. Preprocessing

The preprocessing stage involves reading the collected data, normalizing team and player names, and mapping players to their respective teams. This step ensures consistency and prepares the data for further analysis.

How is the process of Mapping conducted

graph TD
  A[clubs_team_players_v1.csv] --> B[team_to_players mapping]
  C[national_team_players_v1.csv] --> B
  B --> D[matches_v1.json]

  D --> E[matches_with_players.json]

  F[players_rating_v1.csv] --> G[name to rating mapping]
  
  E --> H[football_matches_dataset_v1.json]
  G --> H

Loading

03. Data Cleaning

The data cleaning stage involves handling missing values and ensuring the integrity of the data. This includes:

• Handling missing values by imputing the mean rating for missing player ratings.

04. Exploratory Data Analysis (EDA)

Top 10 Teams with the Most Players	Home Team vs. Away Team Average Rating

Top 10 Players by Frequency of Appearances	Home Team vs Away Team Scores

Home Team Score vs. Home Team Average Rating	Away Team Score vs. Away Team Average Rating

Number of Home vs Away Team Players

05. Model Training

• We use the RandomForestRegressor for prediction due to its robustness and ability to handle multiple output variables efficiently when wrapped in a MultiOutputRegressor.
• Hyperparameters are optimized using RandomizedSearchCV with a predefined parameter distribution and cross-validation

Model Evaluation

The model is evaluated using several metrics:

• Mean Absolute Error (MAE): 1.14
• Mean Squared Error (MSE): 2.4
• Root Mean Squared Error (RMSE): 1.5

06. API Development with Flask

An API is developed using Flask to serve the model predictions. The API allows users to input both teams name and receive predicted scores.

07. Dockerization

The application is containerized using Docker to ensure consistency across different environments. A Dockerfile is created to define the environment and dependencies.

08. Deployment on Azure

The Docker container is deployed on Azure, making the model accessible as an api. Azure provides scalability and reliability for the deployed model.

Final Results:

Football Goals Prediction API

The Football Goals Prediction API provides predictions for football goals based on various data inputs. Below are the details on how to use the API.

Base URL

The base URL for accessing the API is: http://footballgoalspredictionapi.f6cpcjeweuhgacfm.eastus.azurecontainer.io

Endpoints

1. Get Prediction

Endpoint: /predict

Method: POST

Description: This endpoint provides a prediction for football goals based on the provided query parameters.

Query Parameters:

• home_team: Name of the home team (string, required)
• away_team: Name of the away team (string, required)

Example Request:

Using curl command

curl "http://myfootballapi.f7bqcuhbhaenayfk.eastus.azurecontainer.io:5000/predict?home_team=Ac%20Milan&away_team=atalanta%20bc"

Example Response:

{
    "home_team": "Ac Milan",
    "away_team": "atalanta bc",
    "predicted_home_goals": 4,
    "predicted_away_goals": 3
}

• Using Python requests lib

import requests

url = "http://myfootballapi.f7bqcuhbhaenayfk.eastus.azurecontainer.io:5000/predict"  
data = {
  "home_team": "Ac Milan",
  "away_team": "atalanta bc"
}

response = requests.post(url, json=data)  
print(response.json())  

• Using JavaScript fetch API

fetch('http://myfootballapi.f7bqcuhbhaenayfk.eastus.azurecontainer.io:5000/predict?home_team=Ac%20Milan&away_team=atalanta%20bc')
  .then(response => response.json())
  .then(data => console.log(data))
  .catch(error => console.error('Error:', error));

• Using Java HttpClient

import java.net.http.HttpClient;
import java.net.http.HttpRequest;
import java.net.http.HttpResponse;
import java.net.URI;

public class ApiClient {
    public static void main(String[] args) throws Exception {
        HttpClient client = HttpClient.newHttpClient();
        HttpRequest request = HttpRequest.newBuilder()
                .uri(URI.create("http://myfootballapi.f7bqcuhbhaenayfk.eastus.azurecontainer.io:5000/predict?home_team=Ac%20Milan&away_team=atalanta%20bc"))
                .build();

        HttpResponse<String> response = client.send(request, HttpResponse.BodyHandlers.ofString());
        System.out.println(response.body());
    }
}

---