PA-Data-Classification

Problem Statement

Can we develop a predictive model for forecasting the outcomes of International Football/Soccer games based on historical data?

The objective is to to predict the outcome of football matches before-hand by using suitable data about the Countries/Teams that are playing each other.

Data Sources

• Matches dataset from 2006-2018 (Team Country Names, Match Date, Competition Name, Goals scored by both teams)
• Squad dataset of individual competitions (Country, Coach Name, Player Name, Position, Caps)
• FIFA rating (Team Country Names, Match Date, Competition Name, Goals scored by both teams)
• Year-wise Population of Countries (Country, Population from 1960-2022)
• Year-wise GDP of Countries (Country, GDP from 1960-2021)
• Year-wise average years of schooling in Countries (Country, Continent, ISO Code, avg years of schooling from 1990-2021)

Preprocessing

• Replaced Missing Values
• Outlier Detection based on COF
• Normalization
• Encoding (nominal to numeric)

Modelling

Model	Score (%)
KNN	52.21
SVM	53.53
DT	50.46
RF	55.92
GB	56.69
NN	52.70

RapidMiner Screenshots

Folder Contents

.
├── model.rmp               # RapidMiner model
├── Final_Data.csv          # Combined dataset
├── Final_Data_pre.csv      # Combined and pre-processed dataset
├── Data Collection.ipynb   # Code for data cleaning and combining
├── Football                # Football data
│   ├── ...                 # matches, squads, players, fifa ratings
├── Countries               # Country-specific data
│   ├── ...                 # GDP, Population, Average years of schooling
├── Results                 # Prediction files for individual models and Grid Search
│   ├── <model>.csv         # Training file
│   ├── <model>.txt         # Accuracy and Confusion Matrix
├── KNN Scratch.ipynb       # Code for KNN implemented form scratch
└── ...