Data

Primary Data -

• The primary data will be sourced from the pre-approved Football API [https://www.football-data.org/documentation/quickstart]

Secondary Data -

• We’re exploring additional relevant and supporting data to enrich the primary API data.

Data Collection -

• API requests and inputs such as excel, csv, JSON, etc.
• Data Cleaning and Transformation -
• Perform data-wrangling using dplyr, tidyr from Tidyverse package to:
  1. Drop NULL values
  2. Ensure tidy principles are intact
  3. Join different data frames together for analysis

Problem Description

Predict the score and outcome of football matches, using in-game match events rather than the number of goals scored by each team.

What makes the problem result interesting?

• Match results are can be difficult to predict with surprises always popping up.
• A potential approach that can be explored is, by using in-game statistics that can provide us interesting and deeper perspective rather than the simple match results also, creating an opportunity to look beyond the match result itself.

Analytics plan

Exploratory Data Analysis:

• Data summary and visualization would be achieved using various functions, such as str(), summary().
• Using Themes with ggplot2 will contribute in making the visualizations more appealing.

Analysis:

• H2o or Keras will be leveraged to carry out the data analytics.

Evaluation Plan

• The results will be evaluated based on how close the predicted match results are with the actual match results.
• Metrics such as RMSE, R-squared, accuracy, confusion matrix will aid in evaluating our overall model.