This project is a continuation of mtg_scraper project
This project consists of both classification and regression analysis, for which I applied supervised learning based on data previously collected with my scraper.
The project brief was to:
- Identify an industry relevant prediction problem.
- Develop a solution to this problem.
- Present the results.
Deliverables:
- A GitHub repo containing all code.
- A presentation in two parts:
- Non-technical presentation highlighting the problem and the approach to a solution.
- Technical presentation giving details of the techniques applied during data processing and modelling.
Stakeholders for this project include:
- The company owning the game: The plan was to check for the “fairness” and randomisation of the game to see if there are any biases that can lead to unbalanced gaming which -eventually- can lead to a worse experience for the players and a decline in the interest for the game.
- Myself:
I also checked for a correlation between the values of the cards in the secondary market that could potentially lead to identification of under or over valued cards and the profit potential that could derive.
Data cleaning was performed mostly as part of the mtg_scraper. Additional cleaning was needed in order to drop outlier values or features that had no intrinsic value to the specific project
For the classification analysis, 14 different models were trained and compared with a baseline
For the regressions analysis, I only run a Linear regression to get a benchmark.
Classification models need to be enriched with more parametrisation to check if the results can be improved
Regression analysis needs more models so that a proper comparison between models can be performed and better results can be acquired.
main_cat and main_lin_regr files to be merged to one.