- Suhas CV
- Mihir Manjrekar
- Gautam Gadipudi
The Raw Data of the project can be found in Data directory.
The Source code is found src directory.
The following external python libraries were used
Installation: execute the codes in the following order to have dataset installed on your local Mongo DB server:
python3 src/load_player.py
python3 src/load_matches.py
python3 src/update_players.py
python3 src/get_partnerships.py
python3 src/extract_score.py
python3 src/runs_per_match.py
We are going to cluster based on the following attributes per match / game:
runsstrike_ratesixesfours
Firstly, normalize the data. Run the below script to get the normalized data into a collection named clustering_data:
python3 src/normalize_batting_stats.py
Then, run the following to get a plot of k (number of clusters) vs SSE (Sum of Squared Errors):
python3 src/Clustering/plot_sse_k.py
Then, select a particular k at the elbow of the plot (saved under ./Visualizations/Clustering) and run the following:
python3 src/analysis/Clustering/cluster.py <k-value> <iteration-limit>
We are going to perform item set mining (apriori) on batting partnerships which has following attributes:
partners(partner-1 ,partner-2)venuetotal runs
Initially, we will filter the partnerships with total runs > 30. The algorithm works sligthly different from usual apriori, here we have fixed number of items in a partnership(transaction), thus the algorithm stops at level 3. The minimum support is set to 15.
- At Level 1 we will get the the individual players/venues involved in atleast 15 partnerships(30+ run).
- At Level 2 we will have the player-venue/player1-player2 involved in atleast 15 partnerships(30+ run).
- At Level 3 we will have the most frequent partnerships(partner1,partner2,venue).
python3 src/analysis/Association/partnershipVenueMining.py