Pubg Finish Placement Prediction
- In this project, we are going to create an accurate machine learning model that can predict the ranking position of players in a PUBG compeition. Players can be on solo or in teams (groupId) which get ranked at the end of the game (winPlacePerc) based on how many other teams are still alive when they are eliminated. In a game, players have access to different accesories and player moves which were recorded as game stats during a competition.
- We are provided with a large number of anonymized PUBG game stats from over 65,000 PUBG games, formatted so that each row contains one player's post-game stats. The data comes in a zip file containing 3 csv files train, test and sample submission. Since the possible placements are between 0 and 1 we will use CatBoostRegressor and LGBMRegressor model plus GridSearchCV to tune our regressors to get the best performing models.
Downloading and Loading the Data
#Let's install and import some required libraries before we begin. !pip install pandas-profiling numpy matplotlib seaborn pyarrow featuretools --quiet
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 324.4/324.4 kB 8.2 MB/s eta 0:00:00 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 618.0/618.0 kB 31.4 MB/s eta 0:00:00 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 352.3/352.3 kB 17.7 MB/s eta 0:00:00 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 235.5/235.5 kB 11.3 MB/s eta 0:00:00 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 102.7/102.7 kB 6.2 MB/s eta 0:00:00 Preparing metadata (setup.py) ... done ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 679.5/679.5 kB 36.0 MB/s eta 0:00:00 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 296.5/296.5 kB 3.0 MB/s eta 0:00:00 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 455.4/455.4 kB 33.7 MB/s eta 0:00:00 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 4.7/4.7 MB 106.7 MB/s eta 0:00:00 Building wheel for htmlmin (setup.py) ... done