NBA Prediction Project

The objective of the project is to create a predictive model that predicts players for the All-NBA Team and All-Rookie Team based on statistical data.

Results ALL-NBA

All-NBA Team	Player 1	Player 2	Player 3	Player 4	Player 5
First Team	Giannis Antetokounmpo	Luka Doncic	Jayson Tatum	Anthony Davis	Shai Gilgeous-Alexander
Second Team	Anthony Edwards	Kevin Durant	LeBron James	Nikola Jokic	Paolo Banchero
Third Team	Jalen Brunson	De'Aaron Fox	DeMar DeRozan	Domantas Sabonis	Devin Booker

Results ALL-Rookie

All-Rookie Team	Player 1	Player 2	Player 3	Player 4	Player 5
First Team	Victor Wembanyama	Chet Holmgren	Brandon Miller	Keyonte George	Scoot Henderson
Second Team	Jaime Jaquez Jr.	Amen Thompson	Brandin Podziemski	Cason Wallace	Ausar Thompson

Methods Used

• Loading data from a CSV file
• Data preprocessing including handling missing values, standardization, and feature selection
• Modeling using Random Forest Classifier for player classification
• Evaluation of different models including Random Forest, Support Vector Regressor (SVR), and XGBoost

Files Overview

• all_nba.ipynb: Jupyter Notebook containing the data analysis process including data loading, preprocessing, modeling, and evaluation.

Plot and Statistics Descriptions:

1. Number of All-NBA Nominations for Top 20 Players:

   • Displays the number of All-NBA nominations for the top 20 players with the highest number of nominations.

2. Feature Correlation Matrix:

   • Illustrates how various features correlate with each other, aiding in feature selection and understanding feature relationships.

3. Features Most Correlated with All-NBA Nomination:

   • Presents the features (player statistics) most correlated with All-NBA nomination, helping identify key predictors.

4. Average Age of All-NBA Nominated Players in Each Season:

   • Shows the average age of players nominated for All-NBA in each season, indicating if age influences nomination chances.

5. Teams with the Most Players in All-NBA:

   • Displays teams with the highest number of players nominated for All-NBA, highlighting teams with significant impact.

6. Teams with the Most Players in All-NBA in Each Season:

   • Shows teams with the highest number of players nominated for All-NBA in each season, indicating changes in dominant teams over time.

2. Implementation

check files exist

• A function check_files_exist checks if the required files are present in the specified directory. It returns a list of missing files if any.

load data

• The load_data function reads the necessary CSV files into pandas DataFrames for further processing.

preprocess_data

• The preprocess_data function prepares the player statistics data by performing the following steps:

• Dropping Unnecessary Columns: Columns that are not needed for analysis are removed.
• Converting Data Types: The "GP" (games played) column is converted to integers to ensure proper numerical operations.
• Filtering Players: Only players who played more than 40 games in a season are retained.
• Filtering Seasons: The data is filtered to include only the specified seasons.
• Adding All-NBA Nominations: A new column "ALL_NBA_NOMINATION" is added to indicate whether a player received an All-NBA nomination. This column is initially set to 0 for all players.

features and target variables

• The target variable, "ALL_NBA_NOMINATION", is separated from the features.
• Columns that are not needed for the model are removed from the features DataFrame.
• The "DRAFT_YEAR" and "DRAFT_NUMBER" columns are converted to integer type, with undrafted players assigned a value of -1.
• Categorical data in the "TEAM_ABBREVIATION" column is converted to dummy variables.

standardize function

• The standardize_features function standardizes the features using a StandardScaler. This ensures that all features have a mean of 0 and a standard deviation of 1.

train random forest

• The train_random_forest function trains a Random Forest model. The dataset is split into training and testing sets, the model is trained on the training set, and predictions are made on the testing set.

predict for a new season

• The predict_new_season function uses the trained Random Forest model to predict All-NBA nominations for a new season. It preprocesses the new season's data similarly to the training data, standardizes it, and makes predictions.

generate award predictions

• The generate_award_predictions function assigns predicted players to different All-NBA teams based on their predicted probabilities.

save results and model

• The save_results function saves the predicted results to a JSON file.
• The save_model function saves the trained model and the scaler to a file using pickle.

main execution

• The main checks for missing files, loads the data, preprocesses it, trains the model, makes predictions, and saves the results.