NBA Longevity Classification

The National Basketball Association (NBA) is a professional basketball league in North America. The league is composed of 30 teams (29 in the United States and 1 in Canada) and is one of the four major professional sports leagues in the United States and Canada. It is the premier men’s professional basketball league in the world.

Problem Statement

Career longevity is dependent on various factors for any players in all the games and so for NBA Rookies. The factors like games played, count of games played, and other statistics of the player during the game.

Objective

Using machine learning techniques determine if a player’s career will flourish or not.

Code and Resources Used

Python Version: 3.7
Packages: pandas, numpy, sklearn, matplotlib, seaborn, xgboost, lightgbm
Data Source: Data World Data Link: https://data.world/exercises/logistic-regression-exercise-1

Data Dictionary

The values for given attributes are average over all the games played by players

GP: Games Played

MIN: Minutes Played

PTS: Number of points per game

FGM: Field goals made

FGA: Field goals attempt

FG%: field goals percent

3P Made: 3 point made

3PA: 3 points attempt

3P%: 3 point percent

FTM: Free throw made

FTA: Free throw attempts

FT%: Free throw percent

OREB: Offensive rebounds

DREB: Defensive rebounds

REB: Rebounds

AST: Assists

STL: Steals

BLK: Blocks

TOV: Turnovers

Target: 0 if career years played < 5, 1 if career years played >= 5

EDA

I looked at the distributions of the data. Below are a few highlights.

Dataset Shape: (1101, 20)

	Name	dtypes	Missing	Uniques
0	GP	float64	0	274
1	MIN	float64	0	514
2	PTS	float64	0	392
3	FGM	float64	0	289
4	FGA	float64	0	366
5	FG%	float64	0	480
6	3P Made	float64	0	132
7	3PA	float64	0	216
8	3P%	float64	0	349
9	FTM	float64	0	258
10	FTA	float64	0	285
11	FT%	float64	0	554
12	OREB	float64	0	243
13	DREB	float64	0	281
14	REB	float64	0	306
15	AST	float64	0	279
16	STL	float64	0	219
17	BLK	float64	0	201
18	TOV	float64	0	241
19	Target	int64	0	2

Data Pre-Processing

1. Implemented the Quantile Transformer.

Model Building

Split the data into train and tests sets with a test size of 20%.

I tried three different models and evaluated them using Accuracy.

Models:

• Logistic Regression
• XGBoost Classifier
• Light Gradient Boosting Classifier

Model performance

The Light Gradient Boosting model outperformed the other approaches on the test and validation set.

	Model	Accuracy
0	LGBM Classifier	74.66%
1	XGBoost Classifier	71.04%
2	Logistic Regression	70.58%