##Introduction

This program is used to predict the result of 2015-1016 NBA regular season using naive bayesian model. This program did not use any public dataset, all the data were downloaded by python web crawler, from NBA reference 'http://www.basketball-reference.com'. This program also contains all the used web crawler.

##Environments

• Operating System: MAC OS X EI Captian
• Language: python
• Software: pyspark
• Lib: numpy, pyspar.mlib, requests, re, BeautifulSoup

##Programs ###NaiveBayesianModel

####NBA_nb.py

• run on spark, using spark-submit.
• from spark_data folder read the training and testing dataset.
• use training dataset to train Naive Bayesian model.
• calculate the accuracy.

outputfiles:

• result.csv record the win/loss of a game in predict regular season 2015-2016
• testoutput.csv record the true result at the leftmost and predicted result at the rightmost of each game. In each match, the team on the left is the home team.

###runVector

####establishFeatureVector.py

• from seasonmatch folder input the matches data of each game in the selected season.
• from leagues folder input the statical data of each team.
• label each matches and replace the team name using weight vectors.
• save the data in vector folder. The *.csv in this folder can directly used in naive Bayesian model.

###web crawler

This folder contains seven web crawler written on python, for downloading different dataset from 'http://www.basketball-reference.com'.

• player_per_game.py: save players' performance in each game.
• player_season_per36m.py: save players' performance in each season per 36 minute.
• team_history_data.py: save teams' history performance.
• matchWithPer.py: save each match's info.
• establishDictionary.py: establish dictionary used in other web crawler.
• player_info.py: save all the players' info.
• player_season_info.py: save players' performance in each season per game.