Machine learning Score prediction for NBA games

In this repository, I created a machine learning system to predicte Nba games results based on the historical data taken from Basketball Reference

I treated this problem as a Regression model, in order to predict the score for each team

Data Description:

As I said the data was taken was taken from the the link above as csv files I grouped the data of each year in a single file, which mean i created four files:

2014-2015.csv
2015-2016.csv
2016-2017.csv
2017-2018.csv

(It is possible to add more files of course) However the files must respect the structure bellow:

Visitor/Neutral	Home/Neutral	Date	Start (ET)	PTS(Visitor)	PTS(Home)
Houston Rockets	Los Angeles Lakers	Tue Oct 28 2014	10:30 pm	108	90
Orlando Magic	New Orleans Pelicans	Tue Oct 28 2014	8:00 pm	84	101
Dallas Mavericks	San Antonio Spurs	Tue Oct 28 2014	8:00 pm	100	101

Features Extraction

As known, each machine learning model need features to be represented as numbers to make it possible to predict the values needed. (PTS Home and PTS Visitor) the Date and time are easy to be represented as numbers. so for example we can code the day of weeks (Mon, Tue, Wed, ...) with Integers from 1 to 7 (1, 2, 3, ..., 7) the same thing for the months (Jan, Feb, Mar,...) we'll code it with Integers from 1 to 12 (1, 2, 3, ..., 12)

However the big challenge is how to code the the name of each team ? !

If we code it with same method that we did previously, for sure we gonna have wrong predictions. When we coded the Days of Week with Integers, it was significant... Monday was coded with "1": meant that monday is the first day of the week Tuesday was coded with "2": meant that Tuesday is the second day of the week..etc

If we do the same thing with the teams Huston Rockets will be coded as 1 Los Angeles will be coded as 2, etc....

But why those teams will have those values and not another random ones ?, it means nothing for our model. So unfortnantly, this method is not a good one to code our teams.

There are many ways to code Strings. In our case, I choosed a method called One hot encoding

One Hot Encoding

This method is also termed as dummy coding. Using this method, dummy columns are created for each class of a categorical attribute. For each dummy attribute, the presence of the class is represented by 1 and its absence is represented by 0. Example: Suppose that we have the only 3 teams with the initial data:

Visitor/Neutral	Home/Neutral	PTS (Visitor)	PTS(Home)
Team A	Team B	70	83
Team B	Team C	90	50

It will be transformed as the table bellow:

Team A	Team B	Team C	PTS (Visitor)	PTS(Home)
1	1	0	70	83
0	1	1	90	50

Note ! we can change the one values to other values that differs to Zero so for example if we want to code the idea of Home teams and Visitor Teams we can give 2 for the visitors , and 1 for Home teams:

Team A	Team B	Team C	PTS (Visitor)	PTS(Home)
2	1	0	70	83
0	2	1	90	50

As known the NBA league contains 30 teams, therefore 30 teams columns will be added

At the end of the preprocessing part of this project we'll have data represented as follow:

Team A	Team B	Team C	...	Day of week	Month	Day of Month	Year	Start (ET)	PTS (Visitor)	PTS(Home)	Pr(Visitor)	Pr(Home)
2	1	0	...	3	7	23	2	8.50	90	78	1	0
0	2	1	...	4	7	24	2	10.00	80	102	0	1
...	...	...	...	...	...	...	...	...	...	...	...	...

Requierements

You'll need to SBT project manager, Follow the instructions here You'll need also to add Spark libraries to your Build file

Note ! Spark does not support Scala 2.12. Here is my build file:

name := "Spark-trainnings"

version := "1.0"
scalaVersion := "2.11.12"
resolvers += Resolver.sonatypeRepo("releases")

libraryDependencies ++= Seq(
  "org.apache.spark" % "spark-core_2.11" % "2.2.0",
  "org.apache.spark" % "spark-sql_2.11" % "2.2.0",
  "org.apache.spark" % "spark-mllib_2.11" % "2.2.0",
  "org.apache.spark" % "spark-graphx_2.11" % "2.2.0"
)

In the project I created a MLConfig Class to allow to config all the parameters for our predicting model:

object MLConfig {

   val NUM_ITERATIONS = 100
   val STEP_SIZE = 0.01

   val HOME_TEAM = 82
   val VISITOR_TEAM = 83


  final val JAN_MONTH = 1
  final val FEB_MONTH = 2
  final val MAR_MONTH = 3
  final val APR_MONTH = 4
  final val MAI_MONTH = 5
  final val JUN_MONTH = 6
  final val JUL_MONTH = 7
  final val AOU_MONTH = 8
  final val SEP_MONTH = 9
  final val OCT_MONTH = 10
  final val NOV_MONTH = 11
  final val DEC_MONTH = 12


  final val MONDAY   = 1
  final val TUESDAY  = 2
  final val WEDNSDAY = 3
  final val THURSDAY = 4
  final val FRIDAY   = 5
  final val SATURDAY = 6
  final val SUNDAY   = 7

  final val YEAR_2014 = 1
  final val YEAR_2015 = 2
  final val YEAR_2016 = 3
  final val YEAR_2017 = 4
  final val YEAR_2018 = 5
  final val YEAR_2019 = 6



}

mallow66 / ML-Scores-Prediction-Spark-Scala-MLlib