dipunj / ScorePredictor

Machine Learning codes

MachineLearning

This repository contains code for solving the ScorePredictor machine learning problem. The problem Score-Predictor was my first learning problem and hence I didn't use library frameworks. Instead I wrote them by myself in Octave. Others (hopefully in future as well) are written in python or other higher level programming languages.

---

I. File List

---

S.no	Name	Type	Task
1.	Driver.m	MATLAB Script	The main script file containing the main program.
2.	cost.m	Function	Evaluates cost.
3.	expo.m	Function	exponentiate a particular column of a matrix.
4.	import_feature.m	Function	import the data from training.csv...Preprocesses the data.
5.	import_target.m	Function	Imports the G1 G2 G3 i.e target vectors into a numeric matrix.
6.	PlotFig.m	Function	Generates plots for the predicted and actual values for testing purposes
7.	Scaling.m	Function	Causes a particular feature in the feature matrix to be Scaled using Normalised mean
8.	training.csv	.csv(comma delimited)	Column 5 has been preprocessed according to preprocessing scheme
9.	training_set_q1.csv	.csv(comma delimited)	The original(RAW) training file.
10.	test.csv	.csv(comma delimited)	Column 5 has been preprocessed according to preprocessing scheme.
11.	test_set_q1_upload.csv	.csv(comma delimited)	The Original(RAW) test cases file.
12.	output.csv	.csv(comma delimited)	Contains 3 columns G1 predicted, G2 predicted & G3 predicted respectively.
13.	classifier.csv	.csv(comma delimited)	Contains 3 columns containing parameters for G1 , G2 & G3 respectively.

---

II. System Requirements

---

Software Requirements :

1. MATLAB R2015a or later
2. CSV file reader (MS Excel, notepad etc)
3. ASCII text reader

Hardware Requirements :

None in Specific

---

III. Instructions

---

RUNNING THE PROGRAM

1. Launch 'Driver.m'
   Open MATLAB > Navigate to Driver.m and open it.
2. In the command window type 'run Driver.m' (without the single quotes)
3. Wait for the program to train the model
4. After the program finishes to train the model.
   The command window shall print 'TRAINING FINISHED'.

CUSTOMISING THE PROGRAM

* Changing the Size of the training sample
  ------------------------
	Navigate to line# 14 and 17 of the file 'Driver.m' change the last argument

* Changing the test file
  -----------------------
	Navigate to line# 63 and 64 of the file 'Driver.m' change the file 
	name 'test.csv' to the file which you intend to check against,
	and set the first numerical argument to the STARTROW and second 
	numerical argument to ENDROW of your file.

* Changing Lambda
 ---------------
	========== IMPORTANT =========
	If You intend to use full training set then Reset the value of lambda as follows
	In line 43 of the main script
		- Initially lambda is set to for training set (1:470) - - - - - - - - - - - - - - -: [-10,10,20] 
		- When using full training set, the values of lambda best suited are - - -  - - - -: [-10,-10,20]

* Getting Comparision Plot
  -----------------------
	Remove(Or better convert the lines) 82 - 96
	Decomment lines #63, #98 - #105
	make sure you provide the correct target matrix in line 63.

---

IV. Variable Description for 'Driver.m'

---

S.No.	Variable Name	Description	Initialised Value(If Applicable)
1.	alpha	learning rate	0.001
2.	ctr	counter	1
3.	dim	Dimensions of the feature matrix	-
4.	feature_X	Feature matrix	-
5.	hypo	hypothesis values for each training case	-
6.	i	counter variable of for loop	1
7.	lambda	regularisation constant	-
8.	m	no. of training cases	-
9.	Mean_Abs_Error	Contains Value of the mean error	-
10.	myans	contains alternating result	-
11.	n	No of features in the model	-
12.	prdct_G	contains predicted values of subjects	-
13.	t_feat	feature matrix for unused test cases	-
14.	size_t_feat	size of t_feat	-
15.	t_G	target matrix for unused test cases	-
16.	temp	temporary variable used for intermediate data handling	-
17.	target_G	target matrix for training set	-
18.	theta_G	parameter matrix...stores the trained parameters.	-

---

V. Preprocessing Scheme

---

Column No.	Value Input	Assigned to
1.	NITA	1
	IIITA	0
2.	Female	1
	male	0
4.	Urban	1
	Rural	0
5.	GT3	1
	LE3	0
6.	T	1
	A	0
9.	other	0
	at_home	1
	health	2
	service	3
	teacher	4
10.	other	0
	at_home	1
	health	2
	service	3
	teacher	4
11.	other	0
	course	1
	home	2
	reputation	3
12.	other	0
	father	1
	mother	2
16 - 23.	yes	1
	No	0

---

EOF

---