MACHINE LEARNING FOR SUBSIDY ESTIMATION IN HIGHER EDUCATION

MACHINE LEARNING FOR SUBSIDY ESTIMATION IN HIGHER EDUCATION SESSION 10015 JUL 10 2018 SOUTHERN AFRICA ALLIANCE 9 -11 JULY 2018

PRESENTER Innocent Mamvura University of Witwatersrand Innocent. Mamvura@wits. ac. z a Is a Data Scientist with experience in Artificial Intelligence and Machine Learning, Deep Learning, data extraction, preparation, analysis, modeling, data warehousing, and visualizations using Data Science tools. He has also worked on projects involving unstructured datasets, customer prediction models, and market basket analysis. He holds a BSc (Hons) Mathematics, MSc Actuarial Science and Statistics SOUTHERN AFRICA ALLIANCE 9 -11 JULY 2018

UNIVERSITY OF WITWATERSRAND Has a student body of at least 39 k, with 5 faculties. The university is situated in Johannesburg. SOUTHERN AFRICA ALLIANCE 9 -11 JULY 2018

OVERVIEW OPTION 1 – LIST 1. INTRODUCTION An overview of the South African Higher Education enrolment, graduation, subsidy numbers 2. ORACLE MACHINE LEARNING An overview of the oracle machine learning platform 3. MODEL EVALUATION An overview of the machine learning model evaluation 4. SUBSIDY ESTIMATION Application of the model on the subsidy calculations SOUTHERN AFRICA ALLIANCE 9 -11 JULY 2018

INTRODUCTION This section discuss the abstract, definitions and challenges facing higher education SOUTHERN AFRICA ALLIANCE 9 -11 JULY 2018

ABSTRACT The South African Higher Education is facing challenges when it comes to subsidy estimations. Students drop out of university and some take longer to graduate impacting the amount of subsidy that the institution will claim from the department of higher education. Many institutions today, decision leaders are often left to make financial decisions in the dark without proper systems in place. Higher education finance is often viewed as a “black box, ” with revenue generation, spending, and the monitoring of student outcomes often taking place separately from each other. This paper proposes a machine learning system that predicts completion in minimum time using Oracle Machine learning technologies. We then calculate the subsidies based on the system and advise planning officers and decision makers SOUTHERN AFRICA ALLIANCE 9 -11 JULY 2018

DEFINITIONS Artificial Intelligence is the ability of a digital computer or computer-controlled robot to perform tasks commonly associated with intelligent beings. Machine Learning is the science of getting computers to learn and act like humans do, and improve their learning over time in autonomous fashion, by feeding them data and information in the form of observations and real-world interactions Oracle Machine Learning is a collaborative web-based interface that provides a development environment to create data mining notebooks where you can perform data analytics, data discovery and data visualizations SOUTHERN AFRICA ALLIANCE 9 -11 JULY 2018

CHALLENGES FACING HIGHER EDUCATION Rising Cost of Education Declining Completion Rates Dropout and Exclusion Rates New methods and Curricula Rethinking the role of educators SOUTHERN AFRICA ALLIANCE 9 -11 JULY 2018

SOUTH AFRICAN HIGHER EDUCATION HEADCOUNTS Student Headcount 2005 -2017 1200000 1000000 938201 953373 892936, 00001 800000 837779 799387, 00002 761090, 00001 735073 741380 1037024 983698 969154 985212 975837 600000 400000 200000 0 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 SOUTHERN AFRICA ALLIANCE 9 -11 JULY 2018

SOUTH AFRICAN HIGHER EDUCATION GRADUATES Student Graduation Rates 2005 -2017 25, 0% 20, 8% 20, 0% 16, 4% 16, 8% 16, 6% 2005 2006 2007 2008 17, 3% 17, 2% 17, 1% 17, 4% 2009 2010 2011 2012 18, 4% 19, 1% 19, 4% 2014 2015 19, 9% 15, 0% 10, 0% 5, 0% 0, 0% 2013 2016 2017 SOUTHERN AFRICA ALLIANCE 9 -11 JULY 2018

UNIVERSITY OF WITWATERSRAND Wits Graduation Rates 2005 -2017 30 25 21 21 21 23 22 23 2011 2012 2013 21 21 24 23 23 2014 2015 2016 24 19 20 15 10 5 0 2005 2006 2007 2008 2009 2010 2017 EMEA ALLIANCE 11 -12 OCTOBER 2016

GOVERNMENT FUNDING SOUTHERN AFRICA ALLIANCE 9 -11 JULY 2018

REAL GOVERNMENT UNIVERSITY SUBSIDY PER STUDENT IN SA SOUTHERN AFRICA ALLIANCE 9 -11 JULY 2018

ORACLE MACHINE LEARNING This section discuss the oracle machine learning framework SOUTHERN AFRICA ALLIANCE 9 -11 JULY 2018

ORACLE MACHINE LEARNING ENVIRONMENT SOUTHERN AFRICA ALLIANCE 9 -11 JULY 2018

ORACLE MACHINE LEARNING SOUTHERN AFRICA ALLIANCE 9 -11 JULY 2018

NEURAL NETWORK The Artificial Neural Network receives input from the external world in the form of pattern and image in vector form Each input is multiplied by its corresponding weights. Weights are the information used by the neural network to solve a problem. Typically weight represents the strength of the interconnection between neurons inside the neural network. The weighted inputs are all summed up inside computing unit (artificial neuron). In case the weighted sum is zero, bias is added to make the output not- zero or to scale up the system response. Bias has the weight and input always equal to ‘ 1’. The sum corresponds to any numerical value ranging from 0 to infinity. In order to limit the response to arrive at desired value, the threshold value is set up. For this, the sum is passed through activation function SOUTHERN AFRICA ALLIANCE 9 -11 JULY 2018

EXAMPLES OF NEURAL NETWORK EMEA ALLIANCE 11 -12 OCTOBER 2016

MODEL EVALUATION This section discuss the inputs, outputs and the model evaluation SOUTHERN AFRICA ALLIANCE 9 -11 JULY 2018

INPUTS AND LABELS EMEA ALLIANCE 11 -12 OCTOBER 2016

MODEL EVALUATION EMEA ALLIANCE 11 -12 OCTOBER 2016

SUBSIDY ESTIMATION The section discusses the application of the machine learning model on calculating subsidy estimates SOUTHERN AFRICA ALLIANCE 9 -11 JULY 2018

SUBSIDY CALCULATION EMEA ALLIANCE 11 -12 OCTOBER 2016

SUBSIDY ESTIMATION MODEL Actual. Subsidy+TNSubsidy 13 426 999 2014 7 190 415 11 296 243 2013 5 666 400 11 445 658 2012 5 932 397 8 986 827 2011 4 431 005 0 2000000 4000000 6000000 8000000 Actual Subsidy+TNSubsidy 10000000 12000000 14000000 16000000 Actual Subsidy SOUTHERN AFRICA ALLIANCE 9 -11 JULY 2018

SUMMARY AI and Machine can be used to improve subsidies estimations Failed courses is the most important feature followed by Passed courses and Supplementary exams. The model accuracy is 94% and can be used for budget planning purposes. SOUTHERN AFRICA ALLIANCE 9 -11 JULY 2018

CONCLUDING THOUGHTS ANY QUESTIONS? SOUTHERN AFRICA ALLIANCE 9 -11 JULY 2018

THANK YOU! SOUTHERN AFRICA ALLIANCE 9 -11 JULY 2018