Data is Power

Data is the Fuel of Organizations: Opportunities and Challenges in Afghanistan

A car without fuel cannot be driven; a mobile, a laptop or a PC without power cannot be used; a website without feeding won't have any visitors; likewise, an organization without data won’t stand and cannot be developed.

A decade or two ago fossil fuels/oil were considered the most significant resource in the world. Then, water attracted attention of the world. Nowadays, data is considered the fuel of any country, organization and finally the future – as Dave Coplin, Microsoft Chief Envisioning Officer at Xerocon 2016 said, "Data is the fuel of our future. When you have lots of data, it changes things".

Data is power and without data it is not feasible to produce facts and figures, insights, analyses, mining and finally decisions cannot be made – the term "data" exists and plays the crucial role in Data Analyses, Knowledge Discovery from Data (KDD), Data Mining, Data Science, and Big Data. It is the data that enables organizations to explain the past and predict the future through data science and business intelligence tools.

Since 2003, around 1.7 million eligible high school graduates have attended the National University Entrance Exam (Kankor). Their First Names, Last Names, Father Names, and Grand Father Names are recorded in the Kankor system. However, these data are not used at all and one basic example might be to find out the top-n male and female names that are very common in Afghanistan e.g. more than 10,000 Kankor male participants were named "Zabiullah" and more than 8,000 female participants were named "Fatimeh" or the top-n names at a particular province. It is worth mentioning, that even these numbers are not very accurate because in Dari same names can be written in different variations.

Taking these basic data into account leading us to the following practical examples and applications:

• Autofill Missing Gender Values Using Identical Names: Out of 1.7 million records of Kankor data, for around 200,000 records the gender label is missing. This situation poses a challenge for data quality and completeness and should be addressed systematically through a structured process. The current solution is based on the fact that there are many Kankor applicants with the same first names. We assume that the gender for at least one of the identical first names is identified in the master list. That gender value is used as a reference to autofill the gender for the identical names whose gender value is missing. It is worth mentioning that in some rare cases there are first names used both for men and women – the approach described above may take the gender value of a female candidate and autofill for the missing values for the male candidates, which is false. To address this predicament, other approaches were used to guarantee that the pattern will autofill the missing values properly.
• Audit and Match High School Data with Kankor Data: To find out the relation between high school marks and the success rate of candidates in the Kankor exam, it is required to merge and compare the candidates' high school marks with their Kankor data. Since there are no common keys for these two datasets, it would be efficient to use the combination of following common attributes for comparison: first name, last name, province, high school name and graduation year.
• Generate Fake Data: Sometimes it is required to test the performance of databases or web applications with thousands and millions of fake data; or to test and evaluate the feel and look of web applications; or to create good-looking XML documents; or to anonymize data taken from a production service; and other situations. Around 1,000 male candidate names, 1,000 female candidate names, 1,000 family names, all the 34 province names, and more than 400 educational institution names can be taken into account; then a script can be written to find out their ASCII equivalent and to create random Usernames, Emails, and Websites from the ASCII names and family names. Finally, these data can be prepared as a "fa_AF" provider for "Faker", a PHP library that generates fake data.
• Name Entity Recognition: Another great application of these data is in Text Mining, mainly, Name Entity Recognition (NER). Let's suppose we are asked to detect and identify Person Names within the text. The Kankor participants First Names, Last Names, Father Names, Grand Father Names can be considered as representative of all the names that are used in Afghanistan since they cover almost all the common names. These names can be used as a rich dataset in the corpus for identifying the Person Names from large amounts of unstructured text. Likewise, the Province, District, and Village Names can be used to identify the Locations from a given unstructured text. Moreover, High School Names together with Educational Institutions Names can be used to identify almost all the Educational Institutions from a given unstructured text.

There are large amounts of data available for mining purposes in the context of Afghanistan, mainly, in education domain. But the methods that the educational institutions use to store and produce their data only enable them to achieve basic insights which are not useful for decision-makers or policy-makers and which do not help them to guess the future. Their main efforts are to generate (only) basic facts and figures (e.g., total number of students and teachers categorized by gender, location, and some other criteria). However, it turns out that these simple facts and figures do not help policy makers and educational institutions to improve the educational settings. For instance, they cannot be used to predict the right fields of study for high school graduates, or to identify first year university students who are at high risk of attrition or failure, or to recommend appropriate courses for university students.

It is very important to store detailed data and then to apply Data Mining techniques to explain and analyze the past such as what happened and why it happened as well as to predict the future such as what might happen and what should the organizations do.

From the results it can be concluded that there are potential opportunities for educational data mining application in the domain of Afghanistan's education systems. Examples include:

• Predicting suitable fields of study for high school graduates prior to taking the exam.
• Alternatively, the data can provide the necessary tools to policy makers in shaping the education system by introducing specialized education at the high school level.
• Data also provides early warning systems to intervene in identifying first year university students who are at high risk of attrition caused by both capacity related limitations and other external obstacles.

Lack of data availability and accessibility, lack of detailed data, and lack of experts are the main challenges preventing applicability of Data Mining and Educational Data Mining. Education Management Information System (EMIS) at Ministry of Education (MoE) and Higher Education Management Information System (HEMIS) at Ministry of Higher Education (MoHE) together could be appointed to provide the raw data for Data Mining applications to help discern patterns of abilities and behaviors which could be used to help educational institutions. More importantly, the concept of Data Mining is new, and research has not been done about its applications in Afghanistan. This study will be the beginning of a new era for other researchers and policy analysts both at the private and public sectors.

About Kankor

To enter higher education, high school graduates need to pass the National University Entrance Exam (Kankor, from the French word Concours; English: "Contest") . Kankor is held every year, mainly in the capital and large provinces.

The Kankor exam comprises 160 questions about subjects taught at high school mainly in grades 10 to 12 categorized into the following categories: Mathematics, Natural Science, Social Science, and Languages.

Kankor questions are provided both in Pashtu and Dari, in Pashtu and Dari speaking regions respectively. All the questions are multiple choice, and generally, the participants have two to three hours to answer and choose the proper options. Usually, correct answers are worth one to three points. The maximum number of points are between 320 and 370, but more important is the minimum number of points needed to qualify for a university slot. Applicants can choose five favorite fields of study, and each field of study requires a certain score. In general, admission into popular fields of study like medicine, engineering, computer science, economics, and political sciences and into institutes in the capital and main cities requires higher scores. However, if a student does not score high enough for any of his/her chosen fields of study, he/she is dropped altogether and is not assigned any field of study, a result called result-less (benatedja).

The Kankor candidates could opt for ten fields of study they favor, and for one location for each field of study. They also had three chances to participate in Kankor in lifetime until 2011. Since then, they can opt for only five fields of study and have only two chances in the lifetime.