Keywords

1 Introduction

Nowadays, the student attrition is a problematic that worries to the public and private universities. The causes of this problem have not been accurately identified; besides, the way of using the data associated with students in order to generate useful information that can serve to face this problem is a challenge. According to [8], one of the main difficulties facing the current educational system is the desertion, its accumulated value reaches levels of 45% at the national level; one of its main causes is the desertion of academic type. Some authors [16] claim that the main problem facing the Colombian higher education system is the high levels of student attrition. They argue that the number of students that completed their higher education is very low, perceiving that most of these abandoned his studies in the first semester. And, that half the number of matriculated students in a higher education institution cannot complete their academic cycle. Finally, they say that the student attrition in 2004 was estimated at 49% by exhibing the following causes: economic and financial constraints, low academic performance, vocational and professional disorientation and difficulties for adapting to university environment. According to [1], only one of every two students who enroll in an undergraduate program completes his career. The concern is greater if you consider that 39.52% of who drop out his studies argue economic reasons.

The increase in attrition rates has become a problem that has been of interest to higher education institutions and the educational authorities. This problematic has important socio-economic consequences. The loss of students causes serious problems to the universities, since it makes their sources of income are unstable. In addition, student desertion can compromise the future of a country in the medium and long term because the accumulation of scientific and technological knowledge is one of the factors that determine the socio-economic development of a nation [3]. According to [12], school failure is in any case a catastrophe, absolutely desolating on the moral, human and social level, which very often generates exclusions that will mark to young people during their adult life. Besides, desertion and abandonment produce uprooting, loneliness, absence of rites, lack of routines and loss of negotiating capacity with others, social loneliness”. Then, the authors explain that desertion is a problem of the educational system related with the environments of the same, such as educational environments, family situations, environmental and cultural requirements that directly affect to the deserter.

Several studies based on KDD [6] have been developed. In [8] the authors use KDD in the extraction of knowledge from the database of the academic information system (SIA) of the University of Caldas in order to calculate indicators of academic performance. In [9] useful knowledge is generated in order to find possible causes of the student attrition problem of the autonomous University of Manizales from the large amounts of academic information in the academic registry office. The work presented in [1] contribute to the decision-making to reduce student attrition levels in the undergraduate programs of Mariana University applying the KDD process and based on a unified data repository with the socio-economic, personal, academic and institutional information of the students. Other study [2] makes an analysis of the student attrition in the system engineering program of the Simon Bolivar University. There, the causes of desertion are grouped into variables based on information from the academic registry office in order to establish patterns and support the decision making using the data mining techniques and the tool for automatic learning and data mining called WEKA. In this paper we propose a method for analyzing the academic desertion by using techniques, tools and methodologies based on KDD. Unlike the previous works, we are focused on the students of the system and computer engineering program at the University of Quindío. Besides, we include a set of variables provided by the statistical department, which are the result of a proof that assess different cognitive aspects related to the students. These variables are added to the information provided by others information systems such as register office and planning department. The data were provided by three sources: information concerning to the SPADIES taken from the planning and development office, BADyG test provided by the department of statistics and the personal and academic information of the student, taken from the register and control office of the University of Quindío

According to the problematic of the student attrition, some authors [12] say that is obligation of the educational institutions, especially the universities, to establish academic, administrative and adjustment mechanisms to the university life of its students so that they overcome the difficulties of the academic programs and culminate successfully their careers. In addition to the large volume of data available in the higher education institutions related with its students, in [6] the authors propose that a new generation of computational techniques and tools is required to support the extraction of useful knowledge from the rapidly growing volumes of data. These techniques and tools are related with an emerging field of knowledge discovery in databases (KDD) and data mining. The information that can be obtained from academic databases will serve to answer to such questions as: Which are the causes of the student’s retention in the university? Why do students desert? According to [13] automatic data mining techniques can be applied to solve these questions and to facilitate the development of strategies for improving the academic processes and the educational programs. The data mining can offer for the problematic of the student attrition a great variety of statistical and computational methods to investigate the existence of relations and behavior patterns of students in the first year of their university career [10].

The paper is organized as follows: Sect. 2 describes the main concepts to understand the proposal. In Sect. 3 we discuss some related work. In Sect. 4 we describe the proposal together with methodologies and tools used. And finally, in Sect. 5 some conclusions are presented.

2 Theoretical Framework

2.1 Student Attrition

According to [12], student attrition must be understood as the definitive abandonment of the classrooms for different reasons and the non-continuity in the academic formation for each person who begins his studies, hopeful in ending happily the university studies.

2.2 KDD

The term KDD refers to the overall process of discovering useful knowledge from data. Besides, this process focuses on the search for data patterns that are valid, novel, potentially useful and understandable [6].

2.3 BADyG

The term BADyG refers to a test for assessing different cognitive related to the subjects. The theoretical foundation of this test is that the intelligence is composed by a set of differentiated capacities instead of a single capacity [17, 18]. In Table 1 the variables provided by the BADyG test are showed.

Table 1. Evaluated variables in the BADyG test. Source: [17, 18]
Full size table

2.4 Data Mining

Data mining is the exploration and analysis of large quantities of data in order to discover meaningful patterns and rules, allowing for example, a corporation to improve its marketing, sales, and customer support operations through a better understanding of its customers [11].

2.5 CRISP-DM

CRISP-DM is a methodology widely used for analyzing large volumes of data and discover valuable information. The CRISP-DM methodology, according to [19], consists of six phases: understanding the business, understanding the data, preparing the data, modeling, evaluation and implementation.

2.6 Rapid Miner

Rapid Miner is an open-source software (OSS). Rapid Miner provides functionality data preprocessing and visualization, predictive analytics and statistical modeling, in addition of data mining. There are many OSS in the market but the user interface and workflow of Rapid Miner [20] makes it different from other tools.

3 Related Work

Several works uses data mining techniques to identify behavioral patterns that are useful for a particular context. In [14] propose collecting statistical data in order to identify behavioral patterns about travelers visiting the city of Pereira. Also describe their habits and classify those habits regarding trends. This work was carried out by using the CRISP-DM methodology, the KDD process, and the tool called RapidMiner. The work presented in [7] generates useful knowledge from historical data analysis of successful and failed projects on an IT outsourcing company by applying data mining techniques, in order to obtain patterns for allowing the organization to make decisions and guide software projects towards the achievement of success. This work was done by using the CRISP-DM methodology, the KDD process and the SPSS (statistical product and service solutions) tool. In [15] the authors use the data mining techniques to address the problem of student attrition at the Universidad distrital – Francisco José de Caldas in order to determine the causes that lead to desertion the students of the university. This work is intended to generate a decision tree model by implementing the J48 algorithm through the use of the WEKA tool for identifying such causes; also, they use the CRISP-DM methodology to development the work. Finally, the work presented in [5] analyze the academic information related to the academic results of the student and their interaction with the university. The authors identify factors that influence the student desertion of the computer science career at Gastón Dachary University in Argentina. This work apply data mining techniques and use classifications algorithms such as decision trees, Bayesians networks and rules. The work is developed under the KDD process and the WEKA tool.

These studies use the same technology in the same context of our proposal. However they have focused mainly on academic, social, demographic, and economic aspects. Unlike such proposals we also considerer variables provided by the BADyG test.

4 Our Proposal

This proposal consist of a method to analyze the academic desertion, starting from the data provided by the University of Quindío related to the students and making a particular focus on the faculty of engineering. For this proposal, we intend to work with the CRISP-DM methodology, the KDD process, especially in the data mining stage and the tool called RapidMiner. The goal of this proposal is to identify some behavioral patterns and relationships between a large numbers of variables of the students of the University of Quindío. Thus, this work can support decision-making and the creation of plans focused on addressing related problems with desertion in the engineering faculty of the University of Quindío. According to the CRISP-DM methodology, the proposal will handle five phases: understanding the domain, understanding the data, preparation of data, modeling, and analysis and evaluation, the Fig. 1 illustrates this proposal.

Fig. 1.
figure 1

Phases of the proposal.

Full size image

4.1 Understanding the Domain

The development of this phase known in CRIS-DM as “understanding the domain” allows us to obtain a higher level of understanding of the problem related to the case study. In this case the task consists of consulting researches and previous work related to the problem of desertion at the University of Quindío as well as at national and international level. Documentation related to knowledge discovery techniques and their application to desertion themes is also reviewed.

4.2 Understanding the Data

In this phase of CRISP-DM named as “understanding the data” the methodology propose criteria for selecting the data. Such data are obtained and explored in order to identify the elements for allowing the determination of the quality of the same. In this case the following criteria were defined: student´s data provided by the diverse departments at the University of Quindío responsible for collecting and storing the information of the students. Such departments were control and register office, planning department, and statistical department. The source data were provided by three sources: information concerning to the SPADIES taken from the planning and development office, BADyG test provided by the department of statistics and the personal and academic information of the student, taken from the register and control office of the University of Quindío. BADyG test was performed by the department of statistics at the University of Quindío [18]. A sample of the results obtained can be consulted in Figs. 2, 3, and 4.

Fig. 2.
figure 2

Source: [4]

Structure of the information provided by the SPADIES system.

Full size image
Fig. 3.
figure 3

Structure of the information provided by the statistical department of the University of Quindío related to BADyG tests. [18]

Full size image
Fig. 4.
figure 4

Structure of the information provided by the planning and development department of the University of Quindío related to the personal and academic information of the students.

Full size image

4.3 Preparation of the Data

This phase named “Preparation of the data” in CRISP-DM consists of organizing and debugging the data obtained. This phase is composed by four general tasks and four specific tasks. The task 1 allows us to determine the data will be included in the study and to exclude those with low quality. In this case the data related to the name and address of the students were excluded. Such data were classified as sensible information. The task 2 performs the cleaning process of the data starting from the selection criteria previously defined. The task 3 consists of the structuration of the data. In this step de tables for storing de data were structured. Finally, the task 4 allows us the integration of the data.

In previous phases we can see several sets of data which were provide by diverse and heterogeneous sources. So, the information presents great difficulties for its adequate integration. For example, the Table 2 shows the structure and the possible values related to the data provided by the planning and development department.

Table 2. Data structure provided by the planning and development department.
Full size table

As we can see such data as well as the data provided by the others dependencies, present several inconsistencies related to heterogeneous structure, missing values, duplicated records, and redundant information, among others.

In general all data provided by the different sources should be cleaned and integrated. All this process was carried out by using a set of macros in Microsoft Excel ®. Such macros were developed by the authors. Particularly, the information provided by such heterogeneous sources was imported in several excel tables. Then, we use several intermediate dynamic tables where the partial results were recorded. So, each of these tables were generated with Microsoft Excel ® by using macros, formulas and SQL sentences. Finally, the data resulting of the previous tasks were combined and integrated.

4.4 Modeling

In this phase the modeling techniques are selected and applied and their parameters are calibrated to optimal values. In this case we use the classification tree for relating the previously selected attributes. Such attributes were selected taking into account the level of incidence on the decision about the decision to desert or not to desert: BADyG test parameters, genre, age, marital status, victim of conflict, displaced, disabled, and stratum, etc. Such parameters can be consulted in Fig. 5a, b.

Fig. 5.
figure 5

(a) Consolidated final data. (b) Consolidated final data.

Full size image

4.5 Analysis and Evaluation

At this phase the model (or models) obtained are more thoroughly evaluated and the steps executed to construct the model are reviewed. Consequently, we present the results once the information was modeled, analyzed and evaluated. Such results are illustrated in Figs. 6, 7, 8, 9, 10 and 11.

Fig. 6.
figure 6

Civil engineering.

Full size image
Fig. 7.
figure 7

BADyG test at the civil engineering program

Full size image
Fig. 8.
figure 8

Electronic engineering

Full size image
Fig. 9.
figure 9

BADyG test at the electronic engineering program

Full size image
Fig. 10.
figure 10

Systems and computing engineering

Full size image
Fig. 11.
figure 11

BADyG test at the systems and computing engineering program

Full size image

Civil Engineering

The decision tree in Fig. 6 shows that the attributes and their relationships which encourage a student for deserting from civil engineering program are the following:

Age and number of people composing the home. According to analyzed data for this academic program, the students with age between 19 and 21 years old, besides their family group has 3 or more members.

Regarding BADyG test, the most significant attributes which encourage a student for deserting from this academic program, are the following:

MV (Visual memory): this aspect is one of the most relevant since it represents 69.2% percentage of those who deserted and had low score on this aspect.

RA (Speed): the speed to process information is other relevant aspect. Such aspect is a complement of MV, which makes it very difficult for students to learn on this academic program.

Reading comprehension also plays an important role in the determination to desert or not desert. As can be seen, the SV (Complete sentences) and DE (Discrimination of differences) aspects have a high percentage of students who deserted at this program.

Electronic engineering.

The decision tree in Fig. 8 shows that the attributes and their relationships which encourage a student for deserting from electronic engineering program are the following:

A significant weight is associated to the aspect determining if the student work or not. According to the data results provided by the decision tree, the students who study and work at the same time have a high probability for deserting. On the other hand, students who do not work depend on the attributes: number of family members, age, and stratum. Thus, students with stratum 1 and older than 20 trend to desert on the first 3 semesters.

Regarding BADyG test, the most significant attributes which encourage a student for deserting from this academic program, are the following:

SV (Complete sentences) and RA (Speed) demonstrate that students have problems with the reading comprehension, which represents a high probability for deserting.

Systems and Computing Engineering.

The decision tree in Fig. 10 shows that the attributes and their relationships which encourage a student for deserting from systems and computing engineering program are the following:

The number of family members, age, regime, contribute or not, and stratum. According to the Fig. 8, those students who desert economically contribute and have between 4 and 6 family members. Another interesting situation is related to the students who have between 1 and 3 family members and their regime is contributory (EPS, Empresa Prestadora de Salud). The same happens with the students who have ages between 19 and 22 years old, have between 1 and 4 family members and have stratum 1. Finally, students who are under 19 years desert when the stratum is 2 and have between 1 and 3 family members.

Regarding BADyG test, the most significant attributes which encourage a student for deserting from this academic program, are the following:

Students who have a score under 68 in EF (Effectiveness), a score under 19 in Mv (Visual memory), a score under 74 in IG (General intelligence), and have between 1 and 6 family members have a high probability for deserting. Likewise, students with high probability for deserting have a score under 57 in EF (Effectiveness), a score under 6 in Sn (Numerical problems), a score under 74 in IG (General intelligence), a score greater than 6 in Rv (Analog relationships), a score under 68 in Mv (Visual memory), and a number of family members between 1 and 6.

5 Conclusions

In this paper a method for analyzing the student attrition in the faculty of engineering at the University of Quindío was proposed. Thus, an explanation of the process was described by using the phases provided by the methodology CRISP-DM. This methodology favors the obtaining of results because it allows us the complete understanding of the problem in terms of the business and its meaning in terms of data mining.

The advantage of this work is that it takes into account a group of aspects related to intelligence tests. This was achieved by analyzing the data obtained from the BADyG (Battery of General and Differential Aptitudes) tests. The results of this work reflect the need of the university to establish action plans that improve the performance of students. Such as extracurricular advice and accompaniment to strengthen the flaws found. This strategy can be applied even in other universities.

Some inconveniences were found with the application of university policies to determine the status of a student deserter. Which were not very clear or according to reality. Consequently, as future work it is necessary to develop additional studies to improve the classification of deserter students.