International Journal of Scientific & Engineering Research Volume 2, Issue 12, December-2011 1

ISSN 2229-5518

Suitability of Training Programme Based on

Integrity Traits Identification

Mohamad Farhan Mohamad Mohsin, Faudziah Ahmad, Aniza Mohamed Din, Ku Ruhana Ku-Mahamud, Roshidi Din

Abstract- Training programs for teacher’s development in Malaysia are organized ever year. The selection of teachers to attend trainings is currently done randomly, by rotation and not based on their work performance. This poses a problem in selecting the right teacher to attend the right course. Up until now, there is no intelligent model to assist the school management to determine the integrity level of teacher and assign them to the right training program. Thus, this study investigates the integrity traits of teacher using association rule technique with an aim, which can assist the school management to organize a training related to teachers’ integrity performance and to avoid sending the wrong teacher for the training. A dataset of Trainees Integrity Dataset representing 1500 secondary school teachers in Langkawi Island, Malaysia in the year 2009 were pre-processed and mined using apriori. Mining knowledge was analyzed based on demographic and integrity trait of teacher. The finding indicates that adaptability and stability are the weakest integrity trait among teachers. Teachers from the age group of 26 - 30 years are found to have lower integrity performance. However, other demographic factor such as gender, race, and grade position of teachers were not able to reflect their low integrity level in this study. Findings from this study can be used as guideline for school management to propose suitable training programs for teacher to improve their integrity mainly on adaptability and stability traits.

Index Terms— association mining, apriori, big five model, data mining, education training, teacher’s integrity

——————————  ——————————

1 INTRODUCTION

NTEGRITY is a vital trait that everyone must own to per- fectly complete tasks. This includes teachers who play a very important role to nation. They are responsible to de- velop valuable citizens and educate students with a high- quality education. To teach students mainly at the young age is not an easy job. Therefore, teachers must have high integri- ty, and they must establish integrity within the school to ac- complish the goal. Without the presence of integrity, teachers
will definitely fail to do it [1].
To maintain the high level of integrity, the government
through the Ministry of Education (MoE) has arranged many
programs for teacher’s development through training, work-
shop, and seminar. Currently, the training program is ran-
domly organized to fit the yearly calendar without consider- ing the right teacher for the right training. The selection of
trainees to attend training is done randomly, sometimes by rotation and most of the selection is not based on their work performance. Tailored to [20], there are many training pro- grams available but there is no systematic way to ensure all teachers acquire and continue to develop the knowledge they needed.
Even though teachers are evaluated through the yearly per- formance appraisal, the inflexibility of the system to identify less motivated teacher will burden the school management. Up until now, there is no intelligent model in determining the integrity level of teachers and assign them to the right training program.

————————————————

 All authors are researchers at the School of Computing, UUM College of Arts & Sciences, Universiti Utara Malaysia, 06010 UUM Sintok, Kedah Malaysia.

 Inquiries can be emailed to {farhan; fudz; anizamd; roshidi; ruha-

na}@uum.edu.my.

As a result, this will create difficulties for the school man- agement to identify problematic teachers and sending the wrong ones to attend the training. If the requirement of train- ing program is mismatched with the trainee, there is a high possibility that the objective of the training cannot be achieved. Teachers who possess such negative attitudes would not be suitable for an integrity program. If they were selected randomly, there would be a mismatch of supply and demand of needs which would cause failure in the integrity training programs to achieve their goals. According to several studies, selecting a wrong person can increase a cost between 30% and
200% of a person’s annual salary. If a role worth $70,000 a year is filled with the wrong candidate, it could end up costing be- tween $21,000 and $140,000 – a loss most businesses cannot afford to make [2]. Even though there is a keen action form the government to raise the teacher’s integrity, it appears to be a number of reports that some teachers are unable to perform their tasks efficiently when they are having a problem with students and also some of them might involve in crimes [9,
11].
Integrity is viewed as the quality of having an intuitive sense
of honesty and truthfulness of a human in completing task. It is
normally a product of culture with the relation of good traits
and discipline of human beings in all situations [12]. As the re-
sult of high integrity, human beings will be more accountable
and responsible towards his job and struggle to reach his objec- tive. It is the key of success to everyone, and many organiza-
tions have benefited an enormous performance increase when their employees own good integrity behavior. As stated in [1] nothing can work without the presence of integrity. In relation to integrity, it is stressed out that the integrity among teachers is critically important [17] and sometimes the solution towards integrity issues is difficult to be answered. Finding a teacher with high integrity characteristic is not easy since there is no agreement on what makes a teacher have a high quality attitude

IJSER © 2011

http://www.ijser.org

International Journal of Scientific & Engineering Research Volume 2, Issue 12, December-2011 2

ISSN 2229-5518

in school. Such criteria like the experiences that the teacher has, the education level, and the performances achieved by their students can be used to measure his integrity. Besides that, the integrity test such as the Big Five Model [3] and personality test [16] also can be a guideline to measure the level of teacher’s integrity.
High quality teaching is essential to improving student out-
comes and reducing gaps in student achievement. In school,
teachers must uphold the highest integrity and be a good role
models to their students at all times. Their responsibility to- wards students are not limited during school hours but contin-
ued after the class end. Teachers are the nearest people who are examples for students other than their parents. As an idol for students, teachers need to be equipped with a good package of moral quality. For the reason to attract students to behave nice- ly, teachers must have a quality mind and good attitude to im- press students. They need to be familiar with the students as well as the school community to be able to successfully educate students. Good quality teachers must be able to target and plan how to manage their students to become good quality students

and D be a set of transactions, where each transaction T is a set of items such as that T  I. An AR is an implication of form X

→ Y, where X  I, Y  I, and X ∩ Y =Ø. The rule X → Y has support s in the transaction D if s% of transactions id D con-

tain X Y. The rule X → Y holds in the transaction with confi-

dence c if c% of transaction in D that contain X also contain Y.
The algorithm depicts in Fig. 1 is the process of generating
frequent item set in apriori [18, 19]. AR mining’s processes
begin with searching for frequent item set with user-specified
minimum support and later rules are contrasted by binding the frequent item with its values and class. Strong rules are

defined as rules that have confidence more than the minimum confidence threshold. Recently, AR has been used as in vari- ous areas mainly when the problem need to be solved requires the indetification of important relationship among variables such as the work of [21, 22].

Input: database of transactions (D); the mini- mum support count threshold (min sup)

Output: frequent itemsets in D (L)

[10, 13]. Besides that, they also need to be very understanding
and show their respect to the students’ behavior [14]. High

(1) L1

= find_frequent_1-itemsets(D);

quality teaching is essential to improving student outcomes and reducing gaps in student achievement. Stated by [17], an effec- tive teacher is capable of inspiring greater learning gains in their students compared to a teacher with low teaching quality. Those traits have a strong relationship with teacher’s integrity. With the high capacity of integrity, teacher will be motivated to educate student and help the country to develop human capital. It will not only benefit the students and country, but also the teacher itself in providing them with meaningful appraisal, en- courage professional learning and growth [20].
Since the MoE faces problems in discovering problematic teachers related to integrity, a model to identify the weakest integrity trait of teacher is proposed in this paper. The aim of this model is to assist the school management to discover inte- grity traits that are mostly lacking among teachers and proposes a suitable training program for them. To achieve that, data min- ing algorithm called apriori is used to mine Teacher Integrity Dataset (TID). The dataset consists of 1500 information about demographic and integrity score of secondary school teachers in Langkawi Island, Malaysia in the year 2009.
This paper is organized as follows. Section II outlines the ba- sic notion of AR. The model development of the study is dis- cussed in section III. The experiment and result will be pre- sented in section IV and the final section concludes this work.

2 ASSOCIATION RULE

In this section, the basic of association rule (AR) mining is dis- cussed. Association rule mining or AR was first inspired by Agrawal [18] in apriori algorithm. In AR, it relies on the con- cept of ―frequently occurred together‖ which refers to process of the identifying of frequent items that occur in a database of transaction. Each item (ij) in a transaction is an important fea- ture that contributed to the computation of item set and gen- eration of rules. Basically, let I = {i1,i2,…, im} be a set of items

(2) for (k = 2;Lk-1 ≠ Ø;k++){

(3) Ck = apriori_gen (Lk-1);

(4) for each transaction t  D { // scan D

for counts

(5) Ct = subset (Ck, t); // get the subsets of t that are candidates

(6) for each candidate cCt

(7) c.count++;

(8) }

(9) Lk ={cCk|c.count ≥ min_sup}

(10) }

(11) return L = kLk;

Function apriori gen(Lk-1:frequent (k-1)- itemsets)

(1) for each itemset l1  Lk-1

(2) for each itemset l2  Lk-1

(3) if (l1[1]=l2[1])^( l1[2]=

l2[2])^…^( l1[k-2]=l2[k-2])^( l1[k-1] <

l2[k-1]) then {

(4) c = l l ; // join step:

1 2

generate candidates

(5) If

has_infrequent_subset(c, Lk-1) then

(6) delete c; // prune

step: remove infrequent candidate

(7) else add c to Ck; (8) }

(9) return Ck;

Procedure has infrequent subset(c: candidate k-itemset;

Lk-1: frequent (k-1)-itemsets); // use prior knowledge

(1) for each (k-1)-subset s of c

(2) if sLk-1 then

(3) return TRUE;

Fig. 1. Generating frequent item set in apriori [19]

IJSER © 2011

http://www.ijser.org

International Journal of Scientific & Engineering Research Volume 2, Issue 12, December-2011 3

ISSN 2229-5518

3 MODEL DEVELOPMENT

This chapter describes how this study was conducted. The study was divided into three phases which started with data gathering, preparing data for mining and ended with pattern extraction. The details of the phases will be described in the following subsections. Fig. 2 illustrates the data gathering, data preparation for mining and integrity traits extraction of this study.

Fig. 2. Phases of the study

3.1 Data Gathering

The first step of this study is to obtain experiment data from school. In this study, a set of filled questionnaires of secondary school teachers in Langkawi Island, Kedah, Malaysia was se- lected as a case study. The questionnaire was designed by Langkawi Education District Office and it was distributed to teachers while attending training or seminar in the year 2009. There were 1500 respondents and 26 attributes including a decision class. The data set hold two decision class ―Integrity‖ if the accumulative score of integrity test is more than 80% and
―Not Integrity‖ if the score is lesser than 80%. Out of 1500,
36.17% respondents were male and 63.87% were female teach- ers. Table 1 represents the distribution of respondents based on integrity level. The filled questionnaires were put into excel and named as Trainees Integrity Dataset (TID).

TABLE 1

DISTRIBUTION OF RESPONDENTS BASED ON INTEGRITY LAVEL

(BFM) [3] which is needed for stability, extraversion, originali- ty, accommodation, and consolidation. Theoretically, the BFM identifies traits and structure of human personality and it acts as a guideline to measure integrity of a human [3]. The model has been widely used in various areas such as motivating a human being. BFM has been found to produce consistent re- sults over the past years and is the basis of characterizing per- sonality [4, 5, 6]. Measuring the integrity of a training candi- date can help authority such as employer and training provid- er to identify the levels of integrity of a person. Table 2 lists the description of integrity traits as stated in [3].

TABLE 2

INTEGRITY TRAITS BASED ON FIVE FACTOR MODEL [3]

The TID represents two informations that are demographic background and five key traits on integrity of teachers. The integrity elements are inherited from the Big Five Model
Each of the five integrity element (stability, extraversion, originality, accommodation, and consolidation) in TID is represented by four questions. The questions are labeled with A1-A4 for Extraversion, B1-B4 for Consolidation, C1-C4 for Adaptability, D1-D4 for Stability and E1-E4 for Originality. Each integrity traits holds 20 marks where 5 marks was allo-

IJSER © 2011

http://www.ijser.org

International Journal of Scientific & Engineering Research Volume 2, Issue 12, December-2011 4

ISSN 2229-5518

cated for each question. Table 3 displays the list of questions of each integrity trait.

TABLE 3

LIST OF QUESTIONS BASED ON FIVE FACTOR MODEL [3]

For demographic background information, the attributes are gender, age, race, type of school, department, and position grade. Preliminary observation on the raw dataset shows that several attributes were not related to the study; certain values were miss- ing and duplicated. The next step is to pre-process the TID before it can be mined with AR algorithm.

3.2 Data Preparation for Mining

The second step of the study is preparing the data for min- ing. This step involves data preprocessing activities whereby the problem in TID were identified and resolved. During pre- processing task, all dataset were pre-processed where all un- known numeric attributes were replaced with mean value while max value for character attributes. Since the AR algo- rithm in WEKA only accepts nominal type data, the dataset were then discretized using Boolean reasoning technique [15]. To increase the mining speed and accuracy, only important attribute was given priority for mining and those attributes which were not related to the study were ignored. During the selection process, 20 attributes which represent integrity test score were reclassified into five new groups based on the BFM model [3]. The new groups were stability, extraversion, origi- nality, accommodation, and consolidation. Each attributes represents an integrity question with the maximum score was
5. In this process, the sum score of each attribute based on the type of integrity was aggregated and the total score was 20. For example, the attributes A1, A2, A3, and A4 are the set of integrity question for Extraversion. Let say the score is given as A1=1, A2=4, A3=4, A4=1 then the total accumulative score for Extraversion is 10/20. Based on the accumulative score, the value is reclassified into two classes, either ―15-20‖ or ―0-14‖. The ―15-20‖ group was considered as high score and ―0-14‖ as

IJSER © 2011

http://www.ijser.org

International Journal of Scientific & Engineering Research Volume 2, Issue 12, December-2011 5

ISSN 2229-5518

lower. Fig. 3 summarizes the process of attribute selection and reduction.
In the final step of data preprocessing, TID was separated into ―Integrity‖ and ―Not Integrity‖ group and only ―Not In- tegrity‖ group was presented to AR. The reason is to investi- gate the weakest trait of integrity that reflects teachers’ per- formance. Besides, it can concentrate into one group and this step can eliminate longer mining time as well as reducing the number of knowledge generated.

Fig. 3. Model development of the study

After the pre-processed, there were 454 remained records out of 1500 and 10 attributes. The attributes were gender, age, po- sition, grade, department, stability, extraversion, originality, accommodation, and consolidation. Table 4 shows the statistic of ―Not Integrity‖ group based on gender.

TABLE 4

THE STATISTIC OF NOT INTEGRITY GROUP BASED ON GENDER READY FOR MINING

analyzed to finalize the interesting matching among integrity trait. The next section will discuss the finding of this study.

4 FINDING

This section reports the finding of this work. After mining with apriori, there were 1000 rules generated and the algo- rithm also had generated interesting integrity trait with the maximum combination length is 7. Fig. 4 shows the number of interesting trait combination (f_Char) towards frequent item set length. Ln is representing the length and n is the number of attribute combination. For example, if L4, there are four attributes combination. From Fig. 4, the highest f_Char was generated in L4 and it was followed by L3. The number of f_Char is getting lesser when length increases.

Fig. 4. The number of f_Char according to Ln

The results were further analyzed. We continued our analy- sis on the f_Char by focusing at L4 until L7. The L1-L3 groups were ignored since the trait is too compact and short. This ex- ploration was divided into two parts. Firstly, is to investigate the demographic trait and secondly, is to investigate the inte- grity criteria trait. Table 5 and 6 depict the sample of f_Char when the number of integrity trait combination is 4. S% in both tables is the support value.

TABLE 5

THE SAMPLE OF F_CHAR WHEN THE MAKSIMUM TRAIT MATCHING LENGTH IS L=4

3.3 Integrity Traits Extraction

In pattern extraction, AR algorithm called apriori in WEKA data analysis tool was chosen as pattern extraction tool [7]. The clean TID dataset was presented to apriori algorithm and during mining, the length of frequent item set, support, and confidence value of each item set were recorded. In this study, the minimum support value was set to 10% while confidence value was limited to upper than 95%. Besides that, the maxi- mum number of rule was limited up to 1000. The knowledge extracted from apriori was represented in term of frequent teacher’s integrity trait matching. Then, the rules were further

IJSER © 2011

http://www.ijser.org

International Journal of Scientific & Engineering Research Volume 2, Issue 12, December-2011 6

ISSN 2229-5518

TABLE 6

THE SAMPLE OF F_CHAR WHEN THE MAKSIMUM TRAIT MATCHING LENGTH IS L=5

Under demographic analysis, we found that the type of gender is not an important trait to determine the integrity lev- el of teachers. Eventhough the gender attribute had appeared
55.08% in L4, both sexes have similar probability to behave with low integrity. Moreover, the study also found that the
age group within ―26-30‖ was the highest contributor in all f_Char and interestingly, most of them were female. The figure in Table 7 indicates the number of f-Char according to age group. Then, the salary grade and position were found not as a strong trait to determine integrity.

TABLE 7

THE NUMER OF F_CHAR ACCORDING TO AGE

Age f_Char length

L1 L2 L3 L4 L5 L6 L7

TABLE 8

THE RESULT OF SCANNING PROCESS TOWARDS INTEGRITY SCORE

“0-15” IN F_CHAR

f_Char E C A S O

L4 73 54 179 165 61

L5 26 18 145 106 24

L6 2 2 43 40 0

L7 0 0 4 0 0

Total 101 74 371 311 85

E- Extraversion C- Consolidation A- Adaptability S-Stability O-Originality

Further observation was performed on the combination of more than one trait. The observation reveals that adaptability and stability traits were frequently occurred together in all f_Char. For example, in L4, there were 47 combinations of adaptability and stability and 37 were found in L5. While in L6,
10 combinations were recorded. This result indicates teacher who lack adaptability will also have problem related to stabili- ty. In addition, the combination of ―stability and originality‖ and ―adaptability and originality‖ was also high. As can be seen in L4, there are 23 combinations of ―adaptability and ori- ginality‖. The explanation of this part is summarized in Table
9. It indicates the frequency of the weakest trait for two traits combination in L4, L5 and L6.

TABLE 9

THE FREQUENCY OF THE WEAKEST TRAIT FOR TWO TRAITS COMBINA- TION IN L4, L5, AND L6

Under integrity traits, we concentrated our analysis at the weakest score of integrity. The f_Char were scanned and all five integrity traits with the weakest score ―0-15‖ were marked as shown in Tables 5 and 6 in bold. The result of the scanning process is depicted in Table 8.
From Table 8, it can be seen that adaptability is the weakest integrity trait among teachers. In all n, adaptability has the highest score which indicates most teacher are lacking at the adaptability aspect. This is closely followed by stability which also indicates less integrity among teachers. The score of other trait – extraversion, consolidation, and originality is consi- dered low compared to adaptability and stability.

E- Extraversion C- Consolidation A- Adaptability S-Stability O-Originality

For longer combination, association rule generated several interesting traits matching. However, the number of combina- tion is small. Information summarized in Table 5 is the inter- esting trait generated in L4 with the number of combination is three and four. From Table 5, adaptability and stability exist in all combinations. Besides that, extraversion is sent to be fre-

IJSER © 2011

http://www.ijser.org

International Journal of Scientific & Engineering Research Volume 2, Issue 12, December-2011 7

ISSN 2229-5518

quently paired with consolidation traits. The result is tailored with the f_Char in L5 as displayed in Table 6.
However, starting from L6, there are no interesting traits matching found with the condition minimum trait combina-
tion more than 2. As displayed in Table 10, the f_Char in L5 only generates interesting traits matching with maxsimum length L=2. Similar to L7, no interesting combination found.

TABLE 10

THE INTERESTING INTEGRITY TRAIT MATCHING GENERATED IN L6

WITH THE MAXSIMUM MATCHING COMBINATION IS TWO

f_Char (L6) Length S%

In the L7 which represents the longest attribute combina- tion, apriori has produced 5 rules as illustrated in Table 11. Hence, only adaptability trait appears to be the weakest inte- grity trait among teacher. Female teachers from A grade group seem to have the least integrity based on this frequent item set.

TABLE 11

THE LIST OF RULES GENERATED IN L7

In general, the results indicate that adaptability and stabili- ty are the weakest traits among teachers due to the frequency of both traits exit in f_char. The other three characteristics also exist in f_Char but due to their low occurrence, we decided to ignore them as weak integrity traits. According to literature; adaptability refers to the tendency of teacher to be compassio- nate and cooperative rather than suspicious and antagonistic towards others, while stability refers to the ability of teacher responding to stress, high work load and dealing with nega- tive emotions. Since adaptability and stability are the weakest traits among teacher, the school management may organize
suitable trainings to improve both traits.

4 CONCLUSION

Association rule technique was used to discover the weakest integrity trait of teachers. Results showed that adaptability and stability are the weakest traits among the teachers. Find- ings from this study can be used to assist the school manage- ment in identifying integrity traits that are mostly lacking among teachers and proposes a suitable training program for them. In addition, the school management will be able to make a shortlist of candidate for training. School management can use this guideline to propose suitable training program for teachers to improve their integrity mainly on adaptability and stability.

ACKNOWLEDGEMENT

This research work was fully supported by LEADS Grants of Universiti Utara Malaysia (UUM) under the Ministry of- Higher Education of Malaysia.

REFERENCES

[1] K. Christensen, The Magazine of the Rotman School of Management, Fall 2009, pp. 16-20.

[2] M. Dowding. Northern Rivers Business Magazine: Autumn 2011.

[3] P.J. Howard, and J. M. Howard, ―The Owner’s Manual for Personali- ty at Work‖. Austin, Texas: Bard Press, 2001.

[4] L.M. Saulsman and A.C., Page. "The five-factor model and personali- ty disorder empirical literature: A meta-analytic review". Clinical Psychology Review 23 (8): 1055–1085, 2004.

[5] M.R. Barrick, and M.K. Mount. "The Big Five Personality Dimensions and Job Performance: A Meta-Analysis". Personnel Psychology 44: 1–

26, 1991.

[6] M.K. Mount and M. R. Barrick, "Five reasons why the "Big Five" ar- ticle has been frequently cited". Personnel Psychology 51: 849–857,

1998.

[7] I.H., Wittenand and E., Frank. ―Data Mining: Practical machine learn- ing tools and techniques‖. Morgan Kaufmann: San Francisco, 2005.

[9] Azam Ahmad. School teacher suspected rape 16 years old student.

The Star, 2010. (online). www.thestar.com (5th July 2010)

[10] Baker. J. Nobody's Perfect, But You Have To Be: The Power of Per- sonal Integrity in Effective Preaching, New York: Mc Graw-Hill,

1990.

[11] Kamarul Hassan. Guru Sekolah disyaki Merogol Pelajar Tingkatan 2.

Utusan Malaysia, 2010. (Online) www.utusan.com.my (5th July 2010) [12] Popper, K. R. Objective Knowledge in higher educational systems

processes. Proceeding of 5th international conference, University of

Sydney, Australia. pp. 60-69, 1983.

[13] Nillsen, R. The concept of Integrity in Teaching and Learning, Sympo- sium on Promoting Academic Integrity, Newcastle NSW, 2004.

[14] Thornton, M. Working with Integrity, Code Of Business Product, BhpBiliton, 2004. (0nline) www.bhpbilliton.com (1st August 2010)

[15] Nguyen, H.S., ―Descretization problem for rough set methods‖. In Proc of First Int. Conf. on Rough Set and Current Trend in Compu- ting, 1998, pp. 545-552.

IJSER © 2011

http://www.ijser.org

International Journal of Scientific & Engineering Research Volume 2, Issue 12, December-2011 8

ISSN 2229-5518

[16] Daniel, N. "A test of character". The Guardian (London), 2009.

[17] Starling, J.W. The effects of teacher stability on third grade student achievement as measured by the North Carolina End-of-Grade tests in reading and mathematics. Gardner-Webb University. Phd Thesis,

2009.

[18] Agrawal, R. and Srikant, R.. ―Fast algorithm for mining association

rules.―, Proc. Int. Conf. Very Large Databases, pp. 487-449, 1994.

[19] Han, J. & Kamber, M.. Data mining concept and technique. San Fran- cisco: Morgan Kaufmann Publishers, 2006.

[20] Joan, M.R. Carrer Long Teacher Development: Policies that Make Sense. WestEd. 2006. (0nline) http://www.wested.org/online_pubs/teacher_dev/Teac
herDev.pdf (15st August 2010)

[21] Mohamad Mohsin, M.F., Md Norwawi, N, Hibadullah, C.F. & Abdul Wahab, M.H. An Investigation into Influence Factor of Student Pro- gramming Grade Using Association Rule Mining. International Jour- nal of Advances in Information Sciences and Service Sciences. Vol 2,

2010.

[22] Ma, Y., Liu, B., Wong, C.K., Yu, P.S., and Lee, S.M. ―Targeting the Right Students Using Data Mining‖. In Proceedings of the Know- ledge and Data Discovery (KDD2000), Boston, USA. Pp457-464, 2000.

IJSER © 2011

http://www.ijser.org