CHAPTER 1
INTRODUCTION

1.1 PRACTICUM INTRODUCTION

According to Macmillan English Dictionary for Advanced Learners Second Edition (2011), practicum or practical means involving or relating to real situations and events. In the other words, it is the activity which not only about theory but facing the real world works. Students will be trained to apply the theory during this period. In Universiti Utara Malaysia (UUM), practicum is obligated for students in some courses as a fulfillment in the course structure. Practicum program is also designed to provide and expose student to real world problem situations before the end of graduation. From this program, students are given the opportunity to practice and apply their understanding and theoretical knowledge that have been learned in classroom with the help and guidance from the university and the attached organization. It is a platform of training and preparing the student before entering the job market. Practicum program is compulsory for all students pursuing Bachelor of Business Mathematics with Honors in UUM. This program can also be seen as a method to integrate the theoretical knowledge with the hands-on training. In this way, students would have an opportunity to apply their understanding of theories into the real working environment within industries. Practicum program provides opportunities for students to improve two vital skills which are soft kills and hard skills. It aims to help student to apply these two skills in interpersonal relationship with employees from various ages and skill levels. At the end of this program, students are expected to have a strong individual personality with a variety of technical and soft skills and able to present themselves as a dynamic role model in family, community and country.

1.1.1 PRACTICUM OBJECTIVES

After practical training, students will be able to: i. Applying theories and concepts learned in the real world of work environments. i. Identify the structural diversity of the organization. ii. Enhance and strengthen communication skills iii. Establishing good relations among workers in the organization. iv. Ability to work in groups. v. Perform all duties within the organization that lead to the achievement of objectives. vi. Assessing the responsibility of management and certain units in the organization’s objectives. vii. Mastering the appreciation of organizational culture.

1.2 ORGANIZATION BACKGROUND

Johor Government Secretary Office, Pejabat Setiausaha Kerajaan Johor, (JGSO) consists of fifteen departments specializing and being responsible in different fields. These departments’ main responsibilities besides to empower public sector services towards the residents of State of Johor is to perform their designated area of tasks. The state government is focused on the effort to meet the expectations and needs of the people in acquiring the services provided by the Department and the State Agency under the Office of the State Secretary of Johor. The fifteen departments are: 1. Office of Johor Chief Minister (Pejabat Menteri Besar Johor) 2. Johor Government Secretary Office Department of Management
(Pejabat Setiausaha Kerajaan Johor Bahagian Pengurusan) 3. Johor Government Secretary Office Department of Human Resource Management (Pejabat Setiausaha Kerajaan Johor Bahagian Pengurusan Sumber Manusia) 4. Johor Government Secretary Office Department of Housing
(Pejabat Setiausaha Kerajaan Johor Bahagian Perumahan) 5. Johor Government Secretary Office Department of Local Government
(Pejabat Setiausaha Kerajaan Johor Bahagian Kerajaan Tempatan) 6. Johor Government Secretary Office Department of Internal Audit
(Pejabat Setiausaha Kerajaan Johor Bahagian Audit Dalam) 7. State Meeting Council, Branch of the State Assembly Writer
(Majlis Mesyuarat Kerajaan, Cawangan Jurutulis Dewan Negeri) 8. Johor Water Regulatory Body (Badan Kawal Selia Air Johor) 9. Johor State Council of Sport (Majlis Sukan Negeri Johor) 10. Johor State Department of Landscape (Jabatan Landskap Negeri Johor) 11. Johor State Department of Tourism (Jabatan Pelancongan Negeri Johor) 12. Johor Heritage Foundation (Yayasan Warisan Johor) 13. Johor State Economic Planning Unit (Unit Perancang Ekonomi Negeri Johor) 14. Johor State Unit of Science, Technology and ICT
(Unit Sains, Teknologi dan ICT Negeri Johor) 15. Johor National Park Corporation (Perbadanan Taman Negara Johor)
Each department is responsible in managing and development of its department, and is regarded one whole company or firm. The firm that I interned at was JGSO Department of Human Resource Management (DHRM). JGSO DHRM, headed by Head of Assistant Secretary (Ketua Penolong Setiausaha, which hereon will be addressed as HAS), is divided into six different units which are headed by Assistant Secretaries (Penolong Setiausaha, which hereon will be addressed as AS). The six different units are: 1. Organisational Development and Information Management Unit (ODIMU), * To implement planning and posting development that covers researches on organisational restructuring. * To be the source of reference for personnel information management. 2. Promotion, Performance, Disciplinary and Counseling Unit (PPDCU), * To carry out affairs regarding promotion, performance, disciplinary and counseling for all state civil servant. 3. Service, Placement and Appointment Unit (SPAU), * To carry out affairs regarding service, placement, appointment and pension for all state civil servants. 4. Human Resource Development Unit (HRDU), * To organize courses/training and examination to produce excellent civil servants. 5. Administration Unit (AU), and * To administer the company regarding financial, inventory, company property disposal. 6. Integrity and Quality Unit (IQU). * To carry out affairs regarding integrity and quality to improve human capital management in ensuring the constant betterment of civil servants.

1.2.1 CORE VALUES OF ORGANIZATION

In line with the changing expectations for the state to be more innovative and creative, JGSO DHRM has aligned itself to meet the challenging demands as outlined in the New Economic Model (NEM) and to accelerate the state’s transformation and aspiration to achieve a high income economy through enhancing productivity and innovation. JGSO DHRM also functions as front-liner for Iskandar Regional Development Authority (IRDA) in preparing the civil servants as well as the state itself for, among others, the facilities improvements and opportunities prospect enhancement, developed by IRDA. Besides that, JGSO DHRM, corresponding with the implementation of QE 5S, aims to progress the quality of documentation and management of inventory. JGSO DHRM is also responsible in developing state public service organisation through organisations’ restructuring aligned with current responsibility scope to function optimally. JGSO DHRM also ensures the best approach of human capital management is used to improve civil servants’ efficiency and effectiveness through systems development and latest practices based on principles, ethics as well as norms standardization that have been deemed as. Finally but not least, JGSO DHRM ensures the employee-employer relationships are harmonized through information dispersion are wholly done to recuperate mutual understanding between them.

1.2.2 ORGANIZATION’S VISION

To be the leading public sector personnel agency that has high potential and is capable of giving excellent service through policy implementation while complying with State of Johor Public Service Code of Ethics.

1.2.3 ORGANIZATION’S MISSION

To ensure the process of planning, supervising and implementation of all state level programmes are aligned with National Development Policy objectives as well as to prepare macro planning, economic development, citizen, jobs and state income for short, medium and long term projection.

1.2.4 ORGANIZATION’S OBJECTIVES * Providing value-added information on productivity, quality, competitiveness and best practices through research activities and databases; * Developing human capital and organisational excellence for building a knowledge-based society through training, systems development and best practices; * Nurturing innovative and creative culture through QE 5S.
1.2.5 ORGANIZATION’S LOGO

Figure 1: JGSO DHRM Logo

CHAPTER 2
PROJECT DEFINITION

2.1 PROBLEM STATEMENT

Each month, Johor Government Secretary Office, Department of Human Resource Management (JGSO DHRM) spends a huge amount of money for courses, induction, training, development programmes, team building, seminars and others. The expenditure is solely from Human Resources Development Unit (HRDU) alone, not including five other units in JGSO DHRM. This expenditure is spent aiming to gain high return in qualitative aspects. These qualitative aspects among others are job performance, creativity, innovation, cooperation, policy understanding and others. As this expenditure is only from one unit, excluding other units’ expenditure, financial management in this particular public sector is a vital part in managing JGSO DHRM. The objective of training and induction is to enable civil servants to acquire the knowledge, skills, abilities and attitudes necessary to enable them to improve their performance. Staff training and development should focus on the departments’ objectives and goals and staffs’ competencies in achieving them. The funds in ensuring human resources development are efficiently progressing are allocated yearly. These allocations are given after a brief, nonspecific analysis of expenditure is proposed to apply for a yearly budget. The real problem occurs when funds are proven to be insufficient towards the end of the year or there is an excess amount of funds as the year finishes. This shows that there is no formulation or method of finding an optimum value of expenditure per month to fully utilize the allocations given by the state government.

The increment and decrement of expenditures are due to the rate of staff retention in government offices and the rate of staff appointment in government offices. Since there is no formulation or a proper forecast technique used in determining the amount should be used in expenditure of training and induction courses, a new strategy of forecasting should be applied in order to optimize the monthly usage of the yearly funds allocated.
To attain the objectives aforementioned, JGSO DHRM has to produce an accurate and detailed report in order to provide comprehensive analysis of expenditure for civil staffs’ development in the state of Johor and leads to the process of enhancing productivity at every level.
In preparing the forecasting framework, problem statements arises that influence the quantitative aspects are: i. What is the best technique to forecast the expenditures spent on Johor Government employees training course and development programmes? ii. What are the forecasted values of Johor Government employees training course and development programmes expenditure for the month October, November and December of the year 2012?

2.2 PROJECT OBJECTIVE

Objective of the project is to identify the best forecasting model to predict the expenditures spent on Johor Government employees training courses and development programmes for October, November and December of 2012. To compare and measure accuracy of the different forecasting methods, several techniques are going to be used such as Mean Squared Error (MSE), the Root Mean Squared Error (RMSE), Geometric Root Mean Squared Error (GRMSE) and the Mean Absolute Percentage Error (MAPE).

2.3 PROJECT SCOPE

The scope of the study includes the expenditures spent on Johor Government employees training courses and development programmes. Analysis made in monthly basis by firstly predicting the fitted value of productivity growth for the latest three months in the past to determine the forecasting ability. Then the fitted value of the productivity growth will be compared with the actual data to determine the accuracy. 57-months period starting from January 2008 to September 2012 will be used to forecast the the expenditures spent on Johor Government employees training courses and development programmes for October, November and December of 2012. The predicted model that produces most reasonable and accurate forecast will be used to make a prediction for expenditure analysis in the future.

2.4 SIGNIFICANCE OF THE PROJECT

Through this study, I will be able to find the best forecasting technique to identify and calculate expenditure spent in training and induction for JGSO DHRM. This study will assist HRDU, to predict expenditures that will be spent in the future by using simple technique such as analysis of time series using a variable which is the expenditure value. Besides that, this study will help create a budget planner system which helps to build tighter financial controls. This in turn will prevent expenditure overrun, which is the event when the costs for a project go over the amount that was originally estimated to be needed. Finally, this forecast can help in maintaining liquidity of DHRM. While the operating statement is a straightforward snapshot of operation status of DHRM in regards to finance at any given point in time, liquidity is also important. Accurately monitoring the working capital on a monthly basis is ultimately more significant than forecasting longer-term.

CHAPTER 3
METHODOLOGY

3.1 SOURCE OF DATA

The data which will be employed for this project is obtained from financial reports of Administration Unit, JGSO DHRM. This data is a secondary data which shows the expenditures spent on Johor government employees training courses and development programme. The reason for me to use a group of secondary data is because; the primary data of expenditure is with JGSO Department of Internal Audit for auditing purposes, and the only data available for analysis or research are the only secondary data.

3.2 DATA COLLECTION

The data collected for this project is monthly data which contains 57 data points starting from January 2008 to September 2012. Table in the next page shows the data of Expenditures spent on Johor Government Employees for training courses and development programmes for the duration beginning from January 2008 to September 2012.

Table 1: Expenditures spent on Johor Government Employees for training courses and development programmes

Year | Month | Expenditure (RM) | 2008 | 1 | 183393.80 | | 2 | 194382.40 | | 3 | 156399.20 | | 4 | 185293.40 | | 5 | 170483.20 | | 6 | 140171.20 | | 7 | 183174.50 | | 8 | 263102.30 | | 9 | 261289.30 | | 10 | 193298.20 | | 11 | 100397.30 | | 12 | 91391.70 | 2009 | 1 | 197362.40 | | 2 | 256204.20 | | 3 | 249204.30 | | 4 | 492374.30 | | 5 | 294073.20 | | 6 | 449206.10 | | 7 | 396561.30 | | 8 | 418211.40 | | 9 | 394263.30 | | 10 | 153294.30 | | 11 | 197453.40 | | 12 | 491472.30 | 2010 | 1 | 362541.50 | | 2 | 397385.30 | | 3 | 239434.20 | | 4 | 584263.20 | | 5 | 653226.10 | | 6 | 456364.20 | | 7 | 453765.40 | | 8 | 547955.30 | | 9 | 166466.90 | | 10 | 175464.20 | | 11 | 342242.70 | | 12 | 229754.40 | Year | Month | Expenditure (RM) | 2011 | 1 | 543437.30 | | 2 | 398339.30 | | 3 | 202422.20 | | 4 | 589342.20 | | 5 | 578315.80 | | 6 | 548427.30 | | 7 | 591950.80 | | 8 | 378410.10 | | 9 | 348745.20 | | 10 | 454392.20 | | 11 | 379266.80 | | 12 | 601285.10 | 2012 | 1 | 538204.40 | | 2 | 592563.40 | | 3 | 417982.920 | | 4 | 603641.10 | | 5 | 680200.60 | | 6 | 698384.10 | | 7 | 432438.20 | | 8 | 451481.30 | | 9 | 606124.30 |

3.3 ANALYSIS OF DATA: FORECASTING TECHNIQUE

Forecasting is the whole process of developing the necessary methods to generate the future values which subsequently can be used as the inputs for the goals and objectives of the firm (Lazim, 2007). Forecast can also be defined as a statement about what will happen in the future based in information available now. In other words, forecast and forecasting can be defined as the process of obtaining future values which can be used as inputs for the goals and objectives of a firm by using several methods based on information available now. It is essential to be able to correctly define and understand the given data. This includes understanding what caused the numbers to result as it is. By simply relying on the quantitative method of forecasting, the values forecasted would be inaccurate and sometimes illogical. That is why forecasters should also consider the qualitative or subjective method together with the quantitative method while forecasting the given data to produce more accurate values that will benefit the firm involved. The process of collecting data took place before continuing to the forecasting process. For this project, the data used is the Expenditures spent on Johor Government Employees for training courses and development programmes for the month January 2008 to September 2012. The data obtained is in monthly term with 57 data points. The forecast been made is for the next three months; October, November, and December of 2012.

Forecasting process begins by verification of time series components, which four crucial components are trend component, cyclical component, seasonal component and irregular component. This process can be done by doing observation on the original data plot and residual method. Then, identify the forecasting methods which are suitable to be used based on the types of time series components. Three methods were chosen and competed with each other to obtain the best method. Those three methods calculated the fitted value for the latest three years of the actual data. In this project, the latest three years of the actual data which are July, August and September of 2012 had been the left out and their fitted values will be evaluated.

Model evaluation later took place and evaluates the accuracy of the forecasting methods by comparing the actual data with the fitted value. Several techniques were used such as Mean Squared Error (MSE), the Root Mean Squared Error (RMSE), Geometric Root Mean Squared Error (GRMSE) and the Mean Absolute Percentage Error (MAPE). To calculate the error, subtract the fitted value from the actual value. Meanwhile, to obtain the squared error, simply by squaring the error value calculated. For MSE, RMSE, GRMSE, and MAPE, the following formulas have been used:

MSE = m=1net+m2n
RMSE = tnet+12n
GRMSE = tneitM2212n
MAPE = t=1netyt*100n

Where, n = number of forecast error terms generated by the model t = time et=yt-ŷt yt = the actual observed value at time t ŷt = the fitted value at time t. m = number of step (one-step=1, two-step=2, three-step=3) et+1=yt+1-Ft+1 yt+1 = the actual observation at the point t+1 Ft+1 = the forecast value of yt+1

Through the two forecasting methods, they were compared and ranked according to the smallest value of error measure. Based on the overall ranking the best method was chosen to forecast the expenditures that will be spent for the next three month ahead.

3.3.1 SELECTING THE FORECASTING MODEL a) TRIPLE EXPONENTIAL SMOOTHING (HOLT - WINTER’S METHOD)
There are three types of Exponential Smoothing Methods: 1. Single Exponential Smoothing (for series of data that shows no trend component nor seasonality); 2. Double Exponential Smoothing or Holt’s Method (for series of data only shows trend component but no seasonality); 3. Triple Exponential Smoothing or Holt – Winter’s Method (for series of data that shows both trend component and seasonality).
To use Holt – Winter’s Method, there are four important formula needed to be used, which are: * The exponentially smoothed series:

* The trend estimate:

* The seasonality estimate:

* Forecast m period into the future:

Where:
Lt = level of series. = smoothing constant for the data. yt = new observation or actual value in period t. = smoothing constant for trend estimate. bt = trend estimate. = smoothing constant for seasonality estimate.
St =seasonal component estimate. m = Number of periods in the forecast lead period. s = length of seasonality (number of periods in the season) = forecast for m periods into the future.
, , and are the unknowns smoothing constant to be determined with value lying between 0 and 1 and selected by the forecaster. As with all exponential smoothing methods, initial values are needed for the components to start the algorithm. To start the algorithm, the initial values for Lt, the trend bt, and the indices St must be set. To determine initial estimates of the seasonal indices we need to use at least one complete season's data (i.e. s periods). Therefore, we initialize trend and level at period s. * Initialize level as:

* Initialize trend as

* Initialize seasonal indices as:

b) POLYNOMIAL CURVE FITTING (SEASONAL DECOMPOSITION)
Another method is to fit the curve of the data series into a polynomial model. However, two types of model will be produced, one with original data points, and another one with a deseasonalized data, derived from the original data to dispose the seasonality component. Curve fitting differs from the statistical process of regression in that the latter is often the most rational way of achieving the former. In curvefitting, a greater emphasis is placed on the form of the curve which is to be used to match the data, whereas regression often is applied without much thought given to curve selection. Using “Add Trend Line” function will help to produce equations with different degree polynomials. The best curve fitting will be chosen by looking at the R2 value.
A deseasonalized data series can be obtain through a process of five steps:

Step 1: Finding Moving Average
To find moving average of the data, I will start with averaging 12 months data points. This is because in one cycle of year there are 12 12 months. Initializing at I6, where I represents the data of moving average and starting with the 6th because only from that point data can be averaged.

It= yt-5+ yt-4+yt-3+yt-2+ yt-1 yt+ yt+1+ yt+2+ yt+3+ yt+4+ yt+5+ yt+612 t = 6, 7, 8….. 57
3.4 FORECASTING SCENARIO

For model building (diff, techniques – diff model)
Hold out data point historical forecast
Future forecast
Policy planning and control t-2 t-1 t t+1 t+2 n n+1 n+2
T1 estimation period T2 evaluation period T3 (today/present)

2 step
1 step
1 step
2 step
10 steps ahead

3 step

The process of forecasting begins with estimation period, where a few models will be built, depending on the time series components that exist in the data. The estimation period must be long enough for the models to be significant, reliable and justifiable.
Then a few original data points are left out for evaluation. The evaluation period is where error measurements will be used and all the models’ fitted value will be competed against each other to find the least error and the best model.
Finally, the forecasting period is where future forecast are made using the best model that has been decided in estimation period. During this period, it is wise to derive a policy for planning and control.
3.4 ERROR MEASUREMENT

The accuracy of a prediction techniques will be evaluate and calculate by comparing value of the actual data with the projected data (Hanke, Reitsch, J., & A., 1998). Basic formula for an error is: et= yt- Ft
Where,
et = forecast error at time t yt = actual data at time t Ft = projected data at time t

Mean Squared Error (MSE) will be use as a benchmark in evaluating the best forecast technique. Forecast technique that shows the minimum value of MSE is the most appropriate technique that can be used. The formulae for the MSE are as follows:

MSE (Mean Squared Error)= t=1net2n
Where n = no. of sample

Besides other approaches using Mean Absolute Percentage Error (MAPE) also can help researcher to measure forecasting error. According to Ruzleeta Zakaria (2003), the MAPE is calculated based on distribution of the absolute deviation value of each period with the actual value for the period, and the absolute percentage of deviation will be averaged. He added that such an approach will be useful when the size or magnitude of a variable input is important in assessing the accuracy of predictions.

MAPEMean Absolute Percentage Error= t=1netytn*100
Where n = no. of sample

For any scale dependent measure, the existence of an outlier greatly affects the accuracy of the error measure. As a means of overcoming such problem and when confronting with a significantly large error term due to a particularly bad forecast, then the Geometric Root Mean Square Error (GRMSE) is the most useful alternative (Fildes, 1992).

GRMSE (Geometric Root Mean Square Error) = (tneit2)12n
Where n = no. of sample

3.5 SOFTWARE USED

The usage of computer software is able to analyze the time series data, besides the constructed mathematical models to explain the situation and historical features. They are then implemented and applied to predict the productivity growth for the next three years. In this project, the software that will be used is Microsoft Excel. Deseasonalizing data, fitting data into curves and calculations will be done with Microsoft Excel.

CHAPTER 4
DATA ANALYSIS AND RESULT

4.1 IDENTIFICATION OF THE DATA

To get a wider view in understanding the data easily, it is better to study the graph of the data as it shows the fluctuations of ups and downs from year to year more clearly. Figure 2 below shows the graph of productivity growth in Malaysia from year 1981 up to 2011.

Figure 2: Expenditures spent on Johor Government Employees for training courses and development programmes, January 2008 to September 2012 The highest value of expenditures spent on Johor Government Employees for training courses and development programmes was recorded in June 2012 with RM698384.10 spent due to strong sudden increase in number of civil servants employed since 2009. According to Circular of Johor Public Service Commission, Circular No. 241/2001 dated 31.10.2001; every state government employee needs to undergo three compulsory courses within three years of service after been confirmed. The three compulsory courses are: 1. General Induction Courses (Kursus Induksi Umum) a. For all state government employee regardless of department or offices. b. The contents are mainly about National Vision, Policies, Bureaucracy Procedures and so on. 2. Specific Induction Courses (Kursus Induksi Khusus) c. For all state government employees according to their units or departments or offices. d. The contents are mainly on office management, monetary/ allowances claims, applying days off and so on. 3. Character Paradigm Courses (Kursus Paradigma Sahsiah) e. For all state government employees and they will be mixed with staffs from different offices from the whole state. f. Focus mainly on integrity, building team work spirit, creating and conducive working area and so on.
Another high point is May 2012, which is also considered as due month for a lot of state government employees to fulfill their requirements to attend all three courses. Through the 57 months, JGSO DHRM, HRDU has recorded two least expenditure spent, which are on November 2008 and December 2008 with values of expenditure RM100397.30 and RM91391.70 respectively. This is closely related to the fact that Ramadhan, the fasting month was around these months. HRDU will spent less during Ramadhans because there should be no courses organized during fasting months. However, the expenditure exists because instead of organizing courses, during these months, a few spiritual activities were organized by HRDU.

After analysing the data and past history, it is crucial to verify the four main components in forecasting which are trend, cyclical, irregular and seasonal. Since the graph obviously exhibits a nonlinear trend, the pattern of the curve was identified by using Microsoft Excel to find the best trendline. Table 2 below shows the result of the trend identification process.
Table 2: Trend Identification Model | Equation | R2 | Quadratic | y = -39.68x2 + 9551x + 14181 | 0.496 | Cubic | y = 2.956x3 - 296.9x2 + 15571x + 11146 | 0.500 | Linear | y = 7249.x + 16445 | 0.493 | Logarithm | y = 12481ln(x) - 11573 | 0.427 |

The trend line model that were tested are quadratic, cubic, linear and logarithm. The largest R2 was the best suitable trend for the data. From the depicted figure, cubic model was chosen to represent the data trend. Figure 3 below shows the graph and trendline for the data.
Figure 3: : Expenditures spent on Johor Government Employees for training courses and development programmes and Trendline

The cyclical components refer to the rise and fall of the series over unspecified period of time. Moreover, cyclical component is the upward and downward change in the data pattern. In order to identify the component of cyclical, residual method has been used.
The residual method defines the measure of cyclical variation in terms of the percentage of trend and is expressed as:
Percentage of trend = ytTt ×100
Where yt is the actual value in period t and Tt is the estimated trend value in period t.
The trend, Tt, is represented by cubic equation as it is the best trend identified for the data.
The in the next page shows the cyclical characteristic of productivity growth:
Table 3: Cyclical Characteristic of the Productivity Growth Year | Month | Expenditure (yt) | Trend (Tt) | Residual | 2008 | Jan | 183393.8 | 26423.056 | 694.0673 | | Feb | 194382.4 | 41124.048 | 472.6733 | | Mar | 156399.2 | 55266.712 | 282.9899 | | Apr | 185293.4 | 68868.784 | 269.0528 | | May | 170483.2 | 81948 | 208.0383 | | Jun | 140171.2 | 94522.096 | 148.2946 | | Jul | 183174.5 | 106608.808 | 171.8193 | | Aug | 263102.3 | 118225.872 | 222.5421 | | Sep | 261289.3 | 129391.024 | 201.9377 | | Oct | 193298.2 | 140122 | 137.9499 | | Nov | 100397.3 | 150436.536 | 66.73731 | | Dec | 91391.7 | 160352.368 | 56.99429 | 2009 | Jan | 197362.4 | 169887.232 | 116.1726 | | Feb | 256204.2 | 179058.864 | 143.0838 | | Mar | 249204.3 | 187885 | 132.6366 | | Apr | 492374.3 | 196383.376 | 250.721 | | May | 294073.2 | 204571.728 | 143.7507 | | Jun | 449206.1 | 212467.792 | 211.4232 | | Jul | 396561.3 | 220089.304 | 180.182 | | Aug | 418211.4 | 227454 | 183.8664 | | Sep | 394263.3 | 234579.616 | 168.0723 | | Oct | 153294.3 | 241483.888 | 63.48014 | | Nov | 197453.4 | 248184.552 | 79.5591 | | Dec | 491472.3 | 254699.344 | 192.9617 | 2010 | Jan | 362541.5 | 261046 | 138.8803 | | Feb | 397385.3 | 267242.256 | 148.6985 | | Mar | 239434.2 | 273305.848 | 87.60669 | | Apr | 584263.2 | 279254.512 | 209.2225 | | May | 653226.1 | 285105.984 | 229.1169 | | Jun | 456364.2 | 290878 | 156.892 | | Jul | 453765.4 | 296588.296 | 152.995 | | Aug | 547955.3 | 302254.608 | 181.2893 | | Sep | 166466.9 | 307894.672 | 54.06618 | | Oct | 175464.2 | 313526.224 | 55.96476 | | Nov | 342242.7 | 319167 | 107.23 | | Dec | 229754.4 | 324834.736 | 70.72963 | 2011 | Jan | 543437.3 | 330547.168 | 164.4054 | | Feb | 398339.3 | 336322.032 | 118.4398 | | Mar | 202422.2 | 342177.064 | 59.15715 | | Apr | 589342.2 | 348130 | 169.288 | | May | 578315.8 | 354198.576 | 163.2745 | | Jun | 548427.3 | 360400.528 | 152.1716 | | Jul | 591950.8 | 366753.592 | 161.4029 | | Aug | 378410.1 | 373275.504 | 101.3756 | | Sep | 348745.2 | 379984 | 91.77892 | | Oct | 454392.2 | 386896.816 | 117.4453 | | Nov | 379266.8 | 394031.688 | 96.25287 | | Dec | 601285.1 | 401406.352 | 149.7946 | 2012 | Jan | 538204.4 | 409038.544 | 131.5779 | | Feb | 592563.4 | 416946 | 142.1199 | | Mar | 417982.92 | 425146.456 | 98.31504 | | Apr | 603641.1 | 433657.648 | 139.1976 | | May | 680200.6 | 442497.312 | 153.7186 | | Jun | 698384.1 | 451683.184 | 154.6181 | | Jul | 432438.2 | 461233 | 93.75699 | | Aug | 451481.3 | 471164.496 | 95.82244 | | Sep | 606124.3 | 481495.408 | 125.8837 |

The value of residual greater than 100 shows that during that month, the expenditure is higher than the trend line; while the value less than 100 shows that the expenditure is lower than the trend line. Cycle component can be said exist if the growth and decline of the residual are repeated in any period of month. From the table, it can be concluded that there are no cycle existed throughout the period of the data. This is because the growth and decline of the residuals are not consistent with only 14 data points showed decrement, while others show increment, and these flow does not show any pattern. The irregular component is identified when there are major turning points in the data plot. Based from the actual data plot, it can be said that irregular had happened several times. The original data plot shows that productivity growth fluctuates tremendously but the major turning points could be seen during the months November and December of 2008; October and November of 2009; September and October of 2010; August and September of 2011; and July and August of 2012. It is highly believed that during these months, expenditures spending dropped because these were the months of Ramadhan, hence the lower spending.

4.2 FORECASTING THE EXPENDITURE

The important part before forecasting is to find fitted values for the latest three years of the data by using several techniques and conduct the error measurement to test its accuracy. The error measurements that involved were Mean Square Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Percentage Error (MAPE) and Geometric Root Mean Square Error (GRMSE). The fitted values obtained were compared with the actual data by using these error measurements. Then, the most appropriate forecasting model was used to forecast for the year 2012, 2013 and 2014. Table 4 in the next page shows the value of actual data for the latest three years that will then been compared with the fitted values.

Table 4: Actual Values of Expenditures spent on Johor Government Employees for training courses and development programmes July – September 2012

Month | Actual Value | July | 432438.20 | August | 451481.30 | September | 606124.30 |

4.3 THE FORECASTING TECHNIQUES 4.3.1 TRIPLE EXPONENTIAL SMOOTHING (HOLT – WINTER’S METHOD)

For this method, a process of initializing level component, trend component and seasonal indices needs to be done. In order to do so, the following formula will be used:
Initializing level:
L0= 112 (y1+y2+ y3+ y4+ y5+ y6+ y7+ y8+y9+ y10+ y11+ y12)
= 112 (183393.8 + 194382.4 + 156399.2 + 185293.4 + 170483.2 + 140171.2 + 183174.5 + 263102.3 +261289.30+193298.2 + 100397.3 + 91391.7) L0 = 176898.042

Initializing trend: b0= 112 y13- y112 + y14- y212+ y15- y312+y16- y412+ y17- y512+ y18- y612+ y19- y712+ y20- y812+ y21- y912+y22- y1012+y23- y1112+y24- y1212 b0= 112 197362.4- 183393.812 + 256204.2- 194382.412+ 249204.3- 156399.2 12+492374.3- 185293.4 12+ 294073.2- 170483.212+ 449206.1- 140171.212+ 396561.3- 183174.512+ 418211.4- 263102.312+ 394263.3- 261289.3012+153294.3- 193298.212+197453.4- 100397.312+491472.3- 91391.712 b0 = 12964.6111

Initializing seasonal indices (there should be values from S1 to S12 because there are 12 period of months in one cycle of year)
S1= y1L0 S2= y2L0 S3= y3L0 S4= y4L0 S5= y5L0 S6= y6L0 S7= y7L0 S8= y8L0 S9= y9L0 S10= y10L0 S11= y11L0 S12= y12L0

S1= 183393.8176898.042 S2= 194382.4 176898.042 S3= 156399.2 176898.042 S4= 185293.4 176898.042
S5= 170483.2176898.042 S6= 140171.2176898.042 S7= 183174.5176898.042 S8= 263102.3 176898.042 S9= 261289.30176898.042 S10= 193298.2176898.042 S11= 100397.3 176898.042 S12= 91391.7176898.042 S1= 1.03672035 | S2= 1.09883862 | S3= 0.88412058 | S4= 1.04745874 | S5= 0.96373707 | S6= 0.79238413 | S7= 1.03548065 | S8= 1.48731042 | S9= 1.47706157| S10=1.09270967 | S11= 0.56754331 | S12=0.51663489 |
Then, using these formulae to fill the rows of Lt, bt, St and Ft+m : * The exponentially smoothed series:

* The trend estimate:

* The seasonality estimate:

* Forecast m period into the future:
Leaving out three data points, expenditure from July 2012 to September 2012, and forecasting those data points using Holt – Winter’s Method, the following results is obtained:

Table 5: July 2012 – September 2012 Forecast Result Using Holt – Winter’s Method MONTH | FORECAST VALUE | July | 1083028.429 | August | 1813448.489 | September | 2057018.317 | Table 6: Holt – Winter’s Method Error Measurement MSE | RMSE | MAPE (%) | GRMSE | 21.2114 | 4.605588 | 105.9208 | 4.228899 |

4.3.2 POLYNOMIAL CURVE FITTING (SEASONAL DECOMPOSITION)

For this method, two polynomial curve fitting will be used. A forecast projection will be determined by inserting independent variable t, (month) into the equations, to find the forecast expenditure values for the months of July 2012 – September 2012. The first curve fitting is done to the original data, which has been found earlier, which is a cubic curve equation as below: y = 2.956x3 - 296.9x2 + 15571x + 11146
The second curve fitting is obtained by finding a trend line of a deseasonalized data,

Referring to figures in Table 5, weightage 11 shows the smallest value for each MSE, RMSE and GRMSE compared to other weightages with 11.0868, 3.329685 and 2.862153 respectively. Thus, it is ranked at the first place for weightage with smallest error and will be used in constructing forecasting model for productivity growth. However, this error value needs to be compared with other error value from other methods to identify the most appropriate model for forecasting. Calculations of MSE, RMSE, MAPE and GRMSE for each weightage tested in Moving Average method are shown in Appendix A.

4.3.2 SINGLE EXPONENTIAL SMOOTHING

For this method, nine values of α were tested by using trial an error method ranging from 0 to 1. Every fitted value was tested and compared with the actual value in order to select the best α to representing Single Exponential Smoothing in comparing with the other two methods. Table 6 shows the comparison of error measurement between every α.

Table 6: Comparison of Error Measurements between α Using Single Exponential Smoothing Method α | MSE | RMSE | MAPE (%) | GRMSE | 0.1 | 6.482920576 | 2.546158003 | 114.9493346 | 2.455052062 | | | | | | 0.2 | 14.2667583 | 3.777136256 | 122.6061469 | 3.145914877 | 0.3 | 15.60255544 | 3.950007017 | 125.6944582 | 3.307129635 | 0.4 | 17.03403823 | 4.127231303 | 128.2784084 | 3.379099704 | 0.5 | 18.52447033 | 4.304006311 | 129.8635131 | 3.358652356 | 0.6 | 20.07594594 | 4.480618924 | 130.2176425 | 3.207836544 | 0.7 | 21.72843483 | 4.661376924 | 129.3017811 | 2.842925369 | 0.8 | 23.55001421 | 4.852835688 | 127.1528907 | 1.929380653 | | | | | | 0.9 | 25.62891848 | 5.062501207 | 123.8313908 | 2.442985025 | | | | | |

Model with α value 0.1 shows the smallest value for each MSE, RMSE and MAPE compared to other α with 6.482920576, 2.546158003 and 114.9493346% respectively. Thus, it is ranked at the first place for α with smallest error and will be used in constructing forecasting model for productivity growth. Calculations of MSE, RMSE, MAPE and GRMSE for each α tested in Single Exponential Smoothing method is shown in Appendix B.

4.4 COMPARISON BETWEEN MODELS

After running three types of methods, they were then moved to the next step which is competing with each other. Each method competed in terms of error measurements where the method with smallest error was ranked at the first place. This method was then used to forecast the value of Malaysia’s productivity growth for 2012, 2013 and 2014. Table 11 summarizes the comparison and evaluation of the three methods.

Table 11: Comparison and Rank of Error Measurements between Moving Average, Single Exponential Smoothing and Box- Jenkins Methodology MODEL | MSE | RMSE | MAPE (%) | GRMSE | TOTAL RANK | OVERALL RANK | Moving Average | 11.0868 | 3.32969 | 169.2359 | 2.86215 | 12 | 3 | RANK | 3 | 3 | 3 | 3 | | | Single Exponential Smoothing | 6.482921 | 2.546158 | 114.949335 | 2.455052 | 5 | 1 | RANK | 1 | 1 | 1 | 2 | | | Box-Jenkins Methodology | 6.53825 | 2.557 | 143.136 | 2.097244 | 7 | 2 | RANK | 2 | 2 | 2 | 1 | | |

Based on the table, it shows that Single Exponential Smoothing won the competition as it depicted the lowest error for MSE, RMSE and MAPE. Although the GRMSE is higher than that calculated in Box-Jenkins Methodology, the lowest rank of all still favored to Single Exponential Smoothing. In overall, Single Exponential Smoothing was ranked first, Box-Jenkins Methodology in the second place, while Moving Average in third. Thus, Single Exponential Smoothing was chosen to forecast for the next three years ahead.

4.5 CONCLUSION OF ERROR

Through several stages of testing and calculating error measurement, each method produces error values that can be considered large. This indicates that this study needs to be improved in order to produce more reliable and accurate forecast value. Large error occurs because of inconsistent productivity growth in the past history especially when facing with unexpected economic changes such as economic downturn in 1985, 1986, 1998, and 2009. Apart of that, demand from both domestic and external also affecting growth of productivity.

4.6 FORECAST VALUE

In the final stage, the Single Exponential Smoothing method with α of 0.1 was implemented and the forecasted values for the year 2012, 2013 and 2014 are shown in the table below while Figure 6 shows the graphical presentation of the forecasted values.

Table 12: Forecasted Value of Malaysia’s Productivity Growth (%) for the Year 2012, 2013 and 2014 DATE | FORECASTED VALUE (%) | 2012 | 3.7 | 2013 | 3.8 | 2014 | 3.79 |

Figure 5: Graph of Forecasted Value of Malaysia’s Productivity Growth (%) for the Year 2012, 2013 and 2014

CHAPTER 5
CONCLUSION

5.1 SUMMARY

Objective of this project is to identify the best forecasting model to predict Malaysia’s productivity growth in annual basis for the year 2012, 2013 and 2014. With all the methods and stages of works performed through this project, the objective has been achieved in which the Single Exponential Smoothing was chosen as the best technique. Meanwhile, the forecasted values of Malaysia’s productivity growth obtained from this technique for the year 2012, 2013 and 2014 are 3.7%, 3.8% and 3.79% respectively. The accuracy of forecasting results can be tested using several error measurements such as MSE, RMSE, MAPE and GRMSE. However, as shown in previous chapter, the results of this project show quite large error values due to inconsistence and instability of the past actual data. To improve this study, further test and study need to be performed in order to get more reliable and accurate forecasts.

5.2 IMPLICATION
Implications of this project to the organization (MPC) and industries in Malaysia are as follows:

5.2.1 TO THE ORGANIZATION (MPC)

From the results of this project, MPC could consider in adopting another method of calculating productivity growth for Malaysia. Although it needs further study and improvements, the use of Single Exponential Smoothing method is practical as it is easy to calculate, not as tedious as other more complicated methods and widely used and popular among organizations. Furthermore, this study could allow MPC to respond to unexpected events such as when the forecast value of the productivity growth shows a sudden drop. Through the analysis, MPC is able to prepare and provide information and training to industries in Malaysia on how to improve their productivity.

5.2.2 TO THE INDUSTRIES

Once MPC had forecast productivity growth, they will analyse all factors contributing to the changes in growth, in which they will conduct thorough analysis on each industries and economic sectors. They will then provide consultation and training to particular industries in improving and boost their productivity performance. Each industry plays an important role in enhancing Malaysia’s productivity growth. Thus, with a good start in forecasting process, it would help industries in Malaysia to perform better and lead Malaysia to a new outstanding level in the world.

5.3 LIMITATION

Several problems had arisen through completing this project. The main problem that has been faced was lacked of knowledge and experience in applying forecasting technique by using real time series data. Moreover, techniques applied might be too technical as they only take into account the actual figures without considering the reasoning and logical explanation behind the fluctuations such as economic downturn in past history. Besides that, there is lack of data as the data collected only have 31 data points. Furthermore, the forecast value may not be too accurate because in the real world, productivity in Malaysia is influenced by five major factors namely capital, labor, energy, materials and services. Any changes in one of these factors would bring significant changes in productivity performance of this country.

5.4 RECOMMENDATION

Future researchers are encouraged to conduct research on other forecasting techniques available as well, such as Holt’s Method and Time Series Regression. Comparisons among these techniques would assist the researchers in obtaining more accurate result. Moreover, the researchers also need to be alert with the changes in domestic and global economic besides focusing on each factor contributing to productivity performance. In addition of that, the primary data collection process should be done thoroughly to ensure that the values of productivity calculated are as accurate

TABLE OF CONTENTS

ABSTRACT………………………………………………………………………………...v

CHAPTER 1 1
INTRODUCTION 1
1.1 PRACTICUM INTRODUCTION 1 1.1.1 PRACTICUM OBJECTIVES 2
1.2 ORGANIZATION BACKGROUND 3 1.2.1 CORE VALUE OF ORGANIZATION 5 1.2.2 ORGANIZATION’S VISION 5 1.2.3 ORGANIZATION’S MISSION 5 1.2.4 ORGANIZATION’S OBJECTIVES 6 1.2.5 ORGANIZATION’S LOGO 6 1.2.6 BACKGROUND OF MPC HEADQUARTERS Error! Bookmark not defined.

CHAPTER 2 7
PROJECT DEFINITION 7
2.1 PROBLEM STATEMENT 7
2.2 PROJECT OBJECTIVE 8
2.3 PROJECT SCOPE 8
2.4 SIGNIFICANCE OF THE PROJECT 9

CHAPTER 3 10
METHODOLOGY 10
3.1 SOURCE OF DATA 10
3.2 DATA COLLECTION 10
3.3 ANALYSIS OF DATA: FORECASTING TECHNIQUE 12 3.3.1 SELECTING THE FORECASTING MODEL 14
3.4 ERROR MEASUREMENT 16
3.5 SOFTWARE USED 17

CHAPTER 4 18
DATA ANALYSIS AND RESULT 18
4.1 IDENTIFICATION OF THE DATA 18
4.2 FORECASTING THE GROWTH 24
4.3 THE FORECASTING TECHNIQUES 25 4.3.1 MOVING AVERAGE METHOD Error! Bookmark not defined. 4.3.2 SINGLE EXPONENTIAL SMOOTHING 26 4.3.3 BOX- JENKINS METHODOLOGY Error! Bookmark not defined.
4.4 COMPARISON BETWEEN MODELS Error! Bookmark not defined.
4.5 CONCLUSION OF ERROR 26
4.6 FORECAST VALUE 26

CHAPTER 5 26
CONCLUSION 26
5.1 SUMMARY 26
5.2 IMPLICATION 26 5.2.1 TO THE ORGANIZATION (MPC) 26 5.2.2 TO THE INDUSTRIES 26
5.3 LIMITATION 26
5.4 RECOMMENDATION 26

REFERENCES………………………………………………………………………...…viii

APPENDICES…………………………………………………………………………..…ix

LIST OF TABLES

Table 1: Productivity growth in Malaysia...............................................................................9
Table 2: Trend Identification.................................................................................................18
Table 3: Cyclical Characteristic of the Productivity Growth................................................20
Table 4: Actual Values of Productivity Growth in Malaysia for 2009-2011........................22
Table 5: Comparison of Error Measurements between Eleven Weightages Using
Moving Average Method......................................................................................................22
Table 6: Comparison of Error Measurements between α Using Single
Exponential Smoothing Method...........................................................................................23
Table 7: BIC Value for ARIMA (1,0,0)……………………………..…………………......26
Table 8: BIC Value for ARIMA (2,0,0)……………………………………..……………..26
Table 9: BIC Value for ARIMA (2,0,1)…………………………………………..………..26
Table 10: Error Measurements for ARIMA (1,0,0)………………………………..………26
Table 11: Comparison and Rank of Error Measurements between Moving Average,
Single Exponential Smoothing and Box-Jenkins Methodology...........................................27
Table 12: Forecasted Value of Malaysia’s Productivity Growth (%) for the Year 2012,
2013 and 2014.......................................................................................................................28

LIST OF FIGURES

Figure 2: MPC Logo……………………………………………………….…………….....4
Figure 2: Productivity Growth in Malaysia (1981-2011)…………………………...……..17
Figure 3: Productivity Growth in Malaysia (1981-2011) and Trendline..............................19
Figure 4: Residual ACF and PACF Plot..............................................................................25
Figure 5: Graph of Forecasted Value of Malaysia’s Productivity Growth (%) for the
Year 2012, 2013 and 2014...................................................................................................29