FAKULTY OF COMPUTER AND MATHEMATICAL SCIENCES

BACHELOR OF SCIENCE (HONS.) MANAGEMENT MATHEMATICS

ASSIGNMENT 1 : FORECAST INVESTMENT PRICE FOR GAS

(PETRONAS GAS BERHAD)

PREPARED BY :

AMIRAH SYADIAH BINTI MOHAMAD SHUKOR 2020970857

NUR FATIHAH BINTI SUKOR 2020985099

NURUL AINA AFIQAH BINTI AZHAR 2020971259

NURULHUSNA NAJIHAH BINTI HAIRULANUAR 2020964681

PREPARED FOR :

MADAM ISNEWATI BINTI AB MALEK

 INTRODUCTION

Petronas Gas Berhad (PGB) is a Malaysia-based gas infrastructure and utilities company. The
Company and its subsidiaries' major business activities are categorized into four segments: Gas
Processing, Gas Transportation, Utilities and Regasification. Petronas Gas Berhad (PGB) was listed
on the main market of Bursa Malaysia Securities Berhad. It is now a largest company on the local
bourse, in terms of market capitalisation.

Its Gas Processing activities include the processing of natural gas from gas fields offshore the East
Coast of Peninsular Malaysia into sales gas and other by-products, such as ethane, propane and
butane. Its Gas Transportation activities include the transportation of processed gas to end customers
of Petroliam Nasional Berhad (PETRONAS). Its Utilities activities include manufacturing, marketing
and supplying of industrial utilities to the petrochemical complexes in the Kertih and Gebeng Industrial
Area. Its Regasification activities include the regasification of liquefied natural gas (LNG) for
PETRONAS. PGB's subsidiaries include Pengerang LNG (Two) Sdn Bhd, Regas Terminal (Sg.
Udang) Sdn Bhd and Regas Terminal (Pengerang) Sdn Bhd.

In this report, we use the data of the investment prices of gas by years from 2010 until 2020. Basically,
the data showed that there are ups and downs of the investment gas prices as the pandemic Covis-19
is hitting all over the world. It is affecting the investment price for the company because the economic
level of our country is at risk so every investment company is affected. The figure below shows the
graph of the price gas investment on Petronas Gas Berhad (PGAS) throughout the years.
 RESEARCH OBJECTIVES

- To study the pattern investment price of gas (PETRONAS) throughout the year
2010 until 2020
- To identify the most appropriate model and use it for forecasting by analysing
each model's Mean of Square Error (MSE) and Mean Absolute Percentage
Error (MAPE)
- To forecast the investment price of gas for the following month

DATA DESCRIPTION

We obtained the data based on 10 years of the Investment Price for Gas (PETRONAS) in monthly.
The data starts from Dec 2010 until Nov 2020. A sample of size is 120. Estimation part of data is
selected from Dec 2010 - April 2018. Estimation part also has 90 data. Evaluation part of data is
selected from May 2018 - Nov 2020. Evaluation part also has 30 data. From the equation Y=β0 + β1X,
the dependent variable is Investment Price for Gas (PETRONAS) and the independent variable is the
period (monthly).

JAN FEB MAC APRIL MAY JUNE JULY AUG SEPT OCT NOV DEC

2010 - - - - - - - - - - - 11.1

2011 11.18 11.28 11.44 11.26 11.32 13.22 13.48 13.22 12.98 13.1 13.2 15.2

2012 15.68 16.8 16.84 16.78 17.2 18 18.5 19.38 19 19.6 18.4 19.52

2013 18.56 18.32 19 19.72 21.24 20.9 20.84 20.14 22 24.52 23.86 24.28

2014 23.38 23.6 23.8 23.5 24.5 24.5 23.48 22.8 22.92 22.48 22.72 22.16

2015 22.2 23.06 23.02 22.72 21.9 21.26 22.08 21.32 21.96 22.98 22.92 22.7
2016 22.88 22.04 22 21.98 21.64 22 22.18 22.2 21.84 22 20.98 21.3

2017 20.88 20.02 19.76 18.48 18.56 18.54 18.74 18.4 17.9 18.04 15.88 17.48

2018 17.9 17.62 17.84 17.82 17.58 17.3 18.74 18.7 18.9 18.28 18.94 19.2

2019 18.08 18.08 17.62 17.68 17.66 17.36 16.08 16.08 16.38 16.64 15.41 16.52

2020 15.9 16.52 15.4 15.4 19 16.88 16.8 16.5 16.42 15.74 17.12
-
 METHODOLOGY

i- Naive with Trend Model

Estimating methodology in which the realms of the last time are used as the
prediction of this period, without changing them or seeking to ascertain causal factors.
The naive model ( F t+m = y t where m = 1, 2, 3, 4, ...) is updated to take this function into
account. The use of this model is reasonably popular among organisations. Most of the
reasons for this appeal is that it can be used even in a relatively short time series.

One step-ahead forecast:

y
F t+1 = y t y t
t+1

where ;
y t = actual value at time t
y t+1 = actual value in preceding period

ii-​ ​Single Exponential Model

​This model has two names, one is a single exponential smoothing technique,
while the other is an exponential smoothing technique. This is the simplest model in the
exponential smoothing technique family of the whole model. It needs, as an alpha
parameter, that the values and forecasts be generated. This model has good reliability
since it measures the most current prediction. If the latest forecast is found to be correct,
then it is reasonable to base the subsequent forecast on these predictions. The
exponential smoothing method is actually an iterative process. It normally starts from a
certain point of departure. It's much easier to grasp, let's have a look at this case.
 This is the general equation for single exponential
smoothing Where;
F t+m ​= is the single exponential smoothed value in
period t+m for m = 1,2,3,4, ,
Y t ​= is the actual value in time period, t
a = the unknown parameter, alpha, valuing between 0 and 1
F t = the forecast value for period, t

Let's​ ​take a look at m=1, for example. The equation ​is expected to​ ​be,

In general, the above method can be used to obtain the initial value. The key goal
though, is to find a model that not only suits well, but can also be correctly forecasted.
Now to measure the details, we're going to follow these easy steps:

︿
Step 1 : Set y 1 ​ ​, to be the starting value

Step 2 : Calculate next fitted value by using the equation above

Step 3 : Calculate et and et 2

︿
et = y t - y t

Step 4 : Determine the value of a

Step 5 : Calculate error
iii- Double Exponential Model

Often known as the Brown’s method. This approach is very useful where a
sequence consists of linear pattern characteristics. It is analogous to the single
exponential smoothing as we need to find the parameter, alpha, first in order to get
the smallest value of MSE. This approach has an upper hand relative to a single
exponential smoothing in the forecasting of potential prices. This is because the
double exponential smoothing method is capable of multiple-forecasting. In brief, it
will produce a prediction of one, two, three, and so on. These are the notations that
will be used for this method:

Step 1 : The exponentially smoothed series;

S t = ay t + (1 − a) (S t−1 + T t−1

Step 2 : The trend estimates;

T t = β (S t − S t−1 ) + (1 − β )T t−1
 ​ eriod into the future;
Step 3 : Forecasts ​m p

F T +m = S T + T T × m

Step 4 : Determine the value of parameters α and β with value ranging from 0 to 1

Step 5 : Find the error ( e ),

︿
et = y t − y t

error squared ( e2 ),

︿
et 2 = (y t − y t )2

Mean Squared Error (MSE),

Root Mean Squared Error (RMSE),

and Mean Absolute Percentage Error (MAPE)

 ANALYSIS AND RESULT

(a) Naive with Trend Model

From the calculation, the value of Sum Square Error (SSE) for estimation part is 108.821 while
the value of Mean Square Error (MSE) is ​1.236602​. On the other hand, the value for Mean of
Absolute Percentage Error (MAPE) is 4.366234. For the evaluation part, the value of Sum
Square Error (SSE) is 94.15345 while for Mean Square Error (MSE) is 3.128448. The value for
Mean of Absolute Percentage Error (MAPE) is 7.2134. The figure below shows the graph line
of the investment stock price and the forecast number using naive with trend model.
 (b) Single Exponential Smoothing Model

Below are the 𝞪 random values used to fit the model to the data set and resulting Mean Square
Error (MSE);

𝞪 MSE

0.2 1.704556

0.8 0.667774

0.824453 0.667355

0.9 0.671217

1 0.687998

Now that we have found 𝞪 =0.824453 to be the best (as smallest MSE) for this data set when
fitting the single exponential smoothing model.

From the calculation, the value of Sum Square Error (SSE) for the estimation part is 52.61758
while the value of Mean Square Error (MSE) is 0.58464. On the other hand, the value for Mean
of Absolute Percentage Error (MAPE) is 0.520318 . For the evaluation part, the value of Sum
Square Error (SSE) is 27.46505 while for Mean Square Error (MSE) is 0.915502. The value for
Mean of Absolute Percentage Error (MAPE) is 0.349089.The figure below shows the graph line
of the investment gas price and the forecast number using single exponential smoothing
model.
 (c) Double Exponential Smoothing Model

Below are the 𝞪 random values used to fit the model to the data set and resulting Mean Square
Error (MSE);

𝞪 MSE

0.1 1.509714

0.3 0.707644

0.344521 0.700254

0.7 0.935828

0.8 1.083453

Now that we have found 𝞪 =0.344521 to be the best (as smallest MSE) for this data set when
fitting the double exponential smoothing model.

From the calculation, the value of Sum Square Error (SSE) for estimation part is 56.82504
while the value of Mean Square Error (MSE) is 0.631389. On the other hand, the value for
Mean of Absolute Percentage Error (MAPE) is 0.055813. For the evaluation part, the value of
Sum Square Error (SSE) is 27.20548 while for Mean Square Error (MSE) is 0.906849. The
value for Mean of Absolute Percentage Error (MAPE) is 0.076776. Figure below shows the line
graph for double exponential smoothing model.
 CONCLUSION

As the conclusion, we will describe which among the 3 models that we used will
be our best model for this project. Error measures are used while we conduct this
project. As we know, a good forecast model is the one that generally generates
successively good forecasts values so error measure is used to differentiate between a
poor forecast model and a good forecast model. The error measures that we used are
Mean Squared Error (MSE) and Mean Absolute Percentage Error (MAPE).

On the basis of the size of the MSE calculated over the evaluation period, we
can conclude that the best model is the Double Exponential Smoothing with the value
of ​0.906849 ​which are the smallest value of MSE among the three models.

Model type : Model type : Model type:

Naive with Trend Single Double

MSE Exponential Exponential
𝞪 = 0.824453 𝞪 =0.344521

Estimation part 1.236602 0.58464 0.631389

Evaluation part 4.366234 0.915502 0.906849

From the value of MAPE that have been calculated over the evaluation period, we can
conclude that the best model is the Double Exponential Smoothing with the value of
0.076776​ which are the smallest value of among the three model.
 Model type : Model type : Model type:

Naive with Trend Single Double

MAPE Exponential Exponential
𝞪 = 0.824453 𝞪 =0.344521

Estimation part 3.138448 0.520318 0.055813

Evaluation part 7.2134 0.349089 0.076776

The value of MSE and MAPE indicates that double exponential smoothing is the best model
among the three models which are naïve with trend , single exponential smoothing and
double exponential smoothing model.