CODE / COURSE NAME

STQS3113 / STATISTICAL MODELING

SEMESTER / SESSION

SEMESTER 1 / SESSION 2022/2023

ASSIGNMENT 2

LECTURER

DR. RAZIK RIDZUAN BIN MOHD TAJUDDIN

GROUP MEMBERS

NO NAME MATRIC NO.

1 LOO WING LAM A179541
2 TAN CHENG HANG A180122
3 YEE SHU YONG A180847
4 SIA CHANG ANN A181494
5 HENG KAI SHENG A181555
To find an appropriate time series model for the data given by our supervisor, we have
plotted a line chart on the given data as shown above to observe and discover the
discernible pattern of the rate of tourists staying in the hotel. The x-axis represents the
time (4 quarters each year from 2017 to 2022 consecutively) while y-axis represents the
rate of tourists. Based on the chart, we observed a linear upward trend in the rate of
tourists over the 6 years period from 2017 to 2022. A fairly regular seasonal pattern is
detected and there is a small increasing seasonal variation. The magnitude of seasonal
variation increase slightly as the level of the time series increases. Therefore,
multiplicative decomposition method or multiplicative Holts-Winter method can be
used to forecast future rate of tourists staying in the hotel.

After careful consideration, we decided to use multiplicative Holts-Winter

method in the forecasting because the trend and seasonal factors of this given data may
be changing over time. Since the preference and behavior of tourists may change over
time, we must focus on more recent data to predict the future data. By applying
multiplicative Holts-Winter method, we can put more weight on more recent
observations of the rate of tourist staying in the hotel rather than more remote
observations.

1
 Firstly, we obtain the initial values of the level 𝑙0 = 34.0041 and the growth
rate 𝑏0 = 0.5327 from “=intercept()” and “=slope()” function in excel by fitting a least
squares trend line with using 4 years of historical data which for 𝑡 = 1,2, … ,16. Then
𝑦
we detrend the data by computing 𝑆𝑡 = 𝑦̂𝑡 and the average seasonal values for each
𝑡

season which are

𝑆1 + 𝑆5 + 𝑆9 + 𝑆13
̅ =
𝑆[1] = 1.2450
4
̅ = 0.7591
𝑆[2]
̅ = 0.9492
𝑆[3]
̅ = 1.0465
𝑆[4]
for quarter 1,2,3 and 4. Next, we compute the seasonal factors 𝑠𝑛−3, 𝑠𝑛−2, 𝑠𝑛−1 and
𝑠𝑛0 by multiplying the average seasonal values with the normalizing constant 𝐶𝐹 =
𝐿
̅ 𝐶𝐹 and obtain
= 1.00046 such that 𝑠𝑛𝑖−𝐿 = 𝑆[𝑖]
∑𝐿𝑖=1 𝑆̅[𝑖]

𝑠𝑛−3 = 1.2451
𝑠𝑛−2 = 0.7591
𝑠𝑛−1 = 0.9492
𝑠𝑛0 = 1.0466.

After that, we get the point forecast of 𝑦̂1 (0) = (𝑙0 + 𝑏0 )𝑠𝑛1−4 = (34.0041 +
0.5327) ∗ (1.2451) = 43.0007 . The forecast error is 𝑦1 − 𝑦̂1 (0) = 41.7275 −
43.0007 = −1.2733 . We set a predetermined smoothing constant value of 𝛼 =
0.5, 𝛾 = 0.5 and 𝛿 = 0.5 to update the estimates 𝑙𝑡 , 𝑏𝑡 , 𝑠𝑛𝑡 , 𝑦̂𝑡 and the forecast error
with
𝑦𝑡
𝑙𝑡 = 𝛼 ( ) + (1 − 𝛼)(𝑙𝑡−1 + 𝑏𝑡−1 )
𝑠𝑛𝑡−𝐿
𝑏𝑡 = 𝛾(𝑙𝑡 − 𝑙𝑡−1 ) + (1 − 𝛾)𝑏𝑡−1
𝑦𝑡
𝑠𝑛𝑡 = 𝛿 ( ) + (1 − 𝛿)𝑠𝑛𝑡−𝐿
𝑙𝑡
𝑦̂𝑡 (𝑡 − 1) = (𝑙𝑡 + 𝑏𝑡 )𝑠𝑛𝑡−𝐿
𝑓𝑜𝑟𝑒𝑐𝑎𝑠𝑡 𝑒𝑟𝑟𝑜𝑟 = 𝑦𝑡 − 𝑦̂𝑡 (𝑡 − 1)
for 𝑡 = 1,2, … ,24. Later, we use an analysis tool in excel which named solver to solve
the optimum value of 𝛼, 𝛾 and 𝛿 that gives the smallest value of 𝑆𝑆𝐸. In the end we got
𝛼 = 0, 𝛾 = 0.0682 and 𝛿 = 0 which gives 𝑆𝑆𝐸 = 48.1934 . After obtaining the
2
optimum smoothing constant, we started to forecast the rate of tourists for year 2023 by
using 𝑦̂𝑇+𝜏 (𝑇) = (𝑙 𝑇 + 𝜏𝑏𝑇 )𝑠𝑛 𝑇+𝜏−𝐿 for 𝜏 = 1,2,3,4 and obtain
𝑦̂25 (24) = (𝑙24 + 𝑏24 )𝑠𝑛25−4 = 58.9182
𝑦̂26(24) = (𝑙24 + 2𝑏24 )𝑠𝑛26−4 = 35.9216
𝑦̂27(24) = (𝑙24 + 3𝑏24 )𝑠𝑛27−4 = 44.9204
𝑦̂28(24) = (𝑙24 + 4𝑏24 )𝑠𝑛28−4 = 49.5246

The point forecast and the forecast for year 2023 are shown in the graph below.

3
 APPENDIX

Alpha Gamma Delta SSE

0.0000 0.0682 0.0000 48.1934

Point
Year Quarter Rate of tourists t Level Growth Initial Fitted St Q Average S Seasonal Forecast Error
Forecast
-3 1 1.2450 1.2451
-2 2 0.7591 0.7591
-1 3 0.9492 0.9493
0 34.0041 0.5327 4 1.0465 1.0466
2017 1 41.7275 1 34.5368 0.5327 34.5368 1.2082 CF 1.0000 1.2451 43.0007 -1.2733
2 24.0419 2 35.0695 0.5327 35.0695 0.6855 0.7591 26.6213 -2.5795
3 32.3281 3 35.6022 0.5327 35.6022 0.9080 0.9493 33.7959 -1.4678
4 37.3287 4 36.1349 0.5327 36.1349 1.0330 1.0466 37.8174 -0.4887
2018 1 46.2132 5 36.6675 0.5327 36.6675 1.2603 1.2451 45.6537 0.5595
2 29.3463 6 37.2002 0.5327 37.2002 0.7889 0.7591 28.2387 1.1076
3 36.4829 7 37.7329 0.5327 37.7329 0.9669 0.9493 35.8185 0.6644
4 42.9777 8 38.2656 0.5327 38.2656 1.1231 1.0466 40.0473 2.9304
2019 1 48.9015 9 38.7983 0.5327 38.7983 1.2604 1.2451 48.3066 0.5950
2 31.1802 10 39.3310 0.5327 39.3310 0.7928 0.7591 29.8562 1.3240
3 37.7179 11 39.8636 0.5327 39.8636 0.9462 0.9493 37.8412 -0.1233
4 40.4202 12 40.3963 0.5327 40.3963 1.0006 1.0466 42.2773 -1.8571
2020 1 51.2069 13 40.9290 0.5327 40.9290 1.2511 1.2451 50.9595 0.2474
2 31.8872 14 41.4617 0.5327 41.4617 0.7691 0.7591 31.4736 0.4136
3 40.9783 15 41.9944 0.5327 41.9944 0.9758 0.9493 39.8638 1.1145
4 43.7725 16 42.5271 0.5327 42.5271 1.0293 1.0466 44.5073 -0.7348
2021 1 55.5586 17 43.0597 0.5327 1.2451 53.6124 1.9462
2 33.8509 18 43.5924 0.5327 0.7591 33.0911 0.7598
3 42.0764 19 44.1251 0.5327 0.9493 41.8864 0.1899
4 45.6423 20 44.6578 0.5327 1.0466 46.7372 -1.0949
2022 1 59.7668 21 45.1905 0.5327 1.2451 56.2653 3.5015
2 35.1919 22 45.7232 0.5327 0.7591 34.7085 0.4833
3 44.3198 23 46.2559 0.5327 0.9493 43.9091 0.4107
4 47.9137 24 46.7885 0.5327 1.0466 48.9672 -1.0534
2023 1 25 58.9182
2 26 35.9216
3 27 44.9204
4 28 49.5246