#5

Assalamualaikum, hello and salam sejahtera everyone.

In this entry, we will proceed with Time Series ( Forecasting ). To be honest, subject ni merupakan one of feveret subject sebab lecturer kami, Prof Madam Napisah adalah sangat cool dan cara beliau mengajar pun sangatlah details. Cumanya mungkin masa final exam tu tak berapa perform sebab jawapan-jawapan yang iolls bagi tak memuaskan beliau tsk tsk. Buku yang akan dijadikan reference ialah buku tulisan Mohd Alias Lazim, bertajuk " Introductory Business Forecasting, a practical approach. "




Definitions of Time Series data:

• A set of data collected or arranged in a sequences of order over a successive equal increment of time
• Data recorded at equal interval time for certain period
• yt = f(t)

1) It is significant to note that for most economic time series, they have tendency to be influenced either directly or indirectly by one or more interrelated events/factors. 

Contohnya macam kalau production sesebuah kilang ataupun hasil ladang tu mungkin akan ada sedikit kesan akibat weather pattern, ups and down of economic cycles, war and change in the demand level of particular product. 

2) There are also series that exhibit the regular recurring effects, moving up and down at regular, fixed or unfixed, intervals.

Contohnya macam seasonal pattern of road accidents masa musim perayaan, increase in sales of train/flight tickets during school holiday session and etc etc.



3) The effect of occurrence of certain unpredictable events leave varying degrees of impact on the time series, some permanent, some temporary. Some effects take shorter time (short term memory), others may take longer time (long term memory) to diminish.



Objectives:

1. To identify and to describe the underlying structure and the phenomenon as depicted by the sequence of observations in the series
2. To determine the most suitable mathematical model to fit the data series and subsequently use the model to generate forecast values

Approaches:

1. Time domain - commonly used for stochastic observations - moving averages (MA), detrending and regression methods to detect and remove autocorrelation
2. Frequency domain - use spectral analysis, harmonic analysis, periodogram - much more difficult to understand (outside of scope of the book :p)

The Classical Decomposition Method

Component types:

1. Trend component       - Tt     
2. Cyclical component    - Ct
3. Seasonal component   - St 
4. Irregular component    - It

( ps: all 't' are subscript t, indicate that these are time related components for which t=1,2,3......T. )

Relationship between Components

• Multiplicative Effect - If the sizes of seasonal variation increase in accordance with the increase in the level of data series then the following relationship is appropriate -- yt is the product of all component

• Additive Effect - The series interacted in an additive manner, used when the absolute sizes of the seasonal variation are independent of each other. The seasonal variation are not affected by change in the level of the series -- additive assumption is the sum of all components.

And details about each of components (trend, cyclical, seasonal, irregular) semuaaaaaa ada dalam buku, dengan steps macam mana nak calculate using excel spreadsheet, nak lukis graph forecast, reason-reason bagai tu semua confirm ada dalam tu, so all we need to do is refer to the book. Yeah! :p 

See you soon!
Byeeeeeee :]