basic forecasting

R A R A S T YA S N U R I TA

TECHNICAL FORECASTING

PERAMALAN ?

TAHAPAN PERAMALAN

FAKTOR DALAM MEMILIH METODE PERAMALAN

MACAM METODE PERAMALAN

P E R T E M U A N 1 : K O N S E P T E K N I K P E R A M A L A N

OUTLINE

7 September 2012 2 Teknik Peramalan KS091330

SOME QUESTIONS

•  Forecast what? •  Forecast why? •  Forecast when?

7 September 2012 Teknik Peramalan KS091330 3

??

FORECAST?

Predictions of future events and conditions are called

forecasts, and the act of making such predictions is

called forecasting

Purpose : reduce the range of uncertainty within which

management judgements must be made

Difference between forecasting and planning ? •  Forecasting: what we think will happen •  Planning: what we think should happen


USE OF FORECASTING: OPERATIONS DECISIONS

TimeHorizon

AccuracyRequired

Number ofForecasts

ManagementLevel

ForecastingMethod

Processdesign Long Medium Single or few Top Qualitative

or causal

Capacityplanning,facilities

Long Medium Single or few Top Qualitativeand causal

Aggregateplanning Medium High Few Middle Causal and

time series

Scheduling Short Highest Many Lower Time series

Inventorymanagement Short Highest Many Lower Time series


USE OF FORECASTING: MARKETING & FINANCE

TimeHorizon

AccuracyRequired

Number ofForecasts

ManagementLevel

ForecastingMethod

Long-‐rangemarketingprograms

Long Medium Single or few Top Qualitative

Pricingdecisions Short High Many Middle Time series

New productintroduction Medium Medium Single Top Qualitative

and causal

Costestimating Short High Many Lower Time series

Capitalbudgeting Medium Highest Few Top Causal and

time series


MENGAPA HARUS MERAMALKAN?

Peramalan Tumbuh Karena: 1. Meningkatnya kompleksitas organisasi dan lingkungannya. 2. Dengan meningkatnya ukuran organisasi, maka bobot &

kepentingan suatu keputusan telah meningkat pula 3. Lingkungan dari kebanyakan organisasi telah berubah dengan

cepat 4. Pengambilan keputusan telah semakin sistematis 5. Pengembangan metode peramalan & pengetahuan yang

menyangkut aplikasinya telah memungkinkan adanya penerapan secara langsung oleh para praktisi daripada hanya dilakukan oleh para teknisi ahli


FORECASTING STEPS

1

• Problem Formulation • Collect Data • Analyze and Clean Data

2 •  Identify Pattern •  Select Methods

3 • Extrapolate Pattern/ Generate Forecasts • Evaluate Forecasts


CHOOSING FORECASTING TECHNIQUES

•  In choosing a forecasting techniques, the forecaster must consider the following factors: 1.  The time frame 2.  The cost-benefit basis

Berkaitan dengan jumlah item yang diramalkan, lamanya periode peramalan dan metode peramalan yang dipakai

3.  The availability of data 4.  The pattern of data 5.  The accuracy desired 6.  The ease of operation and understanding

Penggunaan metode peramalan yang sederhana, mudah dibuat dan mudah diaplikasikan akan memberi keuntungan bagi perusahaan.


TIME FRAME

•  Forecasts are generated for points in time that may be a number of •  Days (daily) •  Weeks (weekly) •  Months (monthly) •  Quarter (quarterly) •  Years (annually)

is called by Time Frame / Time Horizon


TIME FRAME (2)

•  The length of the time frame is usually categorized as follows: •  Immediate: less than one month •  Short time: one – 3 months •  Medium: > 3 and < 2 years •  Long term: > 2 years

Longer time frame makes accurate forecasting more difficult, with qualitative

techniques more useful as the time lengthens.


JANGKA WAKTU PERAMALAN

Dalam hubungannya dengan horizon waktu peramalan, diklasifikasikan menjadi: 1. Peramalan Jangka Panjang ( 2-10 th )

→ Untuk perencanaan produk & perencanaan sumber daya 2.  Peramalan Jangka Menengah ( 1-24 bulan )

→ Untuk menentukan aliran kas, perencanaan produksi & penentuan anggaran

3. Peramalan Jangka Pendek ( 1-5 minggu ) → Untuk mengambil keputusan dalam hal perlu tidaknya lembur, penjadwalan kerja dan lain-lain melingkupi keputusan kontrol jangka pendek


THE COST OF FORECASTING

•  When choosing a forecasting techniques, several costs are relevant, such as: •  Developing cost •  Actual operation cost •  Complexity

Some forecasting methods are operationally simple, while others are very complex. The degree of complexity can have a definite influence on the total cost of forecasting


THE AVAILABILITY OF DATA

•  Accuracy and the timeliness of the data that are available must be examined

•  The inaccurate or outdated historical data will yield inaccurate predictions

•  If the needed historical data are not available, special data collections procedures may be necessary


TYPE OF DATA

Cross-Sectional

•  Starting salary and GPA for graduates last spring •  Home upkeep cost in the past and current value for homes in an

area •  Labor hours, occupied bed days, and average length of stay for

hospitals last month •  VS Time Series

•  Unit sales of a product over time •  Total dollar sales for a company over time •  Number of unemployed over time •  Population of a city over time •  Daily mean temperature over time


TYPE OF DATA (2)

Cross-sectional VS Time Series

•  Cross-sectional: values observed at one point in time

•  Time series: chronological sequence of observation on a particular variable


TYPE OF DATA (3)

Bi/ Multivariate Univariate

Cross-Sectional Time Series

Time fashion


THE COMPONENTS OF A TIME SERIES

Data

Trend

Irregular

Cycle

Seasonal


THE PATTERN OF DATA

•  The data pattern must be considered when choosing a forecasting model •  Stationary •  Trend •  Seasonal •  Cycle •  Irregular •  Combination


TREND

•  Indicates the very long-term behavior of the time series

•  Typically as a straight line or an exponential curve •  Due to population, technology etc. •  This is useful in seeing the overall picture


SEASONAL

•  Regular pattern of up and down fluctuations within a year

•  Due to weather, customs etc. •  Each time period during the year has its seasonal

index, which indicates how much higher or lower this particular time usually is as compared to the others.


CYCLE

•  Gradually up and down movements that do not repeat each year

•  Peak •  Contraction •  Trough •  Expansion

•  Due to interactions of economic factors •  The cyclic variation is especially difficult to

forecast beyond the immediate future

Mean-R

everting

Time


IRREGULAR FLUCTUATION

•  random, unsystematic, “residual” fluctuations •  Due to random variation or unforeseen events •  Short duration and non-repeating •  A forecast, even in the best situation, can be no

closer (on average) than the typical size of the irregular variation


QUARTERLY DATA BROKEN-DOWN*

Trend

Seasonal Index

Cyclic Behavior

Irregular

Sales Data

*For illustration purpose. 7 September 2012 Teknik Peramalan KS091330 24

Time (a) Trend

Time (d) Trend with seasonal pattern

Time (c) Seasonal pattern

Time (b) Cycle

Dem

and

Dem

and

Dem

and

Dem

and

Random movement

FORMS OF FORECAST MOVEMENT


THE ACCURACY DESIRED

•  In some situations a forecast that is in error by as much as 20% may be acceptable

•  In other situations a forecast that is in error by 1% might be disastrous

The accuracy that can be obtained using any particular forecasting method is always an important consideration


KETEPATAN METODE PERAMALAN

Jika xi merupakan data aktual untuk periode i dan Fi merupakan ramalan untuk periode yang sama, maka kesalahan didefinisikan sebagai :

ei = xi - Fi

Jika terdapat nilai pengamatan ramalan untuk periode waktu, maka akan terdapat n buah kesalahan. Dan ukuran statistik berikut dapat didefinisikan: •  Mean Error

ME = •  Mean Absolute Error/Deviation

MAE/MAD = •  Sum of Squared Error

SSE = •  Mean Squared Error Root Mean Squared Error (RMSE)

MSE = •  Standard Deviation of Error

SDE =

∑=

n

i

ine

1

∑=

n

i

ine

1

∑=

n

iie

1

2

∑=

n

i

ine

1

2

)1/(2−∑ nei


CONTOH

Aktual (x) Forecast (F) Error (e)

3 6 -3

4 4 0

6 5 1

5 3 2

7 6 1

Mean Error: 0.2 Mean Absolute Error: 7/5


KETEPATAN METODE PERAMALAN(2)

•  Tujuan optimasi statistik sering sekali untuk memilih suatu model agar MSE/SSE minimum. Tetapi ukuran ini mempunyai kelemahan karena metode yang berbeda akan menggunakan prosedur yang berbeda dalam fase pencocokan.

•  Ukuran MSE ini juga t idak memudahkan perbandingan antar deret berkala yang berbeda-beda untuk selang waktu yang berlainan, karena MSE merupakan ukuran absolut.


UKURAN-UKURAN RELATIF

Percentage Error •  PEi =

Mean Percentage Error •  MPE =

Mean Absolute Percentage Error

•  MAPE =

( ix ! iFix)(100)

∑ =

n

i i nPE1

∑ =

n

i i nPE1


THE EASE OF OPERATION AND UNDERSTANDING

Manager Responsibility Understood Predict Confidence Decision

That’s why manager’s understanding of the forecasting system is crucial


CONCLUSION

•  The ‘best’ forecasting method for a given situation is not always the ‘most accurate’

•  The greater accuracy, is more complex forecasting techniques, also higher cost

The forecasting method that should be used in one that meets the needs of the situation at the least cost and with the

least inconvenience


METODE PERAMALAN

Teknik peramalan dibagi dalam 2 kategori utama: 1.  Metode KUANTITATIF

•  Deret berkala (9me series) •  Metode kausal/eksplanatoris Contoh :

•  Moving Average, Exponen9al Smoothing •  Dll.

2.  Metode KUALITATIF •  Eksplanatoris •  Norma9f

Contoh : (Judgmental Forecas9ng Method) •  Subjec9ve Curve FiKng •  Delphi Method •  Technological Comparison


PERAMALAN KUANTITATIF

Peramalan kuantitatif dapat diterapkan bila: 1.  Tersedia informasi tentang masa lalu 2.  Informasi tersebut dapat dikuantitatifkan dalam bentuk data

numerik 3.  Dapat diasumsikan bahwa beberapa aspek pola masa lalu

akan terus berlanjut di masa mendatang


PERAMALAN KUANTITATIF (2)

Terdapat 2 jenis model peramalan yang utama: 1. Model Deret Berkala (Time Series)

Pendugaan masa depan dilakukan berdasarkan nilai masa lalu dari suatu variabel dan atau kesalahan masa lalu Tujuannya :

•  Menemukan pola dalam data historis •  Mengekstrapolasikan pola dalam deret data historis •  Mengekstrapolasikan pola tersebut ke masa depan

2. Model Kausal/Eksplanatoris •  Mengasumsikan bahwa faktor yang diramalkan menunjukkan suatu hubungan

sebab akibat dengan satu atau lebih variabel bebas •  Tujuannya untuk menemukan bentuk hubungan tersebut & menggunakannya

untuk meramalkan nilai mendatang dari variabel tak bebas Model deret berkala seringkali dapat digunakan dengan mudah untuk

meramal, sedangkan metode kausal dapat digunakan dengan keberhasilan yang lebih besar untuk pengambilan keputusan dan kebijaksanaan


PERAMALAN TIME SERIES

Langkah terpenting dalam memilih suatu metode time series yang tepat adalah dengan mempertimbangkan jenis pola data.

Sistem diperlakukan sebagai kotak hitam (black box) dan tidak ada

usaha untuk menemukan faktor yang berpengaruh pada perlakuan sistem tersebut.

Terdapat 2 alasan memperlakukan sistem sebagai kotak hitam:

•  Sistem itu mungkin tidak dimengerti dan kalaupun hal itu diketahui mungkin sangat sulit mengukur hubungan yang dianggap mengatur perilaku sistem tersebut

•  Perhatian utamanya mungkin hanya untuk meramalkan apa yang akan terjadi dan bukan mengetahui mengapa itu terjadi.

Sering peramalan dapat menggunakan baik pendekatan kausal

maupun time series.


PERAMALAN KAUSAL/EKSPLANATORIS

Mengasumsikan bahwa terdapat hubungan sebab akibat diantara input dengan output dari suatu sistem. Sistem dapat berupa ekonomi nasional, pasar suatu perusahaan, rumah

tangga, dll. Contoh: Dalam hubungan fisika yang terkenal, yaitu hukum boyle, menyatakan:

P : Tekanan N: Jumlah Molekul V:Volume : Faktor Proporsi

Dengan input diketahui, outputnya dapat diramalkan

VNP ρ=

ρ


EXAMPLE OF TIME SERIES MODEL

t Dt Ft1 120 119.523812 124 121.180953 119 122.83814 124 124.495245 125 126.152386 130 127.809527 129.46667

Intercept (a) 117.8666667Slope (b) 1.657142857

Yt = a + b(t)

F7 = 117.87 + 1.66 (7) = 129.47 = sales forecast for next year

Dt = actual sales

Ft = forecasted sales

t = time period (e.g. year)


EXAMPLE OF CAUSAL MODEL

It Dt Ft34.6 120 121.1503335.7 124 123.7864936.3 119 125.224435.2 124 122.5882435.7 125 123.7864936.4 130 125.4640537.6 128.33987

Intercept (a) 38.23093682Slope (b) 2.396514161

Yt = a + b(t)

F7 = 38.23 + 2.397 (7) = 128.34 = sales forecast for next year (year 7)

Dt = actual sales in year t

Ft = forecasted sales

It = median family income (000’s)


STATISTIK DESKRIPTIF YANG BERGUNA

•  Data Univariat •  Mean / nilai tengah (median) •  Standard deviation (S) •  Variance (S2) •  Mean absolute deviation (MAD) •  Mean Square (MS) •  Sum Square Deviation (SSD) •  Root Mean Square (RMS) •  Kemencengan (Julur) / Skewness


STATISTIK DESKRIPTIF YANG BERGUNA (2)

• Data Bivariat •  Covariance

Covxy =

•  Koefisien Korelasi r =

•  Time Series with lag (k) Auto-cov = (k keterlambatan)

Auto-r = (k keterlambatan)

))((11 YXn yx ii −−− ∑

ss yx

xycov

))((1

11

xxkn xx kt

n

ktt −−

−− −+−∑

2

1

1

)(

))((1

1

x

xxkn

n

tt

kt

n

ktt

x

xx−

−−−−

∑

∑

−

−+−


TASK TO DO..

1.  Cari 2 set data untuk tipe time series dan cross-sectional , juga 5 set data utk pola stationary, trend, seasonal, cycle, dan irregular (total : 7 set data)

•  Sertakan sumber data •  Beri uraian minimal satu paragraf terhadap tiap set data yang telah

diperoleh

2.  Essay mengenai Judgmental Forecasting Method dari berbagai sumber.

3.  Berikan tanggapan mengenai ukuran akurasi yang bermacam-macam, adakah ketentuan penggunaan ukuran akurasi tertentu?

Sifat tugas : individu, softcopy, submit E-Learning, dikumpulkan pertemuan depan 13 Sept 2012 sebelum perkuliahan


CONTACT INFORMATION

• Raras Tyasnurita S.Kom., MBA •  E-mail: [email protected]

Room TC.220 Information Systems Department Faculty of Information Technology Institut Teknologi Sepuluh Nopember


basic forecasting

Documents

the availability of

the pattern of data

the time frame 2

the accuracy desired

penerapan secara langsung

benefit basis berkaitan

tahapan peramalan faktor

memilih metode peramalan