TEII Methodology for Forecasting

1
TEI_at_I Methodologyfor Forecasting

• Shouyang Wang
• Academy of Mathematics and Systems Science
• Chinese Academy of Sciences
• Jointly with Lean Yu and K.K.Lai
• Email sywang_at_amss.ac.cn
• http//madis1.iss.ac.cn and www.amss.ac.cn

2
Outline

• Introduction
• The TEI_at_I methodology for crude oil price
  forecasting
• A simulation study
• Concluding remarks

3
Introduction I

• Importance of oil price forecasting The role of
  oil in the world economy becomes more and more
  significant because nearly two-thirds of the
  worlds energy consumption comes from the crude
  oil and natural gas. For example,
• worldwide consumption of crude oil exceeds 500
  billion, roughly 10 of the USAs GDP.
• crude oil is also the worlds most actively
  traded commodity, accounting for about 10 of
  total world trade.

4
Introduction II

• Determination of oil price Basically, crude oil
  price is determined by its supply and demand, and
  is strongly influenced by many irregular future
  events like the weather, stock levels, GDP
  growth, political aspects and even peoples
  expectation.
• The above facts lead to a strongly fluctuating
  and interacting market whose fundamental
  mechanism governing the complex dynamics is not
  well understood.
• Furthermore, because sharp oil price movements
  are likely to disturb aggregate economic
  activity, researchers have shown considerable
  interests for volatile oil prices.
• Therefore, forecasting oil prices is an important
  and very hard topic due to its intrinsic
  difficulty and practical applications.

5
Introduction III

• Main literature about oil price forecasting
• Watkins, G.C., Plourde, A. How volatile are
  crude oil prices? OPEC Review, 18(4), (1994)
  220-245.
• Hagen, R. How is the international price of a
  particular crude determining? OPEC Review, 18
  (1), (1994) 145-158
• Stevens, P. The determination of oil prices
  1945-1995. Energy Policy, 23(10), (1995) 861-870
• Huntington, H.G. Oil price forecasting in the
  1980s what went wrong? The Energy Journal,
  15(2), (1994) 1-22.
• Abramson, B., Finizza, A. Probabilistic
  forecasts from probabilistic models a case study
  in the oil market. International Journal of
  Forecasting, 11(1), (1995) 63-72
• Morana, C. A semiparametric approach to
  short-term oil price forecasting. Energy
  Economics, 23(3), (2001) 325-338

6
Introduction IV

• Evaluation about literature
• There are only very limited number of related
  papers on oil price forecasting.
• The literature focuses on the oil price
  volatility analysis.
• The literature focuses only on oil price
  determination within the framework of supply and
  demand.
• It is, therefore, very necessary to introduce new
  method for crude oil price forecasting.

7
Outline
B. A New Methodology

• Introduction
• The TEI_at_I methodology for crude oil price
  forecasting
• A simulation study
• Concluding remarks

8
TEI_at_I Introduction (A)

• In view of difficulty and complexity of crude oil
  price forecasting, a new methodology named TEI_at_I
  is proposed in this study to integrate
  systematically text mining, econometrics and
  intelligent techniques and a novel integrated
  forecasting approach with error correction and
  judgmental adjustment within the framework of the
  TEI_at_I methodology is presented for improving
  prediction performance.

9
TEI_at_I Introduction (B)

• TEI_at_I is based on text mining
  econometrics intelligence (intelligent
  algorithms) _at_ integration. Using _at_ to
  replace is to emphasize the functional of
  integrations. The general framework structure is
  shown in the following figure.

10
The general framework of TEI_at_I
11
Man-machine interface (MMI) module

• The man-machine interface (MMI) is a graphical
  window through which users can exchange
  information within the framework of TEI_at_I.
• it handles all input/output between users and the
  TEI_at_I system.
• it can be considered as open platform
  communicating with users and interacting with
  other components of the TEI_at_I system.

12
Web-based text mining module

• Crude oil market is an unstable market with high
  volatility and oil price is often affected by
  many related factors.
• In order to improve forecasting accuracy, these
  related factors should be taken into
  consideration in forecasting.
• Web-based text mining is used to explore the
  related factors.
• In this study, the main goal of web-based text
  mining module is to collect related information
  affecting oil price variability from Internet and
  to provide the collected useful information to
  the rule-based expert
• system forecasting module.

13
The main process of WTM module
14
Rule-based expert system (RES) module

• Expert system module is used to transform the
  irregular events into valuable adjusted
  information.
• That is, rule-based expert system is used to
  extract some rules to judge oil price abnormal
  variability by summarizing the relationships
  between oil price fluctuation and key factors
  affecting oil price volatility.
• See the paper for a detailed discussion.

15
Econometrical forecasting module

• It includes a large number of modeling techniques
  and models, such as autoregressive integrated
  moving average (ARIMA) model, vector
  auto-regression (VAR) model, generalized moment
  method (GMM), etc.
• Autoregressive integrated moving average (ARIMA)
  model is used here.
• ARIMA is used to model the linear pattern of oil
  price time series, while nonlinear component is
  modeled by artificial neural network (ANN).

16
ANN-based time series forecasting module

• The ANN used in this study is a three-layer
  back-propagation neural network (BPNN)
  incorporating the Levenberg- Marquardt algorithm
  for training.
• For an univariate time-series forecasting
  problem, the inputs of the network are the past
  lagged observations of the data series and the
  outputs are the future values.
• BPNN time-series forecasting model performs a
  nonlinear mapping. That is,

17
ANN-based time series forecasting module
18
Bases and bases management module

• The other modules of the TEI_at_I system have a
  strong connection with this module.
• For example, ANN-based forecasting module
  utilizes MB and DB, while the rule-based expert
  system mainly used the KB and DB.
• To summarize, the TEI_at_I system framework is
  developed through an integration of the web-based
  text mining, rule-based expert system and
  ANN-based time series forecasting techniques.

19
Summary

• In this framework, econometrical models (e.g.,
  autoregressive integrated moving average (ARIMA))
  are used to model the linear components of crude
  oil price time series (i.e., the main trends).
• Nonlinear components of crude oil price time
  series (i.e., error term) are modeled by a neural
  network (NN) model.
• the effects of irregular and infrequent future
  events on crude oil price are explored by
  web-based text mining (WTM) and rule-based expert
  systems (RES) techniques.
• MMI and BBM are the auxiliary modules for
  constructing the integrated TEI_at_I system.

20
The nonlinear integrated forecasting approach

• Within the framework of TEI_at_I methodology, a
  novel nonlinear integrated forecasting approach
  is proposed to improve oil price forecasting
  performance.
• The flow chart of the nonlinear integrated
  forecasting approach is shown in the following.

21
The scheme of the TEI_at_I forecasting approach
22
Outline
B. A New Methodology

• Introduction
• The TEI_at_I methodology for crude oil price
  forecasting
• A simulation study
• Concluding remarks

23
A simulation study

• Data and settings
• The crude oil price data used in this study are
  monthly spot prices of West Texas Intermediate
  (WTI) crude oil, covered the period from January
  1970 to December 2003 with a total of n 408
  observations. For the purpose of this study, the
  first 360 observations are used in-sample data
  (including 72 validation data) as training and
  validating sets, while the reminders are used as
  testing ones.

24
Simulation Results (I)

• The forecasting results of crude oil price (Jan.
  2000 - Dec. 2003)

25
Simulation Results (II)
The comparison of hit ratios between nonlinear
integration approach and simple integration
approach
26
Outline
B. A New Methodology

• Introduction
• The TEI_at_I methodology for crude oil price
  forecasting
• A simulation study
• Concluding remarks

27
Concluding Remarks

• In this study, a new TEI_at_I methodology
  integrating web-based text mining rule-based
  expert system techniques, econometrical
  techniques with intelligent forecasting
  techniques is proposed for crude oil price
  forecasting. Based on the TEI_at_I methodology, a
  novel nonlinear integrated forecasting approach
  is presented.
• The simulation results show that the proposed
  nonlinear integrated forecasting approach with
  error correction and judgmental adjustment
  produces a definite improvement in oil price
  forecasting
• The nonlinear integrated forecasts has shown
  superior to the simple integrated forecasts and
  the individual forecasts.
• The novel nonlinear integrated forecasting model
  can be used as an alternative tool for crude oil
  price forecasting to obtain better forecasting
  accuracy than before.

28
??!???????! http//madis1.iss.ac.cn,
?? MADIS??????? MADIS????? MADIS????????????? MA
DIS??????????? MADIS????????pdf????????? MADIS????
?DSSs?MISs???? ISI?SCI??????SSCI?????\NSFC?????
MADIS?????????Call for Papers