Download presentation
Presentation is loading. Please wait.
Published byFelicia McKinney Modified over 9 years ago
1
STIFF: A Forecasting Framework for Spatio-Temporal Data Zhigang Li, Margaret H. Dunham Department of Computer Science and Engineering Southern Methodist University (Abbreviated Version from PAKDD ’02)
2
May 6, 2002Li & Dunham, PAKDD2 Our goal In this paper, we present a novel forecasting framework for spatio-temporal data, in which not only spatial but also temporal characteristics of the data are considered to obtain a more appropriate result.
3
May 6, 2002Li & Dunham, PAKDD3 Presentation Outline Motivation Our Approach: STIFF Combining two approaches to achieve better results: Time Series Analysis and ANNs Performance Future Work
4
May 6, 2002Li & Dunham, PAKDD4 Why There are many application fields which require spatio-temporal forecasting: river hydrology, biological patterns, housing price research, rainfall distribution, waste monitoring, fishery, hotel pickup rate, etc. In spatio-temporal forecasting, both spatial and temporal properties, as well as their mutual correlation, are taken into account.
5
May 6, 2002Li & Dunham, PAKDD5 Flood Forecasting (Our Motivating Application) Catchment Many different types of sensors Predict at one sensor location Water level or Flow rate May not be interested in actual prediction of value
6
May 6, 2002Li & Dunham, PAKDD6 Our approach : Problem definition Δ={α 0, α 1, α 2, … α n } is the research field, composed of n + 1 spatially separated subcomponents, named by α i accordingly. WLOG, α 0 is assumed the target place where forecasting is about to be carried out. For each α i in Δ, there are j observations with equal time intervals between consecutive ones, denoted by Л i ={α i1, α i2, α i3, … α ij }.
7
May 6, 2002Li & Dunham, PAKDD7 Problem definition (Cont.) - Given Δ={α 0, α 1, α 2, … α n }, Л={Л 1, Л 2, …Л n }, the length of observations j and the look-ahead steps of ι, we are expected to find an as good as possible forecasting relationship ƒ that is defined as follows.
8
May 6, 2002Li & Dunham, PAKDD8 Our approach : Algorithm sketch 1) Describe the forecasting problem according the problem definition. Build a time series (ARIMA) model for each α i. Name the forecasting from α 0 time series model as ƒ T. - Construct and train an ANN to capture the spatial correlation and influence over the target subcomponent α 0. Name the forecasting from the neural network as ƒ S. - Combine ƒ T and ƒ S via a statistical regression mechanism.
9
May 6, 2002Li & Dunham, PAKDD9 Find the spatial influence Normally it is much harder to find than its temporal counterpart in the problem. No precise way to convert from the spatial measurement to the value it may change. Time is only 1 dimension while space is 3 (or 2) dimensions. A simple “distance” measure is not enough, other factors are important.
10
May 6, 2002Li & Dunham, PAKDD10 Artificial Neural Network (ANN) Why is ANN used for finding spatial influence? Itself a “black-box” and non-linear technology used to find the hidden pattern. Like human brain, it can self-adjust and learn automatically even if the problem is not defined very well. Practice proves its usefulness [See,1997] found ANN was especially useful in “… situations where the underlying physical relationships are not fully understood …”
11
May 6, 2002Li & Dunham, PAKDD11 ANN Construction Simple 3-layer back-propagation MLP One input node for each sensor value except α 0. Actual input shifted by predicted time lag. The hidden layer has a certain number of neurons that have to be decided by experiment. The output layer has only one neuron that corresponds to the target subcomponent α 0. We also employ a kind of pruning strategy to achieve the most simplicity of ANN structure without harming the efficacy much.
12
May 6, 2002Li & Dunham, PAKDD12 Integrate the two forecasts We have two forecasts so far at the target subcomponent α 0. One is ƒ T, from the time series model, and the other is ƒ S, from ANN. We may - Either dynamically select one from the two as the current forecast; - Or fuse them together since they contribute to the overall forecasting from two different aspects. (That’s what we take in the paper.) The two forecasts are integrated via a very simple linear regression mechanism. Of course other more advanced alternatives can be used instead for better results.
13
May 6, 2002Li & Dunham, PAKDD13 A case study (National River Flow Archive – Great Britain) Here we are going to present a practical case study to demonstrate how the framework works. We will conduct the spatio-temporal forecasting at the outlet gauging station 28010 regarding the river water flow rate (m 3 /s). The basin is shown as follows. The target station is 28010 while its siblings are lying upstream. Derwent Catchment Daily mean flow values
14
May 6, 2002Li & Dunham, PAKDD14 Data transformation Checking the water flow rate data at station 28010 tells us the data is not very stable. The abrupt change is obvious and present roughly about 25% of the whole time. We therefore employ the data transformation first according to the proposed approach discussed before. We empirically vary the value of λ from –1.0 to 1.0 with the step of.1. It turns out λ = 0.0 is the best (relatively). In other words, we will log-transform the original water flow rate data.
15
May 6, 2002Li & Dunham, PAKDD15 Actual Flow at Derwent
16
May 6, 2002Li & Dunham, PAKDD16 Case Study ANN 6 input nodes 1 output node 6 chosen as number of hidden nodes based on experimentation Number of links pruned based on river topology Lag time used for input based on expected flow lag time
17
May 6, 2002Li & Dunham, PAKDD17 Building models Following the framework specification, we then build a time series model based upon the dataset collected from each gauging station. An ANN is constructed after that, with the spatially- induced pruning strategy applied to erase as many as possible unnecessary links while sacrificing little to the forecasting accuracy. The final overall spatio-temporal forecasting is generated then following this simple regression:
18
May 6, 2002Li & Dunham, PAKDD18 STIFF Model 70 23 43 11 55 48 fSfS fTfT x 1 f T + x 2 f S + C
19
May 6, 2002Li & Dunham, PAKDD19 Performance Analysis Compared STIFF to pure time series (C TS ) and pure ANN (C ANN ) Data starting at 10/01/75 30, 60, 120 days Normalized Absolute Ratio Error (NARE)
20
May 6, 2002Li & Dunham, PAKDD20 Forecasting result The forecasting comparison result, measured in NARE, is outlined in the following table. The other two models, built to our best knowledge, are used to compare with STIFF. Here “Over” means overestimation while “Under” for underestimation.
21
May 6, 2002Li & Dunham, PAKDD21 Result 30 Days
22
May 6, 2002Li & Dunham, PAKDD22 Conclusion STIFF has a better forecast accuracy than the normal single time series model and ANN model, and more balanced (over vs. under estimation). Compared with other related work, it avoids the oversimplification. Does not have the large variation problem. STIFF requires much human intervention and interpretation. STIFF is promising for future research.
23
May 6, 2002Li & Dunham, PAKDD23 Future work Extend to multivariate forecasting Use more sophisticated fusing techniques Test on more flood data Compare to other techniques Examine different ANN structures So far, it can only deal with univariate forecasting. Extend to other application domains …..
24
May 6, 2002Li & Dunham, PAKDD24
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.