Download presentation
Presentation is loading. Please wait.
Published byAlison Eaton Modified over 9 years ago
1
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 1 FORECASTING USING NON-LINEAR TECHNIQUES IN TIME SERIES ANALYSIS AN OVERVIEW OF RELATED TECHNIQUES AND MAIN ISSUES Michel Camilleri
2
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 2 EARLY FORECASTING Maltese Stone Age Hunter–Gatherer used Mnajdra to forecast seasons (among other purposes)
3
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 3 FORECASTING TODAY Non linear time series techniques are being used to to forecasting sun spot activity (among other uses) Data/Images acquired at EOS/OCS, CIF-US, Universidad de Sonora, Mexico. Observer(s): M.C. Marianna Lyubarets
4
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 4 APPLICATION AREAS MEDICAL MILITARY MANAGEMENT FINANCE ASTRONOMY DEMOGRAPY …………
5
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 5 TIME SERIES TECHNIQUES LINEARVS NON LINEAR
6
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 6 LINEAR TECHNIQUES Linear methods try to model closely underlying subsystems Require identification & measurement of several system features - seasons, trends, cycles, outliers
7
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 7 NON LINEAR TECHNIQUES Non-linear techniques exploit measurement data and computer power: Mimic dynamic system without having to understand exactly the underlying processes Better results than Linear in certain areas
8
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 8 BASIC STEPS TO FORECASTING COLLECT DATA EXAMINE DATA PREPROCESS DATA OPTIMIZE PARAMETERS APPLY PREDICTION TECHNIQUES MEASURE PREDICTION ERROR REVIEW AND UPDATE
9
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 9 A PRACTICAL EXAMPLE CREATE OWN DATA SET (3000 pts) WITH RANDOM NOISE SEPARATE TRAINING SET, ATTRACTOR, FUTURE (HIDDEN SET) EXAMINE DATA PREPARE DATA PREDICT MEASURE SUCCESS OF PREDICTION OPTIMISE PARAMETERS
10
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 10 DATA CREATION Created function X(t) vs t Where t = DISCRETE VALUES OF TIME (1..3000) And X(t) = A1 * SINE (t * F1) + A2 * COS (t * F2) * Random () * N Amplitude A1 = 0.1, Frequency F1 = 5 Amplitude A2 = 0.2, Frequency F2 = 0.33 Noise factor N = 3
11
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 11 UNDERLYING SUBSYSTEMS SINE FUNCTION + COSINE FUNCTION
12
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 12 MEASUREABLE SIGNAL SUBSYSTEMS + NOISE
13
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 13 SEPARATE THE DATA
14
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 14 FUTURE SET (HIDDEN FROM SYSTEM)
15
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 15 EXAMINE DATA VISUAL INSPECTION STATIONARITY PHASE SPACE MAPPING AUTOCORRELATION LYAPUNOV EXPONENT DELAY SPACE EMBEDDING MINIMAL EMBEDDING DIMENSION
16
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 16 PHASE STATE
17
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 17 PHASE SPACE MAP
18
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 18 AUTO CORRELATION SUM
19
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 19 MAX LYAPUNOV EXPONENT
20
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 20 TIME DELAY EMBEDDING THE ATTRACTOR DIMENSIONS = 100 TIME DELAY = 1PREDICTOR POINT
21
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 21 PREPROCESSING DATA FILTERING NOISE REDUCTION TEMPORAL ABSTRACTIONS CATEGORIZE ETHERNET PACKETS BY SIZE CATEGORIZE ECG SIGNALS BY TYPE
22
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 22 NON LINEAR NOISE REDUCTION Noise reduced by 8 %
23
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 23 APPLY PREDICTION TECHNIQUE Set initial parameters –Time delay, dimensions, distance, box size, number of future steps ahead Choose measure of success and apply it to output (Various) Find optimal set of parameters
24
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 24 COMPARE ATTRACTOR ALONG TRAINING SET
25
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 25 FINDING A NEIGHBOUR
26
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 26 FIND ALL NEIGHBOURS OF SELECTED POINT Neigbour Time point Neighbor 1252391769 22711101770 3447111956 41013121958 51768132145 61203142334 71392152335 81581 ID 9, M=9 Err= 2 NEIGHBOURS FOUND 15
27
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 27 FIND PREDICTED SET FOR NEIGHBOUR
28
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 28 AVERAGE of PREDICTION SETS PREDICTION SETS OF ALL NEIGHBOURS FINAL PREDICTION
29
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 29 FIRST PREDICTION ATTEMPT
30
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 30 NEED TO VARY PARAMETERS I
31
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 31 EXAMINE MORE CLOSELY I
32
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 32 A BETTER ATTRACTOR Time Delay = 9, Dimensions = 9
33
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 33 A BETTER PREDICTION Delay=9,Dim=9,Err=2 neighb=15,rms = 1.09
34
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 34 CHANGE DELAY, DIMENSIONS Delay=1,Dim=20,Err=2 neighb=1,rms = 1.37
35
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 35 CHANGE DISTANCE Delay=1,Dim=20,Err=3 neighb=34,rms = 1.09
36
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 36 PROCESSING CONSIDERATIONS Multiple attempts at prediction,calculation of invariants, noise reduction, require increasing orders of operations Each operation may require comparison of every point on attractor with respective points for each training point. Number of operations to find neighbours can be reduced by comparing attractor only to points in same phase state e.g. Box or Tree assisted neighbour search in Phase space.
37
Forecasting using Non Linear Techniques in Time Series Analysis – Michel Camilleri – September 2004 37 THE END (AS FORECAST)
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.