home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Monster Media 1994 #1
/
monster.zip
/
monster
/
MATH
/
VISSIM.ZIP
/
STOCKMKT.VSM
< prev
next >
Wrap
Text File
|
1993-11-24
|
6KB
|
246 lines
; VisSim Block Diagram Format (VBDF)
; Copyright (C) 1989-1993 Visual Solutions
PV=1.200
PS=0
PE=290
PP=1
PI=170
PX=1
PN=1e-006
Pn=-9,6,14,"MS Sans Serif"
Pc=52
PD=800x600
Pf=0x0
Pr="stub"
Ps=784,0,0,546,0,0
Pd=17
PM=1,1,1,1
N.1="/"@144x56<M>
N.2="const"(2000)@84x70<M>
N.3="/"@150x91<M>
N.4="const"(700000)@84x105<M>
N.5="/"@150x133<M>
N.6="const"(200)@102x147<M>
N.7="/"@144x273<M>
N.8="const"(300)@96x287<M>
N.9="/"@144x315<M>
N.10="const"(600)@96x329<M>
N.11="/"@144x350<M>
N.12="const"(600)@96x364<M>
N.13="/"@150x392<M>
N.14="const"(600)@102x406<M>
N.15="Compound"@114x357#8,8<C>
n="Scaling"
Ms=1560,0,0,1065,0,0
N.16="neuralNet"@240x343#9,1
n="stockmkt.net"
UP=8,1,0,1,14,0.7,0.7,0.5,0,0,1,0,1
N.17="/"@150x14<M>
N.18="const"(4000)@84x28<M>
N.19="const"(4000)@294x119<M>
N.20="*"@372x105<M>
N.21="Compound"@318x245#3,2<C>
n="Sc"
Ms=558,0,0,434,0,0
N.22="comment"@366x392@330x140
C=" BACKGROUND INFORMATION
STOCKMKT.VSM demonstrates the use of
VisSim/Neural-Net in forecasting the future value
of the U.S. Dow Jones Industrial Average (DJIA).
The neural-net is trained on data for 200 working
days since 4 Jan 1982. The neural-net file
STOCKMKT.NET contains the trained weights for
the Backpropagation with Momentum algorithm.
The test data are the 90 days following the training data."
N.23="date"@552x42
N.24="wireLabel"@492x14
n="VisSim/Neural-Net learning the US DJIA"
N.25="comment"@0x14@336x161
C="
TO SEE HOW WELL THE NEURAL NET LEARNED THE
market dynamics, just select GO from the SIMULATE MENU,
and then MAXIMIZE the plot.
TO TRAIN THE NEURAL NETWORK on your own,
see instructions below in this comment block.
The neural net has been trained on 200 days of trading
activity since 4 Jan 1982. From that training, the neural net
shows great ability to predict market activity for the next
90 days, which is roughly 4 months of trading!
If you want to train the neural network, then follow these
instructions:
1. Under Change Parameters in the Simulate
Menu, select Auto Restart.
Also, change Range End to 200.
2. Click the \"button\" block \"ON\" with the right
mouse button.
3. Click right mouse on the STOCKMKT.NET block,
and get the dialogue box. Select LEARN
and SAVE at SIM END. Deselect READ at
Sim Start.
To avoid overwriting the neural network file
that has been pretrained, we recommend that
you specify a new network filename, such
as STOCK.NET.
4. Click right mouse on the Plot block. Change
the Y upper value to .0001, and the Y lower
value to 0.0000
Now you're all set for the neural network to learn.
So go to the simulate menu and select GO!"
N.26="wireLabel"@564x245<M>
n="Scaling inputs into range [0,1]"
N.27="plot"@306x14@234x154<M>
px="Time (sec)"
pax=0
pc=512
pm=10
N.28="plot"@294x273@234x154<M>
px="Time (sec)"
pax=0
pc=512
pm=10
N.29="unitDelay"(0)@420x259<M>
N.30="const"(1)@366x245<M>
N.31="Compound"@192x357#1,1<C>
n="td"
Ms=420,0,0,259,0,0
N.32="wirePositioner"@120x322
N.33="wireLabel"@540x266<M>
n="Scaling the output into the DJIA range"
N.34="wireLabel"@342x182<M>
n="Time delay that presents previous"
N.35="wireLabel"@342x196<M>
n="day's DJIA value to neural net"
N.36="button"(0)@180x245
N.37="merge"@498x77<M>
N.38="wireLabel"@90x210
n="To teach the network, click this ON"
N.39="wireLabel"@90x224
n="and follow the directions listed above"
N.40="merge"@276x413<M>
N.41="const"(0)@222x427<M>
N.42="wireLabel"@540x259<M>
n="which is required by VIsSim/Neural-Net"
N.43="import"@24x357#0,8
n="stockmkt.dat"
di=1
N.44="plot"@384x91@372x280
pt="VisSim/Neural-Net Predicting DJIA"
px="Time (days)"
py="Dow Jones Ind Avg"
pax=0
pf=H
pf=F
pb=1200,700
pbx=290,0
pbY=0,0
pbX=0,0
pc=4096
pm=10
pb.0=2000,0
pL.0="Neural-Net Forecast"
pb.1=2000,0
pL.1="Actual DJIA"
pb.2=0,0
pb.3=0,0
N.45="wirePositioner"@366x182<M>
N.46="wirePositioner"@252x224<M>
N.47="wirePositioner"@342x196<M>
N.48="wirePositioner"@294x210<M>
N.49="wirePositioner"@270x469<M>
N.50="wirePositioner"@246x490<M>
N.51="wirePositioner"@330x434<M>
N.52="wirePositioner"@306x455<M>
I.1.i1=15.i2
I.1.i2=2.o1
I.3.i1=15.i3
I.3.i2=4.o1
I.5.i1=15.i4
I.5.i2=6.o1
I.7.i1=15.i5
I.7.i2=8.o1
I.9.i1=15.i6
I.9.i2=10.o1
I.11.i1=15.i7
I.11.i2=12.o1
I.13.i1=15.i8
I.13.i2=14.o1
G.15=1,2,3,4,5,6,7,8,9,10,11,12,13,14,17,18,26,27,28,42,45,46,47,48,49,50,51,52,
I.15.o1=45.o1
I.15.o2=47.o1
I.15.o3=48.o1
I.15.o4=46.o1
I.15.o5=51.o1
I.15.o6=52.o1
I.15.o7=49.o1
I.15.o8=50.o1
I.15.i1=43.o1
I.15.i2=43.o2
I.15.i3=43.o3
I.15.i4=43.o4
I.15.i5=43.o5
I.15.i6=43.o6
I.15.i7=43.o7
I.15.i8=43.o8
I.16.i1=15.o1
I.16.i2=31.o1
I.16.i3=15.o2
I.16.i4=15.o3
I.16.i5=15.o4
I.16.i6=15.o5
I.16.i7=15.o6
I.16.i8=15.o7
I.16.i9=15.o8
I.17.i1=15.i1
I.17.i2=18.o1
I.20.i1=21.i2
I.20.i2=19.o1
G.21=19,20,33,37,40,41,
I.21.o1=37.o1
I.21.o2=40.o1
I.21.i1=36.o1
I.21.i2=16.o1
I.21.i3=32.o1
I.27.i1=17.o1
I.27.i2=1.o1
I.27.i3=3.o1
I.27.i4=5.o1
I.28.i1=7.o1
I.28.i2=9.o1
I.28.i3=11.o1
I.28.i4=13.o1
I.29.i1=30.o1
I.29.i2=31.i1
G.31=29,30,34,35,
I.31.o1=29.o1
I.31.i1=15.o1
I.32.i1=43.o1
I.37.i1=21.i1
I.37.i2=21.i2
I.37.i3=20.o1
I.40.i1=21.i1
I.40.i2=41.o1
I.40.i3=21.i3
I.44.i1=21.o1
I.44.i2=21.o2
I.45.i1=17.o1
I.46.i1=5.o1
I.47.i1=1.o1
I.48.i1=3.o1
I.49.i1=11.o1
I.50.i1=13.o1
I.51.i1=7.o1
I.52.i1=9.o1