PYTHON PROGAMMING

Project: Exponential Smoothing

Many time series values give spiked results

This is difficult to develop a forecast on due to the erratic nature. Exponential smoothing can smooth

this out so predictions are more stable:

Where alpha = 0.25

Or perhaps this model:

Where alpha = 0.5

Or perhaps this model:

Where alpha = 0.75

What do you believe will happen if alpha is equal to 1?

Formula for Exponential Modeling:

11

t

t

1

1

:note

1) and 0(between constant smoothing

tperiod for timeforecast

valueseries time

1 tperiod for timeforecast

)1(

YF

F

Y

F

where

FYF

t

ttt

=

=

=

=

+=

−+=

+

+

α

αα

In this project your program will use the exponential smoothing to help you predict a value in the future.

In addition your program should come up with the linear regression equation to predict the same value

as was done in exponential smoothing.

Part 1: Obtain the stock price inform from www.nasdaq.com. In this part, go to www.nasdaq.com,

choose a company to analyze, click the “historical quotes” link on the left side after picking a company.

Pick the stock prices for at least 8 months picking one data point out of each month as close to the first

of the month as possible. The x values will be from 1 to 8 (where 1 indicates the first month looked at)

while the y values will be the stock price. For example:

Would give the data point: x = 1, y = 187.21. For the second data point the list is scrolled to find to find

the first historical quote for the next month:

Would give the data point: x = 2, y = 176.02. And so on. Keep in mind that the values for x (1, 2, …, 8)

DO NOT have to correspond to Jan, Feb, etc. x = 1 merely indicates the first month that you decided to

analyze. From there the months should proceed sequentially. Once the data is gathered then you

should have a list of values such as this example:

Time

period

1 2 3 4 5 6 7 8

Value 31.25 32.35 34.15 33.12 37.25 30.19 42.13 44.17

This table should be presented in an Excel document explaining why the choice was made for the

company chosen (why would a model based on the first of the month possible make sense) and the time

frame.

Part 2: Write a Python program that asks the user for the information from part 1 and performs

exponential smoothing based on it. The perfect program will allow the user to input alpha, display the

graph of the original data and the “smoothed data” and have the user verify if this model is appropriate.

If it is not then it should loop asking for new entries for alpha until the user indicates the model is

appropriate. At this point, it should use the exponential smoothing model to predict time period 9 (x =

9). Read the explanation above closely to understand what exponential smoothing provides for the next

month based on the previous month.

Part 3: In the same Python program, the information from NASDAQ should be used to develop a linear

regression model that is used to predict time period 9. It should show the correlation coefficient to

indicate the strength of the model. No other tool is necessary for this project to test the

appropriateness of using a linear regression model.

Category 20 pts 15 pts 10 pts 5 pts 0 pts

Data Collection

Criteria

- Company Identified

and Choice

Explained

- Each data value

obtained

from first

entry for the

month

chosen

- Data was collect from

sequential

months

- Data was delivered in

an Excel

spreadsheet

All 4 criteria

met

Only 3

criteria was

met

Only 2

criteria was

met

Only 1

criteria was

met

None of the

criteria was

met

Exponential

Smoothing Logic

Logic is

100%

correct

Logic is

75%

correct

Logic is

50% correct

Logic is 25%

correct

Logic does

not follow

the

exponential

smoothing

formula

Visual Display of

Smoothed Data

Python

program

prints the

original data

and the

smoothed

data (not

necessarily

in the same

graph)

Python

program

does not

print the

graphs but

writes the

R program

to a file

which can

then be

opened into

R and run.

Python

program

does not

print the

graphs but

outputs the

appropriate

R lines of

code that can

be Edit-

Copied and

Edit-Pasted

into R for

viewing for

graph with

no re-

formatting

necessary in

R.

Python

program

does not

print the

graphs but

outputs the

data that can

be Edit-

Copied and

Edit-Pasted

into R with

manual

reformatting

of the data.

(ex:

5, 8, 7, 10

Which then

has to be

formatted in

R as x <-

c(5,8,7,10)

No attempt

is made to

display the

graphs.

Regression Logic Logic is

100%

correct

Logic is

75%

correct

Logic is

50% correct

Logic is 25%

correct

Logic does

not follow

the

regression

logic from

the last unit

in the

course.

Structure and

Design

Criteria:

- The appropriate

flow

mechanism

is used in the

program

(while loop,

etc.) for ease

of use of the

user

- Code is placed in a

library for

code re-use

- Code is documented

where

appropriate.

- Code is “readable”

(appropriate

variable

names and

structured

programmin

g techniques

used).

All 4 criteria

met

Only 3

criteria was

met

Only 2

criteria was

met

Only 1

criteria was

met

None of the

criteria was

met

Deliverables

- Excel document with data and explanation of company choice

- Python program

- Python library (if a library is implemented)

- R programs (any appropriate R programs)