Scribd
Upload
54 views

Forecasting - CocaColaSales

This document summarizes the results of time series forecasting and modeling of Coca Cola sales data. An exponential trend seasonal model was initially used to forecast sales. Residuals from this model showed no significant autocorrelation. An AR(2) model was then applied to the residuals, eliminating remaining autocorrelation. This improved the initial forecasts by adjusting for error corrections from the AR(2) model, providing a more accurate final forecast for Q3 sales.

Uploaded by

Ravinderpal S Wasu
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
54 views

Forecasting - CocaColaSales

This document summarizes the results of time series forecasting and modeling of Coca Cola sales data. An exponential trend seasonal model was initially used to forecast sales. Residuals from this model showed no significant autocorrelation. An AR(2) model was then applied to the residuals, eliminating remaining autocorrelation. This improved the initial forecasts by adjusting for error corrections from the AR(2) model, providing a more accurate final forecast for Q3 sales.

Uploaded by

Ravinderpal S Wasu
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 6

INDIAN SCHOOL OF BUSINESS

FORECASTING ANALYTICS 1
CBA - B8
Individual Project
Forecasting Analytics 1 : ICE No 3 – June 14, 2017

Ravinderpal S Wasu – 71710004

8) Multiplicative model for Coca Cola sales. (MODEL 4) Consider the residuals.
Use XLMiner: Time Series > and compute/plot the ACF on the Residuals. Comment on the lags (are they significant)?
Solution :
Excel  Coca Cola Sales Exponential Trend Seasonal Model.xlsx
From XL Miner created the Data Partition on the Time series and ran the Prediction as below:
Data partition Prediction setup

From sheet : MLR_TrainingScore  Select the Residual data  XL Miner  ARIMA  Autocorrelations
For 4 lags For 12 lags

Plot for 4 lags

Plot for 12 lags

Here we see that none of the lags are touching / crossing the UCI or LCI bounds. Hence, they are not significant.
9) Run a AR(2) model using the ARIMA (p=2, d=0, q=0) procedure on the residuals R1, R2,….R38 from MODEL 4

a. Is any autocorrelation left? Answer by computing the autocorrelation of error-of-errors. (Run ACF on error of
errors).

Solution:

Create new ARIMA Input sheet with MLR_ValidationScore residuals and t.


Enter the t column and the residual columns as data

Run the ARIMA 2 model for getting the error residuals and forecast errors
ARIMA Model

ARIMA Coeff StErr p-value


Const. term 7.65772E-18 0.004257 1
AR1 0.767112601 0.146852 1.75E-07
AR2 -0.09474209 0.154777 0.540458

Forecast

Quarter t Forecast
Q3_95 Forecast 1 0.045363259
Q4_95 Forecast 2 0.028478568
Q1_96 Forecast 3 0.017548458
Q2_96 Forecast 4 0.010763524

Now going to Arima_Residuals sheet and run ACF again to see autocorrelation in the error of errors
For 4 lags For 12 lags

ACF plot for 4 lags

ACF Values

Lags ACF
0 1
1 0.012059 ACF Plot for Residuals
2 0.019808 0 2 4 6
3 -0.06917 1 0.4
4 -0.11583 ACF
0.5 0.2
ACF

0 0 UCI
-0.5 0 1 2 3 4 -0.2
LCI
-1 -0.4
Lags

ACF plot for 12 lags

ACF Values

Lags ACF
0 1
1 0.012059 ACF Plot for Residuals
2 0.019808
0 5 10 15
3 -0.06917
1 0.4
4 -0.11583 ACF
0.5 0.2
5 0.120418
ACF

0 0 UCI
6 0.056933
-0.5 0 1 2 3 4 5 6 7 8 9 10 11 12 -0.2 LCI
7 0.025883
8 -0.10121 -1 -0.4
Lags
9 0.048069
10 0.01703
11 -0.08724
12 -0.20016

Here we see there are no significant autocorrelation ship left


10) Please get the improved forecast for Q3_95 combining information from MODEL 4 and the information obtained
from the AR(2) model. (Improved forecast based on Layer 2 model)

Solution:
From Arima_Output sheet we see the forecast error corrections as below:

Forecast

Quarter t Forecast
Q3_95 Forecast 1 0.045363
Q4_95 Forecast 2 0.028479
Q1_96 Forecast 3 0.017548
Q2_96 Forecast 4 0.010764

From the MLR_ValidationScore


Calculating the new adjusted values for the forecast using the error corrections
Predicted Actual 95% Confidence Intervals 95% Prediction Intervals
Quarter Residual
Value Value
Lower Upper Lower Upper
Q3_95 8.423997 8.49597 0.071972 8.346027 8.501967758 8.232542 8.61545247
Q4_95 8.348577 8.374015 0.025438 8.270607 8.426547565 8.157122 8.54003228
Q1_96 8.289954 8.34141 0.051456 8.210564 8.369344876 8.097916 8.48199207
Q2_96 8.532015 8.566555 0.03454 8.452624 8.611405271 8.339977 8.72405246

Predicted Values adjustment


Predicted AR2 Adjusted value = 95% Confidence Intervals 95% Prediction Intervals
Predicted
Quarter Value forecasted EXP(Predited
Value Lower Upper Lower Upper
(EXP) Error Value + Error)
Q3_95 8.42399744 4555.076 0.045363 4766.467284 4408.944 5152.982838 3935.942563 5772.24134
Q4_95 8.34857724 4224.167 0.028479 4346.194062 4020.194 4698.629421 3588.899111 5263.28611
Q1_96 8.28995422 3983.652 0.017548 4054.175731 3744.757 4389.160837 3345.812435 4912.51114
Q2_96 8.53201461 5074.659 0.010764 5129.575232 4738.081 5553.417566 4233.313437 6215.59033

You might also like