Scribd
Upload
55 views

Forecasting Final Exam

This document discusses time series analysis and forecasting of sales data. [1] The Dow Jones Index data appears stationary with no clear trends or seasonality. [2] Box-Cox transformation slightly smoothed the trend but did not significantly change the analysis. [3] The sales data shows a long-term increasing trend and strong yearly seasonality with December peaks and January troughs that increase over time. [4] Decomposition of the data clearly shows the trend, seasonality, and remainder components, including the impact of a recession. [5-6] Residuals from ARIMA and STL models differ in distribution. [7] The chosen ARIMA model produced the most accurate forecast based on lowest

Uploaded by

Pedro Pablo Campuzano
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
55 views

Forecasting Final Exam

This document discusses time series analysis and forecasting of sales data. [1] The Dow Jones Index data appears stationary with no clear trends or seasonality. [2] Box-Cox transformation slightly smoothed the trend but did not significantly change the analysis. [3] The sales data shows a long-term increasing trend and strong yearly seasonality with December peaks and January troughs that increase over time. [4] Decomposition of the data clearly shows the trend, seasonality, and remainder components, including the impact of a recession. [5-6] Residuals from ARIMA and STL models differ in distribution. [7] The chosen ARIMA model produced the most accurate forecast based on lowest

Uploaded by

Pedro Pablo Campuzano
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/ 4

Forecasting Final Exam

1.

In the first autoplot of the Dow Jones Index we can see that the data seems to be stationary, with no
trend and no clear cycles or seasonalities. When we check for the residuals, the first plot ilooks like
white noise.

2.

This plot is before using BoxCox transformation.

This plot is after using BoxCox transformation, but we can see that not much has changed, BoxCox
helps for smoothing the trend and make it clearer, but in this case the cycles and trends are very
clear. After the BoxCox transformation we can see a bit more smooth trend and bigger cycles.
3.
a.
After reading the excel we cant see the data, because it is not a time series.

b.
After running a code to make it a time series, we can now plot and see the trends and seasonality.

In this graphs we can see clear trends


and seasonality. The sales have a long
increasing trend along the years and a
yearly seasonality, having a huge
increase every December and a big
decrease for January.

In these two graphs we can see even more clearly the seasonality and how it is increasing over the
years. We can also see how every end of the year there is a huge increase that gets bigger over time,
having the last two years as the biggest increase.
h.

Here we have the two residuals from the train and the test, where we can see that they don’t seem
to be related. The test one is very small, with very small amount of data compared to the train one.

4.
a.
In this decomposition we can clearly see the 4 parameters that we always see in a decomposition;
original data, the trend, the seasonality and the remainder. If we compare the data with the trend,
we can see that there is this upgoing trend, but almost in the end there is a flat area that grows
again by the end. Also the seasonality is very important, we can see here how along the whole trend,
there is this pattern of seasonality repeating.

b.
Yes, the recession is clearly visible in the decomposition: in the original data there is a small
decrease, but when we check at the reminder we can totally see how it affected the whole data, just
to be “normal” again in the last years.

5.

Here we have the two residuals, where we can see that they have some differences. In the STL
method is seems to be much more white noise anda wider distribution of the data.

6. f
7.

The resulted forecast of this Arima model seems to be appropriate, after trying with other models,
this Arima gave the best RMSE (lower value).

You might also like