Sales prominent retailers were examined and a positive

Sales are heavily
influenced by exogenous variables, such as weather, competitor strategies and
sales promotions. This makes forecasts context specific, and therefore
different fore- cast methods yield different accuracy depending on the con-
text. Among the time-series and cross-sectional techniques that Thomassey
mentions, are exponential smoothing, Holt Winther’s model, Box Jenkins 
model,  ARIMA  and SARIMA, and regressions. These have
produced satisfactory results; however, because of reasons described, more
advanced methods have been recently adopted. This includes neural networks
(NN), extreme learning machine algorithms (ELM), Gray relation analysis
integrated with ELM and Fuzzy logic and Fuzzy Inference Systems (FIS). These
improve forecasting accuracy, but never achieved the benchmark of real

There is an
elaborate body of work done on predictive analytics with Big Social Data, and
in the following, we will highlight the most relevant related work for our
paper. One of the key papers developed in the field of business outcomes and
social data, is from Asur and Huberman. In their paper, they use Twitter
activity to predict the revenue of Hollywood movies. Furthermore, they discuss
and analyze how the sentiment of tweets (negative, neutral, positive) affects
the revenue performance after the release of the movies. This study showed that
one could predict more accurately the revenue performance by using social data
than the gold standard of forecasts in a given industry, in this example the
Hollywood Stock Exchange.

We Will Write a Custom Essay Specifically
For You For Only $13.90/page!

order now

Choi and Varian used
search engine data of Google Trends to forecast near-term values of economic
indicators, like sales among others. The query indices are often correlated
with various economic indicators, and therefore they may be helpful for
short-term economic prediction. In their paper, they report that simple
seasonal AR models that include relevant Google Trends data have the ability to
surpass models that do not include this data by 5 percent to 20 percent. In the
management report Using Google Trends to Predict Retail Sales, the benefits of
including search data into retail sales forecasting is explored, and the report
concludes that Google Trends can be a powerful supplement. Search data of
prominent retailers were examined and a positive correlation with both store
and online performance was identified, which can be an indicator for total
brand performance. This is also confirmed by which found that search query
volume can be very predictive of future outcomes.