The joyful festive atmosphere continues to pervade most Chinese families in Asia in the coming week, as we will be celebrating Lunar New Year (Spring Festival) on Jan 23rd . This date is determined by the lunisolar Chinese calendar. Alongside the 12-year cycle of the animal zodiac, 2012 is Year of the Dragon. In ancient China, the dragon represented an emperor and power, as well as being a creature of myth and legend.
While Western readers may find it hard to believe, many women in the world’s most populous nation are expected to time their pregnancies to occur in the Dragon year, since the Chinese zodiac calendar associates it with wealth and power. Once again it is forecasted that there will be a baby boom this year, since during the previous one in 2000, 5 percent more babies were born, which is typical in a dragon year.
How does this relate to business forecasting? Researchers from financial investment firms claim that demand for infant products such as baby formula, diapers, and clothes in China’s market will grow during this Year of the Dragon. This obviously can give us hints on how we manage our investment portfolio this year. Read more about how The Year of the Dragon affects the economy in BusinessWeek.
However, honestly, I am not a money expert; my investment history makes this much clear! The reason I shared the information above is it reminds me how forecasting impacts our everyday life and that times series data is everywhere. Business leaders may be untrained in statistics but are certainly able to interpret the forecasted results and make decisions based on it.
In creating time series forecasting models, “events” can be added to regression, ARIMAX and unobserved components models. “Events” defined as occurrences out of the ordinary that disturb the underlying time series, need to be taken into account. Some events are planned and expected (for example Lunar New Year holidays, promotions, price change, or even a Dragon Year baby boom) and may be recurring; their impact should be assessed and included in future forecasts if appropriate. Some are unplanned (such as strikes or storms); their impact should be isolated so that it is not perpetuated in future forecasts. For a fuller explanation see more details in a SAS white paper on forecasting.
Another way of gaining insight from time series data is a technique called time series data mining (TSDM), which is the combination of forecasting and traditional data mining techniques that uses time dimensions and predictive analytics to make better business decisions. Enterprise Miner 7.1 includes new time series nodes that automate the processes from data preparation and clustering to similarity analysis:
- Time-stamped data preparation – data transformation, accumulation, aggregation, transformation, and transposing, etc.
- Hierarchical clustering – running hierarchical clustering for transformed data for Cubic Clustering Criterion.
- Similarity analysis – comparing time series that exhibit similar characteristics over time.
- Exponential smoothing – forecasting time series and detecting outliers.
For example, with TSDM analysts from mobile operators could use time-stamped information (such as weekly text messages and voice data consumed as inputs), accumulated and transformed for creating customer segmentation (e.g. prepaid subscriber segments) and predictive models (e.g. customer churn, data plan upselling response). TSDM could also help analysts from retail stores quickly reduce the dimensionality of the problem under investigation and extract signals from the noise. SKUs or stores with similar sales trends (from historical time-stamped information) can be combined into segments and business strategy applied to managing these segments. See more details here for how to use TSDM in Enterprise Miner 7.1.
During this Lunar New Year holiday, I will use the TSDM nodes in Enterprise Miner to check if the stock price of a Shanghai-based baby food company exhibits similar characteristics in dragon years. I can share with you the results if you are interested. But I make no claims for predicting the impact on your stock portfolio!