Time Series Forecasting Introduction: The Walt Disney Company is known to be the worlds most admired entertainment company. It has recently decided to open up a new Paxar themed park in California. In order to do so, the company will need to assure their bank that It Is capable of paying back loans In the future as well as reassuring owners and Investors that they will not lose any money In the future.
In order for Walt Disney to carry on with their plan, they need to be able to show their banks, owners and investors a model to predict future values based on historical values. How lucky for them that a group of highly trained time series forecasters are available for a top-dollar price! The group of analysts will decide on a few methods to enter in their data and then determine which technique works best with the corresponding data. They will base their decision by determining which method has the least amount of error as well as the most dependability.
With a company this large and a lot at stake, it is crucial for the results to be as efficient as possible so that the proper decisions can be made to follow. The ingenious analysts will use historical data from the past eight years (31 raters) to determine the revenue of the thirty-second quarter. The forecasting will help banks determine whether It is a good Idea to support Walt Disney with a loan. In Dalton, forecasting for the thirty-second quarter will give Important Information can prepare and make plans.
Data: The company’s historical data involving revenue was collected from the past eight years, a total of thirty-one quarters, from the years 2005 to 2012. Our dependent variable (the variable being predicted) is revenue and our independent variable (used to assess the value of the dependent variable) is time. Revenue was measured in millions and time was measured in quarter periods. Thirty-one observations were included in the dataset in order to receive effective forecasting results.
Our source of data was collected from the Walt Disney quarterly earnings report, which is accessible through the following link: http://telecommunications. Com/investors/ financial-information/earnings. Preliminary Analysis When observing the scatter plot above, we look for whether our data is stationary, meaning that it has no trend or seasonality, whether the data shows a consistent rend, or whether the data shows both a trend and seasonality. Using Disney’s historical data from the past eight years we determined that there is a slight upward trend from looking at the scatter plots.
When forecasting for a positive or negative trend with no seasonal patterns, the Regression method is the technique that works best. In our data set, the variable being predicted is revenue (dependent variable) and time (independent variable) is used to estimate the value of revenue. Plotting these variables allows for us to find the slope of the regression line to determine whether there is a trend. If the slope is zero then there is no trend. If the slope is positive or negative then a trend exists.
A line drawn through our data permits us to calculate the slope and create an equation for the line. The slope of our regression line is 79. 218; meaning $79 million in revenue will change for each one quarter time period. Basically, revenue will increase with each quarter time period. Our data will continue to follow the same trend. Essentially we used actual values of revenue from the past eight years to come up with a regression line to predict future values of quarter time periods.
Forecasting: Forecasting using Naive Forecasting using Exponential Smoothing Forecasting using Classical Decomposition Evaluation: We can evaluate the different errors of each method by using Bias (mean of errors), MAD (mean absolute deviation), MAPLE (mean absolute percentage error), MUSE (mean square error) and SE (standard error) to determine which forecasting method is the best one. Bias can determine whether we have over or under forecasted the outcomes of our Y variables for each method. If there is a positive bias then are under forecasting (meaning we predicted too low of a future revenue).
If there is a negative bias then we are over forecasting (meaning we predicted too high of a future revenue). Bias in a regression method is always zero because half of the errors are positive and the other half are negative logically equaling them out to zero. In the Naive method, our bias is 80. 73 meaning we forecasted too low of a future revenue. In the Exponential Smoothing method, our bias is 473 meaning we forecasted too low of a future revenue. Naturally our bias for regression came out to be zero. MAD gives us the true degree of error in within each method.