1. Build History
We calculated a Neural Network Model which is an iterative fitting optimization that
identifies an optimal point between a hold out sample and the model building database.
The optimal convergence point
2. Model vs. Actual
Time series in broken into a trend and cyclical components. The series below is an
extreme cyclical one where the peaks are generally underestimated.
Missed peaks can be corrected by adding a constant
3. Impacts
This application locates the insignificant factor within our data
Since Season has a greater impact than Time, Time is the insignificant factor
4. Cross Section
This determines how our target will be effected by changing the value of one factor
As the month changes, so does the trans amt.
This graph is cyclical in nature and shows that throughout the summer months the trans
amt peaks
5. 3-D Cross Section
Determines what will happen to our target if two factors were simultaneously changed
It’s a combination of a cyclical and a trend model
This shows that as we change time and season than we can see how the trans amt has
changed
6. Reasoning
By reasoning we are able to deifier all the insignificant factors throughout our data
The unexplained box shows that aprox. 30,000 effects are insignificant
7. Bar Chart
Determines how predictive our model is by showing our error
We can do this by finding the difference of the model chart from the actual chart
9. Anomalies
Views and Locates outliners within our data
That is the data that doesn’t fit patterns within the cycle or trend
In this particular application, the peaks show where error occurred
11. By carefully analyzing the results, we were able to determine that
the trans amount values from hotels with an economy status are the
highest in the summer months. Though we determined that the
trans amount increases throughout the summer months, we also
discovered that there has been a steady decrease in the trans
amount from year 2000 when it peaked- to the present day low
Conclusion