Comparing Predictions for Forecasting Models (Intermediate Data Mining Tutorial)
You have created the following three models:
Predictions for each combination of region and model, based only on data for the individual model and region.
Predictions for all models on a worldwide basis, based on aggregated data.
Predictions for the M200 model in the North America region, based on the aggregated model.
In this final task, you will contrast the predictions for each model to see how using the generalized model affects the results.
Comparing Prediction Results
Remember that the original mining model showed a large gap between certain regions and model lines. The trend line for the M200 model was particularly high, while the trend lines for the T1000 model were low and relatively flat.
.gif "Series predicting M200 and T1000 quantity")
You can create a chart that includes all the predictions by exporting the results and the original data to Microsoft Excel, which provides more sophisticated tools for graphing and managing multiple data series. The following diagram shows the trend lines for just the M200 product models, and compares the predictions from the first mining model against the predictions using the aggregated mining model.
.gif "Excel chart comparing predictions")
From this chart, you can see that the aggregated mining model smoothens the fluctuations in the individual data series. The following table provides a portion of the data series used to create the chart, to aid in comparison.
Series and Mining Model |
7/25/2004 |
8/25/2004 |
9/25/2004 |
10/25/2004 |
11/25/2004 |
|---|---|---|---|---|---|
M200 Europe — aggregated |
143 |
126 |
115 |
119 |
94 |
M200 Europe—specific |
121 |
142 |
152 |
149 |
154 |
M200 North America — aggregated |
208 |
150 |
149 |
151 |
172 |
M200 North America—specific |
163 |
178 |
156 |
173 |
203 |
M200 Pacific — aggregated |
89 |
80 |
71 |
77 |
57 |
M200 Pacific—specific |
46 |
44 |
42 |
42 |
38 |
T1000 Europe — aggregated |
65 |
51 |
54 |
53 |
48 |
T1000 Europe—specific |
42 |
41 |
43 |
42 |
43 |
T1000 North America — aggregated |
103 |
84 |
79 |
85 |
68 |
T1000 North America—specific |
82 |
78 |
78 |
83 |
83 |
T1000 Pacific — aggregated |
68 |
52 |
48 |
56 |
44 |
T1000 Pacific—specific |
38 |
39 |
37 |
38 |
36 |
Conclusion
You have learned how to create a time series model that can be used for prediction, and a generalized model that can be applied to a different data series.