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.

Series predicting M200 and T1000 quantitySeries 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.

Excel chart comparing predictionsExcel 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.