The graphic approach to data classification involves using break points. When all observed values are sorted in ascending order and then plotted in a graph.

The graph is used to look for discontinuities where class limits could be placed. The discontinuities, called break points, can function as class boundaries since they are the natural breaks in the observation series. It is obvious that not all data sets will show break points, and if they do, the chances that there are enough number of classes planned is quite uncertain. You can select the break points by interpretation of:

Frequency diagram: As shown in the figure 1. The frequency diagram can be used when searching for 'breaks'. If found, they can function as class boundaries. A frequency diagram is especially useful when a large number of observations are involved. The chance to find 'breaks in the data' increases when the value along the horizontal axis are grouped together. The existence of break points is strongly influenced by the size of the intervals applied to group the data.

Cumulative frequency diagram: The frequency of occurrence of the observed values is added up. Changes in orientation of the curve indicate break points. The values along the horizontal axis of those break points are the class boundaries. Again, as with the graphic array method, there is a limited chance that the number of class boundaries needed will be found using a graphic approach.

Fig 1. class boundaries selected by looking for gradient changes in the distribution