Yousef Younes^1,, Jun Sang¹, Ahmad Abdullah², Ali Baddour¹

Regression is the data mining process related to estimating a value for a given input by modeling the relationship between the predicators and the response. Choosing the most suitable regression algorithm is the center of big discussion in which the dataset always having the final decision. But when we studied different numerical datasets, we noticed that, data repetition over different intervals is a common property that could be found between any pair of attribute and class values. To exploit this property, this paper begins the journey of finding a new regression approach to address all numeric datasets. The new method uses the minimum bounding geometry to bound the data points in shapes which are used later to suggest values. From the suggested values we choose our targeted prediction value. When we tried this method on different datasets even with the circle as the bounding shape, the results were not perfect but encouraging enough to further elaborate the method. Besides that, the method showed a possibility to do other data mining tasks.

data mining, regression, minimum bounding geometry, classification, association rule discovery