Effectiveness evaluation of regression models for predictive data-mining

In predictive data mining, the choice of technique to use in analysing a data set depends on the understanding of the analyst. In most cases, a lot of time is wasted in trying every single prediction technique in a bid to find the best solution that fits the analyst's needs. Hence, with the advent of improved and modified prediction techniques, there is a need for an analyst to know which tool performs best for a particular type of data set.

This paper studies and proposed a model to evaluate the effectiveness of three data-mining regression techniques for prediction on different and unique data sets. Data-mining techniques, including Multiple Linear Regression MLR, based on the ordinary least-square approach; Principal Component Regression (PCR), an unsupervised technique based on the principal component analysis; Partial Least Squares (PLS), a supervised technique, were applied to each of the data sets. Five criteria used to evaluate the effectiveness of each technique, Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Coefficient of Efficiency (COE), Coefficient of determination (R2) and Number of features or variables used. It also covers the advantages of each technique over the other.