SAS Enterprise Miner

Company has a long list of donors and is involved in the process of sending out mailers to all its donors who have contributed to the Company at one time or another. However, in the interest of making its mailings more fruitful in terms of the returns they are likely to generate, the Company proposes to prune its list down in order to derive a data subset that will take into consideration only those donors in whose case the probability of a donation is higher. Since the list of donors presently is about 5000, the Company is interested in conducting a statistical data analysis and data mining in order to prune down this number. This report will therefore conduct three different forms of statistical data mining to arrive at the information that is sought: (a) decision tree (b) linear regression (c) neural network.
The report generated on a preliminary data analysis on the basis of a paired samples test may be viewed in Appendix A. This analysis uses the variables of the average of all gifts received in the last thirty six months and the number of life time gifts to date reveals a is a positive correlation between the two variables that have been selected in this case – the average gifts received in the past three years and the gifts received during the lifetime of the donors. This would appear to indicate that the most likely donors could be among those who have been making gifts in the past three years, since this is related to a lifetime pattern of giving. This can be correlated with the socio economic variables in order to provide a more comprehensive broad based report of other positive correlations.
In order to analyze this data, the estimated ratio of responses to promotions is a good independent variable, against which the correlation of dependant socio economic variables may be considered. A Pearson correlation matrix helps to establish the correlation between these various socio economic groups and their average gift giving, and this is set out in