Black box modeling of PIDs implemented in PLCs without structural information: a support vector regression approach
Response modeling, which predicts whether each customer will respond or how much each customer will spend based on the database of customers, becomes a key factor of direct marketing. In previous researches, several classification approaches, include Support Vector Machines (SVM) and Neural Networks (NN), have been applied for response modeling. However, there are two drawbacks of conventional approaches: (1) response models only predict classification scores rather than predicting total amount of money spent, (2) too large training data. For the first drawback, we applied Support Vector Regression (SVR) for response modeling to predict total amount of money spent of each respondent. For the second drawback, we employed a pattern selection method designed for SVR. This paper provides experimental results of a direct marketing dataset in terms of model fit, training time complexity and profitability.