Applying feature selection methods to improve the predictive model of a direct marketing problem

The ability to forecast job advertisement demands is vital to enhance the customer retention rate for recruitment companies. On top of that, it is uneconomical to cold call every individual on a regular basis for companies with a large pool of customers. This paper presents a novel approach in predicting the re-ordering demand of a potential group of SMEs customers in a large online recruitment company. Two feature selection techniques, namely Correlation-based Feature Selection (CFS) and Subset Consistency (SC) Feature Selection, were applied to predictive models in this study. The predictive models were compared with other similar models in the absence of feature selections. Results of various experiments show that those models using feature selections generally outperform those without feature selections. The results support the authors' hypothesis that the predictive model can perform better and further ahead than similar methods that exclude feature selection.