Prediction of a promoter is one of the many active areas of bioinformatics. The outcome of a promoter detection algorithm is directly or indirectly influenced by the success of identifying the location of a TATA box in a promoter sequence. A profiling technique is very often used to find putative TATA boxes, but discriminating a TATA box from putative TATA boxes is still a challenging problem. In this work, we formulate the problem and provide solutions using both a linear and a non linear classifiers.