Data mining is the exploration and analysis of large datasets for discovering hidden knowledge and patterns. The various techniques from the field of data mining have been successfully applied to a variety of domains. An important area of data mining and machine learning is feature selection. The goal of feature selection is to find a minimum set of features (attributes) such that the reduced dataset characterizes the data similarly as the original dataset without significantly reducing the accuracy of the classifier. We propose a new feature selection algorithm called Automatic Hybrid Search (AHS) that generates consistent feature subsets and is a hybrid of the filter and the wrapper models. Our experiments have shown that AHS performed well at feature selection with a relatively lower runtime cost, a smaller size of the selected feature subset, and a lower error rate than the more traditional approaches such as exhaustive search, heuristic search, and probabilistic search. The findings suggest that AHS is more sensitive to the number of features than to the number of instances in the dataset.