Abstract: Web proxy caching plays a key role in improving the world wide web performance. However, the difficulty in determining which web objects will be re-visited in the future is still a big problem faced by existing web proxy caching techniques. In this study, we present a new approach which depends on the capability of support vector machine to learn from web proxy log data and predict the classes of objects to be re-visited. Therefore, usage of the cache can be optimized efficiently. Experimental results have revealed that the support vector machine produces similar correct classification rate compared to neuro-fuzzy system. However, the support vector machine achieves much better true positive rate and performs much faster than neuro-fuzzy system for both training and testing in several datasets.