Artificial bee colony algorithm is a new population-based evolutionary method based on the intelligent behavior of honey bee swarm. It has shown more effective than other biological-inspired algorithms. However, there are still insufficiencies in ABC algorithm, which is good at exploration but poor at exploitation and its convergence speed is also an issue in some cases. For these insufficiencies, we propose a novel artificial bee colony algorithm (NABC) for numerical optimization problems in this paper to improve the exploitation capability by incorporating the current best solution into the search procedure. Experiments are conducted on a set of unimodal/multimodal benchmark functions. The experiments results of NABC have been compared with Gbest-guided artificial bee colony algorithm (G-ABC), improved artificial bee colony algorithm (I-ABC), Elitist artificial bee colony algorithm (E-ABC). The results show that NABC is superior to those algorithms in most of the tested functions.