Neural  network (NN) tools are suitable for many tasks such  as classification,  clustering, scheduling and prediction. NN performance depends on the strategy of learning a phenomenon, the number  of hidden nodes, activation function and,  of course,  the behavior  of  the data. There are many techniques used for training NN, while the  social insect’s techniques become more  focused  by researchers because of its  natural  behavioral  processing.  The Artificial Bee Colony (ABC) algorithm has produced  an  easy way for solving combinatorial, statistical problems and for training NN by different organized agents.  The objective of training NN  is to adjust the weights so that application of a set of inputs produces the desired set of outputs. Normally, NN is  trained by a  standard back-propagation (BP)  algorithm;  however,  BP  is too  slow  for many applications and trapping in a local minima problem. To recover the above gap, the hybrid technique was used for training NN here. The hybrid of natural  behavior agent ant and bee  techniques  was  used for training  NN. The simulation result of a  Hybrid Ant Bee Colony (HABC) was compared with, ABC, BP Levenberg-Mardquart (LM) and BP Gradient Descent (GD)  learning  algorithms.  According to experimental results,  the proposed HABC algorithm did  improve the classification accuracy  for  the  Boolean function, and prediction of volcano time-series data, which was used to train the MLP.