Neural networks usually suffer from model performance problems due to high variance. Ensemble techniques refer to the method of combining multiple models resulting in improving the overall performance of the model. The overall performance can be improved by combining the predictions from multiple models having a good accuracy instead of highly tuning the models, to reduce the variance of predictions and reduce the error due to generalization. In this paper, the different types of ensemble techniques such as stacking ensemble, horizontal voting, and weight average ensemble are being discussed which can be used to improve the model accuracy. The numerous disadvantages of hyperparameter tuning like the wrong choice of parameters, overfitting, and other inefficient optimizing strategies can be overcome using ensemble techniques. The ensemble techniques have numerous advantages such as improving the model performances, reduction of model variance, and so on. Accuracy is improved because results are obtained using the mean predictions of the number of sub-models and the performance is improved with the learning weights of each of the models.