Software Maintenance is of utmost importance for any industry. So, to predict the value of software maintenance beforehand also becomes very important, hence many software maintenance prediction algorithms have been devised previously. Here we are also trying to calculate the software maintenance prediction values by taking current and previous datasets of the same software and comparing the lines of code in both of them. In this paper, performance of various machine learning algorithms and ensemble learning using deep neural networks has been compared. The machine learning algorithms used here are Decision Trees, Linear regression, Support Vector Machines, XGBoost, Catboost, and Random forest. The ensemble learning technique used is Stacking and each model of the ensemble model is of deep neural network. The models are evaluated with RMSE and MSE. By observing the error values of all the algorithms, the ensemble deep neural network gave the minimum error and outperformed all the other ML algorithms.