Autonomous Software Installation using a Sequenceof Predictions from Bayesian Networks

The idea of automated installation/un-installation is a direct consequence of the tedious and time consuming manual efforts put into installing or uninstalling multiple software over hundreds of machines. In this work we propose what is to the best of our knowledge the first learnable method of autonomous software installation/un-installation. The method leverages text classification using as data textual guidelines given for users on the installation window. This is used to arrive at the Next/Pause/Abort decisions for each installation window using multiple classifier schemes. We report the best results using a full Bayesian Network with accuracy level of 94%, while Na¨ive Bayes and rule-based inference accuracy was 42% and 88%. We attribute this to the sequential nature of the Bayesian network that corresponds to the sequential nature of natural language data.