Automation of Grievance Registration using Transfer Learning

Abstract

Grievance redressal is an indispensable service but involves a lot of issues, which can be resolved if a proper automated application is introduced which involves grievance classification and location fetching mechanism. To arrive at the solution, machine learning techniques can be used, but another major facet of this application is that it should be compatible and transportable. Hence the solution needs to be in the form of a mobile application. The machine learning model must be incorporated into the mobile application. Since mobile phones have minimal computational power to run a model, an architecture which uses minimal resources must be used. MobileNet V2 is an architecture which is specially designed to incorporate Deep learning (DL) algorithm especially Image classification. MobileNet uses minimal computational resources, and interoperability is achieved through Google's Teachable machine learning, which provides a tft lite (TensorFlow Lite) model for our trained dataset and the model can be imported in to the project's asset. Location manager of android's architecture can be used to fetch the user's current latitude and longitude, which can be used by grievance redressal organization to navigate. On achieving this solution, a lot of tedious processes in our existing grievance management system can be automated. Both the public and the government can be benefited and as a result, a lot of data will be in hand which is of prominent importance now a days.